Network Working Group                                          Z. Wang
Request for Comments: 1335                                J. Crowcroft
                                            University College London
                                                             May 1992


            A Two-Tier Address Structure for the Internet:
        A Solution to the Problem of Address Space Exhaustion

Status of this Memo

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

Abstract

  This RFC presents a solution to problem of address space exhaustion
  in the Internet.  It proposes a two-tier address structure for the
  Internet.  This is an "idea" paper and discussion is strongly
  encouraged.

Introduction

  Address space exhaustion is one of the most serious and immediate
  problems that the Internet faces today [1,2].  The current Internet
  address space is 32-bit.  Each Internet address is divided into two
  parts: a network portion and a host portion.  This division
  corresponds the three primary Internet address classes: Class A,
  Class B and Class C.  Table 1 lists the network number statistics as
  of April 1992.

                     Total       Allocated     Allocated (%)
  Class A              126            48            54%
  Class B            16383          7006            43%
  Class C          2097151         40724             2%

         Table 1: Network Number Statistics (April 1992)

  If recent trends of exponential growth continue, the network numbers
  in Class B will soon run out [1,2].  There are over 2 million Class C
  network numbers and only 2% have been allocated.  However, a Class C
  network number can only accommodate 254 host numbers which is too
  small for most networks.  With the rapid expansion of the Internet
  and drastic increase in personal computers, the time when the 32-bit
  address space is exhausted altogether is also not too distant [1-3].

  Recently several proposals have been put forward to deal with the



Wang & Crowcroft                                                [Page 1]

RFC 1335      Two-Tier Address Structure for the Internet       May 1992


  immediate problem [1-4].  The Supernetting and C-sharp schemes
  attempt to make the Class C numbers more usable by re-defining the
  way in which Class C network numbers are classified and assigned
  [3,4].  Both schemes require modifications to the exterior routing
  algorithms and global coordination across the Internet may be
  required for the deployment.  The two schemes do not expand the total
  number of addresses available to the Internet and therefore can only
  be used as a short-term fix for next two or three years.  Schemes
  have also been put forwarded in which the 32-bit address field is
  replaced with a field of the same size but with different meaning and
  the gateways on the boundary re-write the address when the packet
  crossed the boundary [1,2,5].  Such schemes, however, requires
  substantial changes to the gateways and the exterior routing
  algorithm.

  In this paper, we present an alternative solution to the problem of
  address space exhaustion.  The "Dual Network Addressing (DNA)" scheme
  proposed here is based on a two-tier address structure and sharing of
  addresses.  It requires no modifications to the exterior routing
  algorithms and any networks can adopt the scheme individually at any
  time without affecting other networks.

The Scheme

  The DNA scheme attempts to reduce the waste in using the Internet
  addresses.  A useful analogy to our scheme is the extension system
  used in the telephone system.  Many large organizations usually have
  extensive private telephone networks for internal use and at the mean
  time hire a limited number of external lines for communications with
  the outside world.  In such a telephone system, important offices may
  have direct external lines and telephones in the public areas may be
  restricted to internal calls only.  The majority of the telephones
  can usually make both internal calls and external calls.  But they
  must share a limited number of external lines.  When an external call
  is being made, a pre-defined digit has to be pressed so that an
  external line can be allocated from the poll of external lines.

  In the DNA scheme, there are two types of Internet addresses:
  Internal addresses and External addresses.  An internal address is an
  Internet address only used within one network and is unique only
  within that network.  An interface with an internal address can only
  communicate with another interface with an internal address in the
  same network.  An external address is unique in the entire Internet
  and an interface with an external address can communicate directly to
  another interface with an external address over the Internet.  All
  current Internet addresses are external addresses.

  In effect, the external addresses form one global Internet and the



Wang & Crowcroft                                                [Page 2]

RFC 1335      Two-Tier Address Structure for the Internet       May 1992


  internal addresses form many private Internets.  Within one network,
  the external addresses are only used for inter-network communications
  and internal addresses for intra-network communications.  An External
  Address Sharing Service (EASS) is needed to manage the sharing of
  external addresses.  An EASS server reserves a number of external
  addresses.  When a machine that only has an internal address wants to
  communicate a machine with an external address in other networks, it
  can send a request to an EASS server to obtain a temporary external
  address.  After the use, the machine can return the external address
  to the EASS server.

  We believe that, with the DNA scheme, a network can operate with a
  limited number of external addresses.  The reasons are as follows:

  *  In most networks, the majority of the traffic is confined to
     its local area networks.  This is due the nature of
     networking applications and the bandwidth constraints on
     inter-network links.

  *  The number of machines which act as Internet servers, i.e.,
     running programs waiting to be called by machines in other
     networks, is often limited and certainly much smaller than
     the total number of machines.  These machines include mail
     servers, domain name servers, ftp archive servers, directory
     servers, etc.

  *  There are an increasingly large number of personal machines