Network Working Group                                       E. Crawley
Request for Comments: 2386                              Argon Networks
Category: Informational                                        R. Nair
                                                           Arrowpoint
                                                       B. Rajagopalan
                                                              NEC USA
                                                           H. Sandick
                                                         Bay Networks
                                                          August 1998


          A Framework for QoS-based Routing in the Internet

Status of this Memo

  This memo provides information for the Internet community.  It does
  not specify an Internet standard of any kind.  Distribution of this
  memo is unlimited.

Copyright Notice

  Copyright (C) The Internet Society (1998).  All Rights Reserved.

ABSTRACT

  QoS-based routing has been recognized as a missing piece in the
  evolution of QoS-based service offerings in the Internet. This
  document describes some of the QoS-based routing issues and
  requirements, and proposes a framework for QoS-based routing in the
  Internet. This framework is based on extending the current Internet
  routing model of intra and interdomain routing to support QoS.

1. SCOPE OF  DOCUMENT & PHILOSOPHY

  This document proposes a framework for QoS-based routing, with the
  objective of fostering the development of an Internet-wide solution
  while encouraging innovations in solving the many problems that
  arise.  QoS-based routing has many complex facets and it is
  recommended that the following two-pronged approach be employed
  towards its development:

   1. Encourage the growth and evolution of novel intradomain QoS-based
      routing architectures. This is to allow the development of
      independent, innovative solutions that address the many QoS-based
      routing issues. Such solutions may be deployed in autonomous
      systems (ASs), large and small, based on their specific needs.





Crawley, et. al.             Informational                      [Page 1]

RFC 2386           A Framework for QoS-based Routing         August 1998


   2. Encourage simple, consistent and stable interactions between ASs
      implementing routing solutions developed as above.

  This approach follows the traditional separation between intra and
  interdomain routing. It allows solutions like QOSPF [GKOP98, ZSSC97],
  Integrated PNNI [IPNNI] or other schemes to be deployed for
  intradomain routing without any restriction, other than their ability
  to interact with a common, and perhaps simple, interdomain routing
  protocol. The need to develop a single, all encompassing solution to
  the complex problem of QoS-based routing is therefore obviated. As a
  practical matter, there are many different views on how QoS-based
  routing should be done. Much overall progress can be made if an
  opportunity exists for various ideas to be developed and deployed
  concurrently, while some consensus on the interdomain routing
  architecture is being developed.  Finally, this routing model is
  perhaps the most practical from an evolution point of view. It is
  superfluous to say that the eventual success of a QoS-based Internet
  routing architecture would depend on the ease of evolution.

  The aim of this document is to describe the QoS-based routing issues,
  identify basic requirements on intra and interdomain routing, and
  describe an extension of the current interdomain routing model to
  support QoS. It is not an objective of this document to specify the
  details of intradomain QoS-based routing architectures.  This is left
  up to the various intradomain routing efforts that might follow.  Nor
  is it an objective to specify the details of the interface between
  reservation protocols such as RSVP and QoS-based routing. The
  specific interface functionality needed, however, would be clear from
  the intra and interdomain routing solutions devised.  In the
  intradomain area, the goal is to develop the basic routing
  requirements while allowing maximum freedom for the development of
  solutions. In the interdomain area, the objectives are to identify
  the QoS-based routing functions, and facilitate the development or
  enhancement of a routing protocol that allows relatively simple
  interaction between domains.

  In the next section, a glossary of relevant terminology is given. In
  Section 3, the objectives of QoS-based routing are described and the
  issues that must be dealt with by QoS-based Internet routing efforts
  are outlined. In Section 4, some requirements on intradomain routing
  are defined. These requirements are purposely broad, putting few
  constraints on solution approaches. The interdomain routing model and
  issues are described in Section 5 and QoS-based multicast routing is
  discussed in Section 6.  The interaction between QoS-based routing
  and resource reservation protocols is briefly considered in Section
  7. Security considerations are listed in Section 8 and related work
  is described in Section 9. Finally, summary and conclusions are
  presented in Section 10.



Crawley, et. al.             Informational                      [Page 2]

RFC 2386           A Framework for QoS-based Routing         August 1998


2.  GLOSSARY

  The following glossary lists the terminology used in this document
  and an explanation of what is meant. Some of these terms may have
  different connotations, but when used in this document, their meaning
  is as given.

  Alternate Path Routing : A routing technique where multiple paths,
  rather than just the shortest path, between a source and a
  destination are utilized to route traffic. One of the objectives of
  alternate path routing is to distribute load among multiple paths in
  the network.

  Autonomous System (AS): A routing domain which has a common
  administrative authority and consistent internal routing policy. An
  AS may employ multiple intradomain routing protocols internally and
  interfaces to other ASs via a common interdomain routing protocol.

  Source: A host or router that can be identified by a unique unicast
  IP address.

  Unicast destination: A host or router that can be identified by a
  unique unicast IP address.

  Multicast destination: A multicast IP address indicating all hosts
  and routers that are members of the corresponding group.

  IP flow (or simply "flow"): An IP packet stream from a source to a
  destination (unicast or multicast) with an associated Quality of
  Service (QoS) (see below) and higher level demultiplexing
  information. The associated QoS could be "best-effort".

  Quality-of-Service (QoS): A set of service requirements to be met by
  the network while transporting a flow.

  Service class: The definitions of the semantics and parameters of a
  specific type of QoS.

  Integrated services:  The Integrated Services model for the Internet
  defined in RFC 1633 allows for integration of QoS services with the
  best effort services of the Internet.  The Integrated Services
  (IntServ) working group in the IETF has defined two service classes,
  Controlled Load Service [W97] and Guaranteed Service [SPG97].

  RSVP:  The ReSerVation Protocol [BZBH97].  A QoS signaling protocol
  for the Internet.

  Path: A unicast or multicast path.



Crawley, et. al.             Informational                      [Page 3]

RFC 2386           A Framework for QoS-based Routing         August 1998


  Unicast path: A sequence of links from an IP source to a unicast IP
  destination, determined by the routing scheme for forwarding packets.

  Multicast path (or Multicast Tree): A subtree of the network topology
  in which all the leaves and zero or more interior nodes are members
  of the same multicast group. A multicast path may be per-source, in
  which case the subtree is rooted at the source.

  Flow set-up: The act of establishing state in routers along a path to
  satisfy the QoS requirement of a flow.

  Crankback: A technique where a flow setup is recursively backtracked
  along the partial flow path up to the first node that can determine
  an alternative path to the destination.

  QoS-based routing: A routing mechanism under which paths for flows
  are determined based on some knowledge of resource availability in
  the network as well as the QoS requirement of flows.

  Route pinning: A mechanism to keep a flow path fixed for a duration
  of time.

  Flow Admission Control (FAC): A process by which it is determined
  whether a link or a node has sufficient resources to satisfy the QoS
  required for a flow. FAC is typically applied by each node in the
  path of a flow during flow set-up to check local resource
  availability.

  Higher-level admission control: A process by which it is determined
  whether or not a flow set-up should proceed, based on estimates and
  policy requirements of the overall resource usage by the flow.
  Higher-level admission control may result in the failure of a flow
  set-up even when FAC at each node along the flow path indicates
  resource availability.

3.  QOS-BASED ROUTING: BACKGROUND AND ISSUES

3.1  Best-Effort and QoS-Based Routing

  Routing deployed in today's Internet is focused on connectivity and
  typically supports only one type of datagram service called "best
  effort" [WC96]. Current Internet routing protocols, e.g. OSPF, RIP,
  use "shortest path routing", i.e. routing that is optimized for a
  single arbitrary metric, administrative weight or hop count. These
  routing protocols are also "opportunistic," using the current
  shortest path or route to a destination. Alternate paths with
  acceptable but non-optimal cost can not be used to route traffic
  (shortest path routing protocols do allow a router to alternate among



Crawley, et. al.             Informational                      [Page 4]

RFC 2386           A Framework for QoS-based Routing         August 1998


  several equal cost paths to a destination).

  QoS-based routing must extend the current routing paradigm in three
  basic ways.  First, to support traffic using integrated-services
  class of services, multiple paths between node pairs will have to be
  calculated. Some of these new classes of service will require the
  distribution of additional routing metrics, e.g. delay, and available
  bandwidth. If any of these metrics change frequently, routing updates
  can become more frequent thereby consuming network bandwidth and
  router CPU cycles.

  Second, today's opportunistic routing will shift traffic from one
  path to another as soon as a "better" path is found.  The traffic
  will be shifted even if the existing path can meet the service
  requirements of the existing traffic.  If routing calculation is tied
  to frequently changing consumable resources (e.g. available
  bandwidth) this change will happen more often and can introduce
  routing oscillations as traffic shifts back and forth between
  alternate paths. Furthermore, frequently changing routes can increase
  the variation in the delay and jitter experienced by the end users.

  Third, as mentioned earlier, today's optimal path routing algorithms
  do not support alternate routing.   If the best existing path cannot
  admit a new flow, the associated traffic cannot be forwarded even if
  an adequate alternate path exists.

3.2 QoS-Based Routing and Resource Reservation

  It is important to understand the difference between QoS-based
  routing and resource reservation.  While resource reservation
  protocols such as RSVP [BZBH97] provide a method for requesting and
  reserving network resources, they do not provide a mechanism for
  determining a network path that has adequate resources to accommodate
  the requested QoS.  Conversely, QoS-based routing allows the
  determination of a path that has a good chance of accommodating the
  requested QoS, but it does not include a mechanism to reserve the
  required resources.

  Consequently, QoS-based routing is usually used in conjunction with
  some form of resource reservation or resource allocation mechanism.
  Simple forms of QoS-based routing have been used in the past for Type
  of Service (TOS) routing [M98].  In the case of OSPF, a different
  shortest-path tree can be computed for each of the 8 TOS values in
  the IP header [ISI81]. Such mechanisms can be used to select
  specially provisioned paths but do not completely assure that
  resources are not overbooked along the path.  As long as strict
  resource management and control are not needed, mechanisms such as
  TOS-based routing are useful for separating whole classes of traffic



Crawley, et. al.             Informational                      [Page 5]

RFC 2386           A Framework for QoS-based Routing         August 1998


  over multiple routes.  Such mechanisms might work well with the
  emerging Differential Services efforts [BBCD98].

  Combining a resource reservation protocol with QoS-based routing
  allows fine control over the route and resources at the cost of
  additional state and setup time. For example, a protocol such as RSVP
  may be used to trigger QoS-based routing calculations to meet the
  needs of a specific flow.

3.3  QoS-Based Routing: Objectives

  Under QoS-based routing,  paths for flows would be determined based
  on some knowledge of resource availability in the network, as well as
  the QoS requirement of flows. The main objectives of QoS-based
  routing are:

  1.  Dynamic determination of feasible paths:  QoS-based routing can
      determine a path, from among possibly many choices, that has a
      good chance of accommodating the QoS of the given flow. Feasible
      path selection may be subject to policy constraints, such as path
      cost, provider selection, etc.

  2.  Optimization of resource usage: A network state-dependent QoS-
      based routing scheme can aid in the efficient utilization of
      network resources by improving the total network throughput. Such
      a routing scheme can be the basis for efficient network
      engineering.

  3.  Graceful performance degradation: State-dependent routing can
      compensate for transient inadequacies in network engineering
      (e.g., during focused overload conditions), giving better
      throughput and a more graceful performance degradation as
      compared to a state-insensitive routing scheme [A84].

  QoS-based routing in the Internet, however, raises many issues:

  -  How do routers determine the QoS capability of each outgoing link
     and reserve link resources? Note that some of these links may be
     virtual, over ATM networks and others may be broadcast multi-
     access links.

  -  What is the granularity of routing decision (i.e., destination-
     based, source and destination-based, or flow-based)?

  -  What routing metrics are used and how are QoS-accommodating paths
     computed for unicast flows?





Crawley, et. al.             Informational                      [Page 6]

RFC 2386           A Framework for QoS-based Routing         August 1998


  -  How are QoS-accommodating paths computed for multicast flows with
     different reservation styles and receiver heterogeneity?

  -  What are the performance objectives while computing QoS-based
     paths?

  -  What are the administrative control issues?

  -  What factors affect the routing overheads?, and

  -  How is scalability achieved?

  Some of these issues are discussed briefly next. Interdomain routing
  is discussed in Section 5.

3.4  QoS Determination and Resource Reservation

  To determine whether the QoS requirements of a flow can be
  accommodated on a link, a router must be able to determine the QoS
  available on the link. It is still an open issue as to how the QoS
  availability is determined for broadcast multiple access links (e.g.,
  Ethernet). A related problem is the reservation of resources over
  such links.  Solutions to these problems are just emerging [GPSS98].

  Similar problems arise when a router is connected to a large non-
  broadcast multiple access network, such as ATM. In this case, if the
  destination of a flow is outside the ATM network, the router may have
  multiple egress choices. Furthermore, the QoS availability on the ATM
  paths to each egress point may be different. The issues then are,

     o   how does a router determine all the egress choices across the
         ATM network?
     o   how  does it determine what QoS is available over the path to
         each egress point?, and
     o   what QoS value does the router advertise for the ATM link.

  Typically, IP routing over ATM (e.g., NHRP) allows the selection of a
  single egress point in the ATM network, and the procedure does not
  incorporate any knowledge of the QoS required over the path. An
  approach like I-PNNI [IPNNI] would be helpful here, although it
  introduces some complexity.

  An additional problem with resource reservation is how to determine
  what resources have already been allocated to a multicast flow. The
  availability of this information during path computation improves the
  chances of finding a path to add a new receiver to a multicast flow.
  QOSPF [ZSSC97] handles this problem by letting routers broadcast
  reserved resource information to other routers in their area.



Crawley, et. al.             Informational                      [Page 7]

RFC 2386           A Framework for QoS-based Routing         August 1998


  Alternate path routing [ZES97] deals with this issue by using probe
  messages to find a path with sufficient resources. Path QoS
  Computation (PQC) method, proposed in [GOA97], propagates bandwidth
  allocation information in RSVP PATH messages. A router receiving the
  PATH message gets an indication of the resource allocation only on
  those links in the path to itself from the source.  Allocation for
  the same flow on other remote branches of the multicast tree is not
  available. Thus, the PQC method may not be sufficient to find
  feasible QoS-accommodating paths to all receivers.

3.5  Granularity of Routing Decision

  Routing in the Internet is currently based only on the destination
  address of a packet.  Many multicast routing protocols require
  routing based on the source AND destination of a packet. The
  Integrated Services architecture and RSVP allow QoS determination for
  an individual flow between a source and a destination. This set of
  routing granularities presents a problem for QoS routing solutions.

  If routing based only on destination address is considered, then an
  intermediate router will route all flows between different sources
  and a given destination along the same path. This is acceptable if
  the path has adequate capacity but a problem arises if there are
  multiple flows to a destination that exceed the capacity of the link.

  One version of QOSPF [ZSSC97] determines QoS routes based on source
  and destination address.  This implies that all traffic between a
  given source and destination, regardless of the flow, will travel
  down the same route.  Again, the route must have capacity for all the
  QoS traffic for the source/destination pair.  The amount of routing
  state also increases since the routing tables must include
  source/destination pairs instead of just the destination.

  The best granularity is found when routing is based on individual
  flows but this incurs a tremendous cost in terms of the routing
  state.  Each QoS flow can be routed separately between any source and
  destination. PQC [GOA97] and alternate path routing [ZES97], are
  examples of solutions which operate at the flow level.

  Both source/destination and flow-based routing may be susceptible to
  packet looping under hop-by-hop forwarding. Suppose a node along a
  flow or source/destination-based path loses the state information for
  the flow.  Also suppose that the flow-based route is different from
  the regular destination-based route. The potential then exists for a
  routing loop to form when the node forwards a packet belonging to the
  flow using its destination-based routing table to a node that occurs





Crawley, et. al.             Informational                      [Page 8]

RFC 2386           A Framework for QoS-based Routing         August 1998


  earlier on the flow-based path. This is because the latter node may
  use its flow-based routing table to forward the packet again to the
  former and this can go on indefinitely.

3.6   Metrics and Path Computation

3.6.1 Metric Selection and Representation

  There are some considerations in defining suitable link and node
  metrics [WC96]. First, the metrics must represent the basic network
  properties of interest. Such metrics include residual bandwidth,
  delay and jitter.  Since the flow QoS requirements have to be mapped
  onto path metrics, the metrics define the types of QoS guarantees the
  network can support.  Alternatively, QoS-based routing cannot support
  QoS requirements that cannot be meaningfully mapped onto a reasonable
  combination of path metrics.  Second, path computation based on a
  metric or a combination of metrics must not be too complex as to
  render them impractical. In this regard, it is worthwhile to note
  that path computation based on certain combinations of metrics (e.g.,
  delay and jitter) is theoretically hard. Thus, the allowable
  combinations of metrics must be determined while taking into account
  the complexity of computing paths based on these metrics and the QoS
  needs of flows. A common strategy to allow flexible combinations of
  metrics while at the same time reduce the path computation complexity
  is to utilize "sequential filtering". Under this approach, a
  combination of metrics is ordered in some fashion, reflecting the
  importance of different metrics (e.g., cost followed by delay, etc.).
  Paths based on the primary metric are computed first (using a simple
  algorithm, e.g., shortest path) and a subset of them are eliminated
  based on the secondary metric and so forth until a single path is
  found. This is an approximation technique and it trades off global
  optimality for path computation simplicity (The filtering technique
  may be simpler, depending on the set of metrics used. For example,
  with bandwidth and cost as metrics, it is possible to first eliminate
  the set of links that do not have the requested bandwidth and then
  compute the least cost path using the remaining links.)

  Now, once suitable link and node metrics are defined, a uniform
  representation of them is required across independent domains -
  employing possibly different routing schemes - in order to derive
  path metrics consistently (path metrics are obtained by the
  composition of link and node metrics). Encoding of the maximum,
  minimum, range, and granularity of the metrics are needed. Also, the
  definitions of comparison and accumulation operators are required. In
  addition, suitable triggers must be defined for indicating a
  significant change from a minor change.  The former will cause a
  routing update to be generated. The stability of the QoS routes would




Crawley, et. al.             Informational                      [Page 9]

RFC 2386           A Framework for QoS-based Routing         August 1998


  depend on the ability to control the generation of updates. With
  interdomain routing, it is essential to obtain a fairly stable view
  of the interconnection among the ASs.

3.6.2  Metric Hierarchy

  A hierarchy can be defined among various classes of service based on
  the degree to which traffic from one class can potentially degrade
  service of traffic from lower classes that traverse the same link. In
  this hierarchy, guaranteed constant bit rate traffic is at the top
  and "best-effort" datagram traffic at the bottom.  Classes providing
  service higher in the hierarchy impact classes providing service in
  lower levels. The same situation is not true in the other direction.
  For example, a datagram flow cannot affect a real-time service. Thus,
  it may be necessary to distribute and update different metrics for
  each type of service in the worst case.  But, several advantages
  result by identifying a single default metric.  For example, one
  could derive a single metric combining the availability of datagram
  and real-time service over a common substrate.

3.6.3  Datagram Flows

  A delay-sensitive metric is probably the most obvious type of metric
  suitable for datagram flows. However, it requires careful analysis to
  avoid instabilities and to reduce storage and bandwidth requirements.
  For example, a recursive filtering technique based on a simple and
  efficient weighted averaging algorithm [NC94] could be used. This
  filter is used to stabilize the metric. While it is adequate for
  smoothing most loading patterns, it will not distinguish between
  patterns consisting of regular bursts of traffic and random loading.
  Among other stabilizing tools, is a minimum time between updates that
  can help filter out high-frequency oscillations.

3.6.4 Real-time Flows

  In real-time quality-of-service, delay variation is generally more
  critical than delay as long as the delay is not too high.  Clearly,
  voice-based applications cannot tolerate more than a certain level of
  delay. The condition of varying delays may be expected to a greater
  degree in a shared medium environment with datagrams, than in a
  network implemented over a switched substrate.  Routing a real-time
  flow therefore reduces to an exercise in allocating the required
  network resources while minimizing fragmentation of bandwidth. The
  resulting situation is a bandwidth-limited minimum hop path from a
  source to the destination.  In other words, the router performs an
  ordered search through paths of increasing hop count until it finds
  one that meets all the bandwidth needs of the flow. To reduce
  contention and the probability of false probes (due to inaccuracy in



Crawley, et. al.             Informational                     [Page 10]

RFC 2386           A Framework for QoS-based Routing         August 1998


  route tables), the router could select a path randomly from a
  "window" of paths which meet the needs of the flow and satisfy one of
  three additional criteria: best-fit, first-fit or worst-fit. Note
  that there is a similarity between the allocation of bandwidth and
  the allocation of memory in a multiprocessing system. First-fit seems
  to be appropriate for a system with a high real-time flow arrival
  rates; and worst-fit is ideal for real-time flows with high holding
  times.  This rather nonintuitive result was shown in [NC94].

3.6.5  Path Properties

  Path computation by itself is merely a search technique, e.g.,
  Shortest Path First (SPF) is a search technique based on dynamic
  programming. The usefulness of the paths computed depends to a large
  extent on the metrics used in evaluating the cost of a path with
  respect to a flow.

  Each link considered by the path computation engine must be evaluated
  against the requirements of the flow, i.e., the cost of providing the
  services required by the flow must be estimated with respect to the
  capabilities of the link. This requires a uniform method of combining
  features such as delay, bandwidth, priority and other service
  features.  Furthermore, the costs must reflect the lost opportunity
  of using each link after routing the flow.

3.6.6  Performance Objectives

  One common objective during path computation is to improve the total
  network throughput.  In this regard, merely routing a flow on any
  path that accommodates its QoS requirement is not a good strategy. In
  fact, this corresponds to uncontrolled alternate routing [SD95] and
  may adversely impact performance at higher traffic loads.  It is
  therefore necessary to consider the total resource allocation for a
  flow along a path, in relation to available resources, to determine
  whether or not the flow should be routed on the path.  Such a
  mechanism is referred to in this document as "higher level admission
  control". The goal of this is to ensure that the "cost" incurred by
  the network in routing a flow with a given QoS is never more than the
  revenue gained.  The routing cost in this regard may be the lost
  revenue in potentially blocking other flows that contend for the same
  resources. The formulation of the higher level admission control
  strategy, with suitable administrative hooks and with fairness to all
  flows desiring entry to the network, is an issue.  The fairness
  problem arises because flows with smaller reservations tend to be
  more successfully routed than flows with large reservations, for a
  given engineered capacity.  To guarantee a certain level of





Crawley, et. al.             Informational                     [Page 11]

RFC 2386           A Framework for QoS-based Routing         August 1998


  acceptance rate for "larger" flows, without over-engineering the
  network, requires a fair higher level admission control mechanism.
  The application of higher level admission control to multicast
  routing is discussed later.

3.7   Administrative Control

  There are several administrative control issues. First, within an AS
  employing state-dependent routing, administrative control of routing
  behavior may be necessary. One example discussed earlier was higher
  level admission control. Some others are described in this section.
  Second, the control of interdomain routing based on policy is an
  issue.  The discussion of interdomain routing is defered to Section
  5.

  Two areas that need administrative control, in addition to
  appropriate routing mechanisms, are handling flow priority with
  preemption, and resource allocation for multiple service classes.

3.7.1  Flow Priorities and Preemption

  If there are critical flows that must be accorded higher priority
  than other types of flows, a mechanism must be implemented in the
  network to recognize flow priorities. There are two aspects to
  prioritizing flows.  First, there must be a policy to decide how
  different users are allowed to set priorities for flows they
  originate. The network must be able to verify that a given flow is
  allowed to claim a priority level signaled for it. Second, the
  routing scheme must ensure that a path with the requested QoS will be
  found for a flow with a probability that increases with the priority
  of the flow. In other words, for a given network load, a high
  priority flow should be more likely to get a certain QoS from the
  network than a lower priority flow requesting the same QoS. Routing
  procedures for flow prioritization can be complex.  Identification
  and evaluation of different procedures are areas that require
  investigation.

3.7.2 Resource Control

  If there are multiple service classes, it is necessary to engineer a
  network to carry the forecasted traffic demands of each class. To do
  this, router and link resources may be logically partitioned among
  various service classes. It is desirable to have dynamic partitioning
  whereby unused resources in various partitions are dynamically
  shifted to other partitions on demand [ACFH92]. Dynamic sharing,
  however, must be done in a controlled  fashion in order to prevent
  traffic under some service class from taking up more resources than




Crawley, et. al.             Informational                     [Page 12]

RFC 2386           A Framework for QoS-based Routing         August 1998


  what was engineered for it for prolonged periods of time. The design
  of such a resource sharing scheme, and its incorporation into the
  QoS-based routing scheme are significant issues.

3.8   QoS-Based Routing for Multicast Flows

  QoS-based multicast routing is an important problem, especially if
  the notion of higher level admission control is included. The
  dynamism in the receiver set allowed by IP multicast, and receiver
  heterogeneity add to the problem. With straightforward implementation
  of distributed heuristic algorithms for multicast path computation
  [W88, C91], the difficulty is essentially one of scalability. To
  accommodate QoS, multicast path computation at a router must have
  knowledge of not only the id of subnets where group members are
  present, but also the identity of branches in the existing tree. In
  other words, routers must keep flow-specific state information. Also,
  computing optimal shared trees based on the shared reservation style
  [BZBH97], may require new algorithms.  Multicast routing is discussed
  in some detail in Section 6.

3.9    Routing Overheads

  The overheads incurred by a routing scheme depend on the type of the
  routing scheme, as well as the implementation. There are three types
  of overheads to be considered: computation, storage and
  communication. It is necessary to understand the implications of
  choosing a routing mechanism in terms of these overheads.

  For example, considering link state routing, the choice of the update
  propagation mechanism is important since network state is dynamic and
  changes relatively frequently. Specifically, a flooding mechanism
  would result in many unnecessary message transmissions and
  processing.  Alternative techniques, such as tree-based forwarding
  [R96], have to be considered. A related issue is the quantization of
  state information to prevent frequent updating of dynamic state.
  While coarse quantization reduces updating overheads, it may affect
  the performance of the routing scheme.  The tradeoff has to be
  carefully evaluated.  QoS-based routing incurs certain overheads
  during flow establishment, for example, computing a source route.
  Whether this overhead is disproportionate compared to the length of
  the sessions is an issue. In general, techniques for the minimization
  of routing-related overheads during flow establishment must be
  investigated. Approaches that are useful include pre-computation of
  routes, caching recently used routes, and TOS routing based on hints
  in packets (e.g., the TOS field).






Crawley, et. al.             Informational                     [Page 13]

RFC 2386           A Framework for QoS-based Routing         August 1998


3.10   Scaling by Hierarchical Aggregation

  QoS-based routing should be scalable, and hierarchical aggregation is
  a common technique for scaling (e.g., [PNNI96]). But this introduces
  problems with regard to the accuracy of the aggregated state
  information [L95]. Also, the aggregation of paths under multiple
  constraints is difficult. One of the difficulties is the risk of
  accepting a flow based on inaccurate information, but not being able
  to support the QoS requirements of flow because the capabilities of
  the actual paths that are aggregated are not known during route
  computation.  Performance impacts of aggregating path metric
  information must therefore be understood. A way to compensate for
  inaccuracies is to use crankback, i.e., dynamic search for alternate
  paths as a flow is being routed. But crankback increases the time to
  set up a flow, and may adversely affect the performance of the
  routing scheme under some circumstances. Thus, crankback must be used
  judiciously, if at all, along with a higher level admission control
  mechanism.

4. INTRADOMAIN ROUTING REQUIREMENTS

  At the intradomain level, the objective is to allow as much latitude
  as possible in addressing the QoS-based routing issues. Indeed, there
  are many ideas about how QoS-based routing services can be
  provisioned within ASs. These range from on-demand path computation
  based on current state information, to statically provisioned paths
  supporting a few service classes.

  Another aspect that might invite differing solutions is performance
  optimization. Based on the technique used for this, intradomain
  routing could be very sophisticated or rather simple. Finally, the
  service classes supported, as well as the specific QoS engineered for
  a service class, could differ from AS to AS. For instance, some ASs
  may not support guaranteed service, while others may. Also, some ASs
  supporting the service may be engineered for a better delay bound
  than others. Thus, it requires considerable thought to determine the
  high level requirements for intradomain routing that both supports
  the overall view of QoS-based routing in the Internet and allows
  maximum autonomy in developing solutions.

  Our view is that certain minimum requirements must be satisfied by
  intradomain routing in order to be qualified as "QoS-based" routing.
  These are:

  - The routing scheme must route a flow along a path that can
    accommodate its QoS requirements, or indicate that the flow cannot
    be admitted with the QoS currently being requested.




Crawley, et. al.             Informational                     [Page 14]

RFC 2386           A Framework for QoS-based Routing         August 1998


  - The routing scheme must indicate disruptions to the current route
    of a flow due to topological changes.

  - The routing scheme must accommodate best-effort flows without any
    resource reservation requirements. That is, present best effort
    applications and protocol stacks need not have to change to run in
    a domain employing QoS-based routing.

  - The routing scheme may optionally support QoS-based multicasting
    with receiver heterogeneity and shared reservation styles.

  In addition, the following capabilities are also recommended:

  - Capabilities to optimize resource usage.

  - Implementation of higher level admission control procedures to
    limit the overall resource utilization by individual flows.

  Further requirements along these lines may be specified. The
  requirements should capture the consensus view of QoS-based routing,
  but should not preclude particular approaches (e.g., TOS-based
  routing) from being implemented. Thus, the intradomain requirements
  are expected to be rather broad.

5. INTERDOMAIN ROUTING

  The fundamental requirement on interdomain QoS-based routing is
  scalability.  This implies that interdomain routing cannot be based
  on highly dynamic network state information. Rather, such routing
  must be aided by sound network engineering and relatively sparse
  information exchange between independent routing domains. This
  approach has the advantage that it can be realized by straightforward
  extensions of the present Internet interdomain routing model. A
  number of issues, however, need to be addressed to achieve this, as
  discussed below.
















Crawley, et. al.             Informational                     [Page 15]

RFC 2386           A Framework for QoS-based Routing         August 1998


5.1 Interdomain QoS-Based Routing Model

  The interdomain QoS-based routing model is depicted below:

         AS1                   AS2             AS3
     ___________        _____________      ____________
    |           |      |             |    |            |
    |           B------B             B----B            |
    |           |      |             |    |            |
     -----B-----       B-------------      --B---------
           \         /                      /
            \       /                      /
         ____B_____B____         _________B______
        |               |       |                |
        |               B-------B                |
        |               |       |                |
        |               B-------B                |
         ---------------         ----------------
              AS4                           AS5

  Here, ASs exchange standardized routing information via border nodes
  B.  Under this model, each AS can itself consist of a set of
  interconnected ASs, with standardized routing interaction. Thus, the
  interdomain routing model is hierarchical.  Also, each lowest level
  AS employs an intradomain QoS-based routing scheme (proprietary or
  standardized by intradomain routing efforts such as QOSPF). Given
  this structure, some questions that arise are:

  - What information is exchanged between ASs?

  - What routing capabilities does the information exchange lead to?
    (E.g., source routing, on-demand path computation, etc.)

  - How is the external routing information represented within an AS?

  - How are interdomain paths computed?

  - What sort of policy controls may be exerted on interdomain path
    computation and flow routing?, and

  - How is interdomain QoS-based multicast routing accomplished?

  At a high level, the answers to these questions depend on the routing
  paradigm. Specifically, considering link state routing, the
  information exchanged between domains would consist of an abstract
  representation of the domains in the form of logical nodes and links,
  along with metrics that quantify their properties and resource
  availability.  The hierarchical structure of the ASs may be handled



Crawley, et. al.             Informational                     [Page 16]

RFC 2386           A Framework for QoS-based Routing         August 1998


  by a hierarchical link state representation, with appropriate metric
  aggregation.

  Link state routing may not necessarily be advantageous for
  interdomain routing for the following reasons:

  - One advantage of intradomain link state routing is that it would
    allow fairly detailed link state information be used to compute
    paths on demand for flows requiring QoS. The state and metric
    aggregation used in interdomain routing, on the other hand, erodes
    this property to a great degree.

  - The usefulness of keeping track of the abstract topology and
    metrics of a remote domain, or the interconnection between remote
    domains is not obvious. This is especially the case when the remote
    topology and metric encoding are lossy.

  - ASs may not want to advertise any details of their internal
    topology or resource availability.

  - Scalability in interdomain routing can be achieved only if
    information exchange between domains is relatively infrequent.
    Thus, it seems practical to limit information flow between domains
    as much as possible.

  Compact information flow allows the implementation QoS-enhanced
  versions of existing interdomain protocols such as BGP-4. We look at
  the interdomain routing issues in this context.

5.2  Interdomain Information Flow

  The information flow between routing domains must enable certain
  basic functions:

  1.  Determination of reachability to various destinations

  2.  Loop-free flow routes

  3.  Address aggregation whenever possible

  4.  Determination of the QoS that will be supported on the path to a
      destination. The QoS information should be relatively static,
      determined from the engineered topology and capacity of an AS
      rather than ephemeral fluctuations in traffic load through the
      AS. Ideally, the QoS supported in a transit AS should be allowed
      to vary significantly only under exceptional circumstances, such
      as failures or focused overload.




Crawley, et. al.             Informational                     [Page 17]

RFC 2386           A Framework for QoS-based Routing         August 1998


  5.  Determination, optionally, of multiple paths for a given
      destination, based on service classes.

  6.  Expression of routing policies, including monetary cost, as a
      function of flow parameters, usage and administrative factors.

  Items 1-3 are already part of existing interdomain routing. Item 5 is
  also a straightfoward extension of the current model. The main
  problem areas are therefore items 4 and 6.

  The QoS of an end-to-end path is obtained by composing the QoS
  available in each transit AS.  Thus, border routers must first
  determine what the locally available QoS is in order to advertise
  routes to both internal and external destinations. The determination
  of local "AS metrics" (corresponding to link metrics in the
  intradomain case) should not be subject to too much dynamism. Thus,
  the issue is how to define such metrics and what triggers an
  occasional change that results in re-advertisements of routes.

  The approach suggested in this document is not to compute paths based
  on residual or instantaneous values of AS metics (which can be
  dynamic), but utilize only the QoS capabilities engineered for
  aggregate transit flows.  Such engineering may be based on the
  knowledge of traffic to be expected from each neighboring ASs and the
  corresponding QOS needs.  This information may be obtained based on
  contracts agreed upon prior to the provisioning of services. The AS
  metric then corresponds to the QoS capabilities of the "virtual path"
  engineered through the AS (for transit traffic) and a different
  metric may be used for different neighbors. This is illustrated in
  the following figure.

         AS1                   AS2             AS3
     ___________        _____________      ____________
    |           |      |             |    |            |
    |           B------B1           B2----B            |
    |           |      |             |    |            |
     -----B-----       B3------------      --B---------
           \         /
            \       /
         ____B_____B____
        |               |
        |               |
        |               |
        |               |
         ---------------
              AS4





Crawley, et. al.             Informational                     [Page 18]

RFC 2386           A Framework for QoS-based Routing         August 1998


  Here, B1 may utilize an AS metric specific for AS1 when computing
  path metrics to be  advertised to AS1. This metric is based on the
  resources engineered in AS2 for transit traffic from AS1. Similarly,
  B3 may utilize a different metric when computing path metrics to be
  advertised to AS4.  Now, it is assumed that as long as traffic flow
  into AS2 from AS1 or AS4 does not exceed the engineered values, these
  path metrics would hold.  Excess traffic due to transient
  fluctuations, however, may be handled as best effort or marked with a
  discard bit.

  Thus, this model is different from the intradomain model, where end
  nodes pick a path dynamically based on the QoS needs of the flow to
  be routed.  Here, paths within ASs are engineered based on presumed,
  measured or declared traffic and QoS requirements. Under this model,
  an AS can contract for routes via multiple transit ASs with different
  QoS requirements. For instance, AS4 above can use both AS1 and AS2 as
  transits for same or different destinations. Also, a QoS contract
  between one AS and another may generate another contract between the
  second and a third AS and so forth.

  An issue is what triggers the recomputation of path metrics within an
  AS.  Failures or other events that prevent engineered resource
  allocation should certainly trigger recomputation. Recomputation
  should not be triggered in response to arrival of flows within the
  engineered limit.

5.3   Path Computation

  Path computation for an external destination at a border node is
  based on reachability, path metrics and local policies of selection.
  If there are multiple selection criteria (e.g., delay, bandwidth,
  cost, etc.), mutiple alternaives may have to be maintained as well as
  propagated by border nodes. Selection of a path from among many
  alternatives would depend on the QoS requests of flows, as well as
  policies. Path computation may also utilze any heuristics for
  optimizing resource usage.

5.4  Flow Aggregation

  An important issue in interdomain routing is the amount of flow state
  to be processed by transit ASs. Reducing the flow state by
  aggregation techniques must therefore be seriously considered. Flow
  aggregation means that transit traffic through an AS is classified
  into a few aggregated streams rather than being routed at the
  individual flow level. For example, an entry border router may
  classify various transit flows entering an AS into a few coarse
  categories, based on the egress node and QoS requirements of the
  flows.  Then, the aggregated stream for a given traffic class may be



Crawley, et. al.             Informational                     [Page 19]

RFC 2386           A Framework for QoS-based Routing         August 1998


  routed as a single flow inside the AS to the exit border router. This
  router may then present individual flows to different neighboring ASs
  and the process repeats at each entry border router. Under this
  scenario, it is essential that entry border routers keep track of the
  resource requirements for each transit flow and apply admission
  control to determine whether the aggregate requirement from any
  neighbor exceeds the engineered limit. If so, some policy must be
  invoked to deal with the excess traffic. Otherwise, it may be assumed
  that aggregated flows are routed over paths that have adequate
  resources to guarantee QoS for the member flows. Finally, it is
  possible that entry border routers at a transit AS may prefer not to
  aggregate flows if finer grain routing within the AS may be more
  efficient (e.g., to aid load balancing within the AS).

5.5   Path Cost Determination

  It is hoped that the integrated services Internet architecture would
  allow providers to charge for IP flows based on their QoS
  requirements.  A QoS-based routing architecture can aid in
  distributing information on expected costs of routing flows to
  various destinations via different domains. Clearly, from a
  provider's point of view, there is a cost incurred in guaranteeing
  QoS to flows.  This cost could be a function of several parameters,
  some related to flow parameters, others based on policy. From a
  user's point of view, the consequence of requesting a particular QoS
  for a flow is the cost incurred, and hence the selection of providers
  may be based on cost. A routing scheme can aid a provider in
  distributing the costs in routing to various destinations, as a
  function of several parameters, to other providers or to end users.
  In the interdomain routing model described earlier, the costs to a
  destination will change as routing updates are passed through a
  transit domain. One of the goals of the routing scheme should be to
  maintain a uniform semantics for cost values (or functions) as they
  are handled by intermediate domains. As an example, consider the cost
  function generated by border node B1 in domain A and passed to node
  B2 in domain B below.  The routing update may be injected into domain
  B by B2 and finally passed to B4 in domain C by router B3. Domain B
  may interpret the cost value received from domain A in any way it
  wants, for instance, adding a locally significant component to it.
  But when this cost value is passed to domain C, the meaning of it
  must be what domain A intended, plus the incremental cost of
  transiting domain B, but not what domain B uses internally.









Crawley, et. al.             Informational                     [Page 20]

RFC 2386           A Framework for QoS-based Routing         August 1998


   Domain A                    Domain B           Domain C
    ____________          ___________      ____________
   |            |        |           |    |            |
   |            B1------B2          B3---B4            |
   |            |        |           |    |            |
    ------------          -----------      ------------

  A problem with charging for a flow is the determination of the cost
  when the QoS promised for the flow was not actually delivered.
  Clearly, when a flow is routed via multiple domains, it must be
  determined whether each domain delivers the QoS it declares possible
  for traffic through it.

6. QOS-BASED MULTICAST ROUTING

  The goals of QoS-based multicast routing are as follows:

  - Scalability to large groups with dynamic membership

  - Robustness in the presence of topological changes

  - Support for receiver-initiated, heterogeneous reservations

  - Support for shared reservation styles, and

  - Support for "global" admission control, i.e., administrative
    control of resource consumption by the multicast flow.

  The RSVP multicast flow model is as follows. The sender of a
  multicast flow advertises the traffic characteristics periodically to
  the receivers.  On receipt of an advertisement, a receiver may
  generate a message to reserve resources along the flow path from the
  sender. Receiver reservations may be heterogeneous. Other multicast
  models may be considered.

  The multicast routing scheme attempts to determine a path from the
  sender to each receiver that can accommodate the requested
  reservation.  The routing scheme may attempt to maximize network
  resource utilization by minimizing the total bandwidth allocated to
  the multicast flow, or by optimizing some other measure.

6.1   Scalability, Robustness and Heterogeneity

  When addressing scalability, two aspects must be considered:

    1.  The overheads associated with receiver discovery. This overhead
        is incurred when determining the multicast tree for forwarding
        best-effort sender traffic characterization to receivers.



Crawley, et. al.             Informational                     [Page 21]

RFC 2386           A Framework for QoS-based Routing         August 1998


    2.  The overheads associated with QoS-based multicast path
        computation.  This overhead is incurred when flow-specific
        state information has to be collected by a router to determine
        QoS-accommodating paths to a receiver.

  Depending on the multicast routing scheme, one or both of these
  aspects become important. For instance, under the present RSVP model,
  reservations are established on the same path over which sender
  traffic characterizations are sent, and hence there is no path
  computation overhead. On the other hand, under the proposed QOSPF
  model [ZSSC97] of multicast source routing, receiver discovery
  overheads are incurred by MOSPF [M94] receiver location broadcasts,
  and additional path computation overheads are incurred due to the
  need to keep track of existing flow paths. Scaling of QoS-based
  multicast depends on both these scaling issues. However, scalable
  best-effort multicasting is really not in the domain of QoS-based
  routing work (solutions for this are being devised by the IDMR WG
  [BCF94, DEFV94]). QoS-based multicast routing may build on these
  solutions to achieve overall scalability.

  There are several options for QoS-based multicast routing. Multicast
  source routing is one under which multicast trees are computed by the
  first-hop router from the source, based on sender traffic
  advertisements.  The advantage of this is that it blends nicely with
  the present RSVP signaling model. Also, this scheme works well when
  receiver reservations are homogeneous and the same as the maximum
  reservation derived from sender advertisement.  The disadvantages of
  this scheme are the extra effort needed to accommodate heterogeneous
  reservations and the difficulties in optimizing resource allocation
  based on shared reservations.

  In these regards, a receiver-oriented multicast routing model seems
  to have some advantage over multicast source routing. Under this
  model:

    1.  Sender traffic advertisements are multicast over a best-effort
        tree which can be different from the QoS-accommodating tree for
        sender data.

    2.  Receiver discovery overheads are minimized by utilizing a
        scalable scheme (e.g., PIM, CBT), to multicast sender traffic
        characterization.

    3.  Each receiver-side router independently computes a QoS-
        accommodating path from the source, based on the receiver
        reservation. This path can be computed based on unicast routing
        information only, or with additional multicast flow-specific
        state information. In any case, multicast path computation is



Crawley, et. al.             Informational                     [Page 22]

RFC 2386           A Framework for QoS-based Routing         August 1998


        broken up into multiple, concurrent nunicast path computations.

    4.  Routers processing unicast reserve messages from receivers
        aggregate resource reservations from multiple receivers.

  Flow-specific state information may be limited in Step 3 to achieve
  scalability [RN98]. In general, limiting flow-specific information in
  making multicast routing decisions is important in any routing model.
  The advantages of this model are the ease with which heterogeneous
  reservations can be accommodated, and the ability to handle shared
  reservations. The disadvantages are the incompatibility with the
  present RSVP signaling model, and the need to rely on reverse paths
  when link state routing is not used. Both multicast source routing
  and the receiver-oriented routing model described above utilize per-
  source trees to route multicast flows. Another possibility is the
  utilization of shared, per-group trees for routing flows. The
  computation and usage of such trees require further work.

  Finally, scalability at the interdomain level may be achieved if
  QoS-based multicast paths are computed independently in each domain.
  This principle is illustrated by the QOSPF multicast source routing
  scheme which allows independent path computation in different OSPF
  areas. It is easy to incorporate this idea in the receiver-oriented
  model also. An evaluation of multicast routing strategies must take
  into account the relative advantages and disadvantages of various
  approaches, in terms of scalability features and functionality
  supported.

6.2    Multicast Admission Control

  Higher level admission control, as defined for unicast, prevents
  excessive resource consumption by flows when traffic load is high.
  Such an admission control strategy must be applied to multicast flows
  when the flow path computation is receiver-oriented or sender-
  oriented. In essence, a router computing a path for a receiver must
  determine whether the incremental resource allocation for the
  receiver is excessive under some administratively determined
  admission control policy. Other admission control criteria, based on
  the total resource consumption of a tree may be defined.

7.    QOS-BASED ROUTING AND RESOURCE RESERVATION PROTOCOLS

  There must clearly be a well-defined interface between routing and
  resource reservation protocols. The nature of this interface, and the
  interaction between routing and resource reservation has to be
  determined carefully to avoid incompatibilities. The importance of
  this can be readily illustrated in the case of RSVP.




Crawley, et. al.             Informational                     [Page 23]

RFC 2386           A Framework for QoS-based Routing         August 1998


  RSVP has been designed to operate independent of the underlying
  routing scheme. Under this model, RSVP PATH messages establish the
  reverse path for RESV messages.  In essence, this model is not
  compatible with QoS-based routing schemes that compute paths after
  receiver reservations are received. While this incompatibility can be
  resolved in a simple manner for unicast flows, multicast with
  heterogeneous receiver requirements is a more difficult case.  For
  this, reconciliation between RSVP and QoS-based routing models is
  necessary. Such a reconciliation, however, may require some changes
  to the RSVP model depending on the QoS-based routing model [ZES97,
  ZSSC97, GOA97]. On the other hand, QoS-based routing schemes may be
  designed with RSVP compatibility as a necessary goal. How this
  affects scalability and other performance measures must be
  considered.

8. SECURITY CONSIDERATIONS

  Security issues that arise with routing in general are about
  maintaining the integrity of the routing protocol in the presence of
  unintentional or malicious introduction of information that may lead
  to protocol failure [P88]. QoS-based routing requires additional
  security measures both to validate QoS requests for flows and to
  prevent resource-depletion type of threats that can arise when flows
  are allowed to make arbitratry resource requests along various paths
  in the network. Excessive resource consumption by an errant flow
  results in denial of resources to legitimate flows. While these
  situations may be prevented by setting up proper policy constraints,
  charging models and policing at various points in the network, the
  formalization of such protection requires work [BCCH94].

9. RELATED WORK

  "Adaptive" routing, based on network state, has a long history,
  especially in circuit-switched networks. Such routing has also been
  implemented in early datagram and virtual circuit packet networks.
  More recently, this type of routing has been the subject of study in
  the context of ATM networks, where the traffic characteristics and
  topology are substantially different from those of circuit-switched
  networks [MMR96]. It is instructive to review the adaptive routing
  methodologies, both to understand the problems encountered and
  possible solutions.

  Fundamentally, there are two aspects to adaptive, network state-
  dependent routing:

    1.  Measuring and gathering network state information, and
    2.  Computing routes based on the available information.




Crawley, et. al.             Informational                     [Page 24]

RFC 2386           A Framework for QoS-based Routing         August 1998


  Depending on how these two steps are implemented, a variety of
  routing techniques are possible. These differ in the following
  respects:

  -  what state information is used
  -  whether local or global state is used
  -  what triggers the propagation of state information
  -  whether routes are computed in a distributed or centralized manner
  -  whether routes are computed on-demand, pre-computed, or in a
     hybrid manner
  -  what optimization criteria, if any, are used in computing routes
  -  whether source routing or hop by hop routing is used, and
  -  how alternate route choices are explored

  It should be noted that most of the adaptive routing work has focused
  on unicast routing. Multicast routing is one of the areas that would
  be prominent with Internet QoS-based routing. We treat this
  separately, and the following review considers only unicast routing.
  This review is not exhaustive, but gives a brief overview of some of
  the approaches.

9.1 Optimization Criteria

  The most common optimization criteria used in adaptive routing is
  throughput maximization or delay minimization. A general formulation
  of the optimization problem is the one in which the network revenue
  is maximized, given that there is a cost associated with routing a
  flow over a given path [MMR96, K88]. In general, global optimization
  solutions are difficult to implement, and they rely on a number of
  assumptions on the characteristics of the traffic being routed
  [MMR96]. Thus, the practical approach has been to treat the routing
  of each flow (VC, circuit or packet stream to a given destination)
  independently of the routing of other flows. Many such routing
  schemes have been implemented.

9.2  Circuit Switched Networks

  Many adaptive routing concepts have been proposed for circuit-
  switched networks. An example of a simple adaptive routing scheme is
  sequential alternate routing [T88]. This is a hop-by-hop
  destination-based routing scheme where only local state information
  is utilized.  Under this scheme, a routing table is computed for each
  node, which lists multiple output link choices for each destination.
  When a call set-up request is received by a node, it tries each
  output link choice in sequence, until it finds one that can
  accommodate the call. Resources are reserved on this link, and the
  call set-up is forwarded to the next node. The set-up either reaches
  the destination, or is blocked at some node. In the latter case, the



Crawley, et. al.             Informational                     [Page 25]

RFC 2386           A Framework for QoS-based Routing         August 1998


  set-up can be cranked back to the previous node or a failure
  declared. Crankback allows the previous node to try an alternate
  path.  The routing table under this scheme can be computed in a
  centralized or distributed manner, based only on the topology of the
  network. For instance, a k-shortest-path algorithm can be used to
  determine k alternate paths from a node with distinct initial links
  [T88]. Some mechanism must be implemented during path computation or
  call set-up to prevent looping.

  Performance studies of this scheme illustrate some of the pitfalls of
  alternate routing in general, and crankback in particular [A84, M86,
  YS87]. Specifically, alternate routing improves the throughput when
  traffic load is relatively light, but adversely affects the
  performance when traffic load is heavy. Crankback could further
  degrade the performance under these conditions. In general,
  uncontrolled alternate routing (with or without crankback) can be
  harmful in a heavily utilized network, since circuits tend to be
  routed along longer paths thereby utilizing more capacity. This is an
  obvious, but important result that applies to QoS-based Internet
  routing also.

  The problem with alternate routing is that both direct routed (i.e.,
  over shortest paths) and alternate routed calls compete for the same
  resource.  At higher loads, allocating these resources to alternate
  routed calls result in the displacement of direct routed calls and
  hence the alternate routing of these calls. Therefore, many
  approaches have been proposed to limit the flow of alternate routed
  calls under high traffic loads. These schemes are designed for the
  fully-connected logical topology of long distance telephone networks
  (i.e., there is a logical link between every pair of nodes). In this
  topology, direct routed calls always traverse a 1-hop path to the
  destination and alternate routed calls traverse at most a 2-hop path.

  "Trunk reservation" is a scheme whereby on each link a certain
  bandwidth is reserved for direct routed calls [MS91]. Alternate
  routed calls are allowed on a trunk as long as the remaining trunk
  bandwidth is greater than the reserved capacity. Thus, alternate
  routed calls cannot totally displace direct routed calls on a trunk.
  This strategy has been shown to be very effective in preventing the
  adverse effects of alternate routing.

  "Dynamic alternate routing" (DAR) is a strategy whereby alternate
  routing is controlled by limiting the number of choices, in addition
  to trunk reservation [MS91]. Under DAR, the source first attempts to
  use the direct link to the destination. When blocked, the source
  attempts to alternate route the call via a pre-selected neighbor. If
  the call is still blocked, a different neighbor is selected for
  alternate routing to this destination in the future. The present call



Crawley, et. al.             Informational                     [Page 26]

RFC 2386           A Framework for QoS-based Routing         August 1998


  is dropped. DAR thus requires only local state information. Also, it
  "learns" of good alternate paths by random sampling and sticks to
  them as long as possible.

  More recent circuit-switched routing schemes utilize global state to
  select routes for calls. An example is AT&T's Real-Time Network
  Routing (RTNR) scheme [ACFH92]. Unlike schemes like DAR, RTNR handles
  multiple classes of service, including voice and data at fixed rates.
  RTNR utilizes a sophisticated per-class trunk reservation mechanism
  with dynamic bandwidth sharing between classes. Also, when alternate
  routing a call, RTNR utilizes the loading on all trunks in the
  network to select a path. Because of the fully-connected topology,
  disseminating status information is simple under RTNR; each node
  simply exchanges status information directly with all others.

  From the point of view of designing QoS-based Internet routing
  schemes, there is much to be learned from circuit-switched routing.
  For example, alternate routing and its control, and dynamic resource
  sharing among different classes of traffic. It is, however, not
  simple to apply some of the results to a general topology network
  with heterogeneous multirate traffic. Work in the area of ATM network
  routing described next illustrates this.

9.3 ATM Networks

  The VC routing problem in ATM networks presents issues similar to
  that encountered in circuit-switched networks. Not surprisingly, some
  extensions of circuit-switched routing have been proposed. The goal
  of these routing schemes is to achieve higher throughput as compared
  to traditional shortest-path routing. The flows considered usually
  have a single QoS requirement, i.e., bandwidth.

  The first idea is to extend alternate routing with trunk reservation
  to general topologies [SD95].  Under this scheme, a distance vector
  routing protocol is used to build routing tables at each node with
  multiple choices of increasing hop count to each destination. A VC
  set-up is first routed along the primary ("direct") path. If
  sufficient resources are not available along this path, alternate
  paths are tried in the order of increasing hop count. A flag in the
  VC set-up message indicates primary or alternate routing, and
  bandwidth on links along an alternate path is allocated subject to
  trunk reservation. The trunk reservation values are determined based
  on some assumptions on traffic characteristics. Because the scheme
  works only for a single data rate, the practical utility of it is
  limited.

  The next idea is to import the notion of controlled alternate routing
  into traditional link state QoS-based routing [GKR96]. To do this,



Crawley, et. al.             Informational                     [Page 27]

RFC 2386           A Framework for QoS-based Routing         August 1998


  first each VC is associated with a maximum permissible routing cost.
  This cost can be set based on expected revenues in carrying the VC or
  simply based on the length of the shortest path to the destination.
  Each link is associated with a metric that increases exponentially
  with its utilization. A switch computing a path for a VC simply
  determines a least-cost feasible path based on the link metric and
  the VC's QoS requirement.  The VC is admitted if the cost of the path
  is less than or equal to the maximum permissible routing cost. This
  routing scheme thus limits the extent of "detour" a VC experiences,
  thus preventing excessive resource consumption. This is a practical
  scheme and the basic idea can be extended to hierarchical routing.
  But the performance of this scheme has not been analyzed thoroughly.
  A similar notion of admission control based on the connection route
  was also incorporated in a routing scheme presented in [ACG92].

  Considering the ATM Forum PNNI protocol [PNNI96], a partial list of
  its stated characteristics are as follows:

           o   Scales to very large networks
           o   Supports hierarchical routing
           o   Supports QoS
           o   Uses source routed connection setup
           o   Supports multiple metrics and attributes
           o   Provides dynamic routing

  The PNNI specification is sub-divided into two protocols: a signaling
  and a routing protocol. The PNNI signaling protocol is used to
  establish point-to-point and point to multipoint connections and
  supports source routing, crankback and alternate routing. PNNI source
  routing allows loop free paths.  Also, it allows each implementation
  to use its own path computation algorithm. Furthermore, source
  routing is expected to support incremental deployment of future
  enhancements such as policy routing.

  The PNNI routing protocol is a dynamic, hierarchical link state
  protocol that propagates topology information by flooding it through
  the network.  The topology information is the set of resources (e.g.,
  nodes, links and addresses) which define the network. Resources are
  qualified by defined sets of metrics and attributes (delay, available
  bandwidth, jitter, etc.) which are grouped by supported traffic
  class.  Since some of the metrics used will change frequently, e.g.,
  available bandwidth, threshold algorithms are used to determine if
  the change in a metric or attribute is significant enough to require
  propagation of updated information.  Other features include, auto
  configuration of the routing hierarchy, connection admission control
  (as part of path calculation) and aggregation and summarization of
  topology and reachability information.




Crawley, et. al.             Informational                     [Page 28]

RFC 2386           A Framework for QoS-based Routing         August 1998


  Despite its functionality, the PNNI routing protocol does not address
  the issues of multicast routing, policy routing and control of
  alternate routing. A problem in general with link state QoS-based
  routing is that of efficient broadcasting of state information. While
  flooding is a reasonable choice with static link metrics it may
  impact the performance adversely with dynamic metrics.

  Finally, Integrated PNNI [I-PNNI] has been designed from the start to
  take advantage of the QoS Routing capabilities that are available in
  PNNI and integrate them with routing for layer 3.  This would provide
  an integrated layer 2 and layer 3 routing protocol for networks that
  include PNNI in the ATM core.  The I-PNNI specification has been
  under development in the ATM Forum and, at this time, has not yet
  incorporated QoS routing mechanisms for layer 3.

9.4   Packet Networks

  Early attempts at adaptive routing in packet networks had the
  objective of delay minimization by dynamically adapting to network
  congestion.  Alternate routing based on k-shortest path tables, with
  route selection based on some local measure (e.g., shortest output
  queue) has been described [R76, YS81]. The original ARPAnet routing
  scheme was a distance vector protocol with delay-based cost metric
  [MW77]. Such a scheme was shown to be prone to route oscillations
  [B82]. For this and other reasons, a link state delay-based routing
  scheme was later developed for the ARPAnet [MRR80]. This scheme
  demonstrated a number of techniques such as triggered updates,
  flooding, etc., which are being used in OSPF and PNNI routing today.
  Although none of these schemes can be called QoS-based routing
  schemes, they had features that are relevant to QoS-based routing.

  IBM's System Network Architecture (SNA) introduced the concept of
  Class of Service (COS)-based routing [A79, GM79].  There were several
  classes of service:  interactive, batch, and network control.  In
  addition, users could define other classes. When starting a data
  session an application or device would request a COS.  Routing would
  then map the COS into a statically configured route which marked a
  path across the physical network.  Since SNA is connection oriented,
  a session was set up along this path and the application's or
  device's data would traverse this path for the life of the session.
  Initially, the service delivered to a session was based on the
  network engineering and current state of network congestion. Later,
  transmission priority was added to subarea SNA.  Transmission
  priority allowed more important traffic (e.g. interactive) to proceed
  before less time-critical traffic (e.g. batch) and improved link and
  network utilization. Transmission priority of a session was based on
  its COS.




Crawley, et. al.             Informational                     [Page 29]

RFC 2386           A Framework for QoS-based Routing         August 1998


  SNA later evolved to support multiple or alternate paths between
  nodes.  But, although assisted by network design tools, the network
  administrator still had to statically configure routes. IBM later
  introduced SNA's Advanced Peer to Peer Networking (APPN) [B85]. APPN
  added new features to SNA including dynamic routing based on a link
  state database. An application would use COS to indicate it traffic
  requirements and APPN would calculate a path capable of meeting these
  requirements.  Each COS was mapped to a table of acceptable metrics
  and parameters that qualified the nodes and links contained in the
  APPN topology Database.  Metrics and parameters used as part of the
  APPN route calculation include, but are not limited to:  delay, cost
  per minute, node congestion and security.  The dynamic nature of APPN
  allowed it to route around failures and reduce network configuration.

  The service delivered by APPN was still based on the network
  engineering, transmission priority and network congestion.  IBM later
  introduced an extension to APPN, High Performance Routing
  (HPR)[IBM97]. HPR uses a congestion avoidance algorithm called
  adaptive rate based (ARB) congestion control.  Using predictive
  feedback methods, the ARB algorithm prevents congestion and improves
  network utilization.  Most recently, an extension to the COS table
  has been defined so that HPR routing could recognize and take
  advantage of ATM QoS capabilities.

  Considering IP routing, both IDRP [R92] and OSPF support  type of
  service (TOS)-based routing. While the IP header has a TOS field,
  there is no standardized way of utilizing it for TOS specification
  and routing. It seems possible to make use of the IP TOS feature,
  along with TOS-based routing and proper network engineering, to do
  QoS-based routing. The emerging differentiated services model is
  generating renewed interest in TOS support. Among the newer schemes,
  Source Demand Routing (SDR) [ELRV96] allows  on-demand path
  computation by routers and the implementation of strict and loose
  source routing. The Nimrod architecture [CCM96] has a number of
  concepts built in to handle scalability and specialized path
  computation. Recently, some work has been done on QoS-based routing
  schemes for the integrated services Internet. For example, in [M98],
  heuristic schemes for efficient routing of flows with bandwidth
  and/or delay constraints is described and evaluated.

9. SUMMARY AND CONCLUSIONS

  In this document, a framework for QoS-based Internet routing was
  defined.  This framework adopts the traditional separation between
  intra and interdomain routing. This approach is especially meaningful
  in the case of QoS-based routing, since there are many views on how
  QoS-based routing should be accomplished and many different needs.
  The objective of this document was to encourage the development of



Crawley, et. al.             Informational                     [Page 30]

RFC 2386           A Framework for QoS-based Routing         August 1998


  different solution approaches for intradomain routing, subject to
  some broad requirements, while consensus on interdomain routing is
  achieved. To this end, the QoS-based routing issues were described,
  and some broad intradomain routing requirements and an interdomain
  routing model were defined. In addition, QoS-based multicast routing
  was discussed and a detailed review of related work was presented.

  The deployment of QoS-based routing across multiple administrative
  domains requires both the development of intradomain routing schemes
  and a standard way for them to interact via a well-defined
  interdomain routing mechanism. This document, while outlining the
  issues that must be addressed, did not engage in the specification of
  the actual features of the interdomain routing scheme. This would be
  the next step in the evolution of wide-area, multidomain QoS-based
  routing.

REFERENCES

  [A79]    V. Ahuja, "Routing and Flow Control in SNA", IBM Systems
           Journal, 18 No. 2, pp.  298-314, 1979.

  [A84]    J. M. Akinpelu, "The Overload Performance of Engineered
           Networks with Non-Hierarchical Routing", AT&T Technical
           Journal, Vol. 63, pp. 1261-1281, 1984.

  [ACFH92] G. R. Ash, J. S. Chen, A. E. Frey and B. D. Huang, "RealTime
           Network Routing in a Dynamic Class-of-Service Network",
           Proceedings of ITC 13, 1992.

  [ACG92]  H. Ahmadi, J. Chen, and R. Guerin, "Dynamic Routing and Call
           Control in High-Speed Integrated Networks", Proceedings of
           ITC-13, pp. 397-403, 1992.

  [B82]    D. P. Bertsekas, "Dynamic Behavior of Shortest Path Routing
           Algorithms for Communication Networks", IEEE Trans. Auto.
           Control, pp. 60-74, 1982.

  [B85]    A. E. Baratz, "SNA Networks of Small Systems", IEEE JSAC,
           May, 1985.

  [BBCD98] Black, D., Blake, S., Carlson, M., Davies, E., Wang, Z., and
           W. Weiss, "An Architecture for Differentiated Services",
           Work in Progress.

  [BCCH94] Braden, R., Clark, D., Crocker, D., and C. Huitema, "Report
           of IAB Workshop on Security in the Internet Architecture",
           RFC 1636, June 1994.




Crawley, et. al.             Informational                     [Page 31]

RFC 2386           A Framework for QoS-based Routing         August 1998


  [BCF94]  A. Ballardie, J. Crowcroft and P. Francis, "Core-Based
           Trees: A Scalable Multicast Routing Protocol", Proceedings
           of SIGCOMM `94.

  [BCS94]  Braden, R., Clark, D., and S. Shenker, "Integrated Services
           in the Internet Architecture: An Overview", RFC 1633, July
           1994.

  [BZ92]   S. Bahk and M. El Zarki, "Dynamic Multi-Path Routing and How
           it Compares with Other Dynamic Routing Algorithms for High
           Speed Wide Area Networks", Proc. SIGCOMM `92, pp. 53-64,
           1992.

  [BZBH97] Braden, R., Zhang, L., Berson, S., Herzog, S., and S. Jamin,
           "Resource ReSerVation Protocol (RSVP) -- Version 1
           Functional Spec", RFC 2205, September 1997.

  [C91]    C-H. Chow, "On Multicast Path Finding Algorithms",
           Proceedings of the IEEE INFOCOM `91, pp. 1274-1283, 1991.

  [CCM96]  Castineyra, I., Chiappa, J., and M. Steenstrup, "The Nimrod
           Routing Architecture", RFC 1992, August 1996.

  [DEFV94] S. E. Deering, D. Estrin, D. Farinnacci, V. Jacobson, C-G.
           Liu, and L. Wei, "An Architecture for Wide-Area Multicast
           Routing", Technical Report, 94-565, ISI, University of
           Southern California, 1994.

  [ELRV96] Estrin, D., Li, T., Rekhter, Y., Varadhan, K., and D.
           Zappala, "Source Demand Routing: Packet Format and
           Forwarding Specification (Version 1)", RFC 1940, May 1996.

  [GKR96]  R. Gawlick, C. R. Kalmanek, and K. G. Ramakrishnan, "On-Line
           Routing of Permanent Virtual Circuits", Computer
           Communications, March, 1996.

  [GPSS98] A. Ghanwani, J. W. Pace, V. Srinivasan, A. Smith and M.
           Seaman, "A Framework for Providing Integrated Services over
           Shared and Switched IEEE 802 LAN Technologies", Work in
           Progress.

  [GM79]   J. P. Gray, T. B. McNeil, "SNA Multi-System Networking", IBM
           Systems Journal, 18 No. 2, pp.  263-297, 1979.

  [GOA97]  Y. Goto, M. Ohta and K. Araki, "Path QoS Collection for
           Stable Hop-by-Hop QoS Routing", Proc. INET '97, June, 1997.





Crawley, et. al.             Informational                     [Page 32]

RFC 2386           A Framework for QoS-based Routing         August 1998


  [GKOP98] R. Guerin, S. Kamat, A. Orda, T. Przygienda, and D.
           Williams, "QoS Routing Mechanisms and OSPF extensions", work
           in progress, March, 1998.

  [IBM97]  IBM Corp, SNA APPN - High Performance Routing Architecture
           Reference, Version 2.0, SV40-1018, February 1997.

  [IPNNI]  ATM Forum Technical Committee. Integrated PNNI (I-PNNI) v1.0
           Specification. af-96-0987r1, September 1996.

  [ISI81]  Postel, J., "Internet Protocol", STD 5, RFC 791, September
           1981.

  [JMW83]  J. M. Jaffe, F. H. Moss, R. A. Weingarten, "SNA Routing:
           Past, Present, and Possible Future", IBM Systems Journal,
           pp.  417-435, 1983.

  [K88]    F.P. Kelly, "Routing in Circuit-Switched Networks:
           Optimization, Shadow Prices and Decentralization", Adv.
           Applied Prob., pp. 112-144, March, 1988.

  [L95]    W. C. Lee, "Topology Aggregation for Hierarchical Routing in
           ATM Networks", ACM SIGCOMM Computer Communication Review,
           1995.

  [M86]    L. G. Mason, "On the Stability of Circuit-Switched Networks
           with Non-hierarchical Routing", Proc. 25th Conf. On Decision
           and Control, pp. 1345-1347, 1986.

  [M98]    Moy, J., "OSPF Version 2", STD 54, RFC 2328, April 1998.

  [M94]    Moy, J., "MOSPF: Analysis and Experience", RFC 1585,  March
           1994.

  [M98]    Q. Ma, "Quality-of-Service Routing in Integrated Services
           Networks", PhD thesis, Computer Science Department, Carnegie
           Mellon University, 1998.

  [MMR96]  D. Mitra, J. Morrison, and K. G. Ramakrishnan, "ATM Network
           Design and Optimization: A Multirate Loss Network
           Framework", Proceedings of IEEE INFOCOM `96, 1996.

  [MRR80]  J. M. McQuillan, I. Richer and E. C. Rosen, "The New Routing
           Algorithm for the ARPANET", IEEE Trans.  Communications, pp.
           711-719, May, 1980.






Crawley, et. al.             Informational                     [Page 33]

RFC 2386           A Framework for QoS-based Routing         August 1998


  [MS91]   D. Mitra and J. B. Seery, "Comparative Evaluations of
           Randomized and Dynamic Routing Strategies for Circuit
           Switched Networks", IEEE Trans. on Communications, pp. 102-
           116, January, 1991.

  [MW77]   J. M. McQuillan and D. C. Walden, "The ARPANET Design
           Decisions", Computer Networks, August, 1977.

  [NC94]   Nair, R. and Clemmensen, D. : "Routing in Integrated
           Services Networks", Proc. 2nd International Conference on
           Telecom.  Systems  Modeling and Analysis, March 1994.

  [P88]    R. Perlman, "Network Layer Protocol with Byzantine
           Robustness", Ph.D. Thesis, Dept. of EE and CS, MIT, August,
           1988.

  [PNNI96] ATM Forum PNNI subworking group, "Private Network-Network
           Interface Spec.  v1.0 (PNNI 1.0)", afpnni-0055.00, March
           1996.

  [R76]    H. Rudin, "On Routing and "Delta Routing": A Taxonomy and
           Performance Comparison of Techniques for Packet-Switched
           Networks", IEEE Trans. Communications, pp. 43-59, January,
           1996.

  [R92]    Y. Rekhter, "IDRP Protocol Analysis: Storage Overhead", ACM
           Comp.  Comm.  Review, April, 1992.

  [R96]    B. Rajagopalan, "Efficient Link State Routing", Work in
           Progress, available from [email protected].

  [RN98]   B. Rajagopalan and R. Nair, "Multicast Routing with Resource
           Reservation", to appear in J. of High Speed Networks, 1998.

  [SD95]   S. Sibal and A. Desimone, "Controlling Alternate Routing in
           General-Mesh Packet Flow Networks", Proceedings of ACM
           SIGCOMM, 1995.

  [SPG97]  Shenker, S., Partridge, C., and R. Guerin, "Specification of
           Guaranteed Quality of Service", RFC 2212, September 1997.

  [T88]    D. M. Topkis, "A k-Shortest-Path Algorithm for Adaptive
           Routing in Communications Networks", IEEE Trans.
           Communications, pp.  855-859, July, 1988.

  [W88]    B. M. Waxman, "Routing of Multipoint Connections", IEEE
           JSAC, pp. 1617-1622, December, 1988.




Crawley, et. al.             Informational                     [Page 34]

RFC 2386           A Framework for QoS-based Routing         August 1998


  [W97]   Wroclawski, J., "Specification of the Controlled-Load Network
           Element Service", RFC 2211, September 1997.

  [WC96]   Z. Wang and J. Crowcroft, "QoS Routing for Supporting
           Resource Reservation", IEEE JSAC, September, 1996.

  [YS81]   T. P. Yum and M. Schwartz, "The Join-Based Queue Rule and
           its Application to Routing in Computer Communications
           Networks", IEEE Trans. Communications, pp. 505-511, 1981.

  [YS87]   T. G. Yum and M. Schwartz, "Comparison of Routing Procedures
           for Circuit-Switched Traffic in Nonhierarchical Networks",
           IEEE Trans. Communications, pp. 535-544, May, 1987.

  [ZES97]  Zappala, D., Estrin, D., and S. Shenker, "Alternate Path
           Routing and Pinning for Interdomain Multicast Routing", USC
           Computer Science Technical Report #97-655, USC, 1997.

  [ZSSC97] Zhang, Z., Sanchez, C., Salkewicz, B., and E. Crawley, "QoS
           Extensions to OSPF", Work in Progress.































Crawley, et. al.             Informational                     [Page 35]

RFC 2386           A Framework for QoS-based Routing         August 1998


AUTHORS' ADDRESSES

  Bala Rajagopalan
  NEC USA, C&C Research Labs
  4 Independence Way
  Princeton, NJ 08540
  U.S.A

  Phone: +1-609-951-2969
  EMail: [email protected]


  Raj Nair
  Arrowpoint
  235 Littleton Rd.
  Westford, MA 01886
  U.S.A

  Phone: +1-508-692-5875, x29
  EMail: [email protected]


  Hal Sandick
  Bay Networks, Inc.
  1009 Slater Rd., Suite 220
  Durham, NC 27703
  U.S.A

  Phone: +1-919-941-1739
  EMail: [email protected]


  Eric S. Crawley
  Argon Networks, Inc.
  25 Porter Rd.
  Littelton, MA 01460
  U.S.A

  Phone: +1-508-486-0665
  EMail: [email protected]











Crawley, et. al.             Informational                     [Page 36]

RFC 2386           A Framework for QoS-based Routing         August 1998


Full Copyright Statement

  Copyright (C) The Internet Society (1998).  All Rights Reserved.

  This document and translations of it may be copied and furnished to
  others, and derivative works that comment on or otherwise explain it
  or assist in its implementation may be prepared, copied, published
  and distributed, in whole or in part, without restriction of any
  kind, provided that the above copyright notice and this paragraph are
  included on all such copies and derivative works.  However, this
  document itself may not be modified in any way, such as by removing
  the copyright notice or references to the Internet Society or other
  Internet organizations, except as needed for the purpose of
  developing Internet standards in which case the procedures for
  copyrights defined in the Internet Standards process must be
  followed, or as required to translate it into languages other than
  English.

  The limited permissions granted above are perpetual and will not be
  revoked by the Internet Society or its successors or assigns.

  This document and the information contained herein is provided on an
  "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING
  TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING
  BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION
  HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF
  MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
























Crawley, et. al.             Informational                     [Page 37]


You are viewing proxied material from gopher.fnord.one. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.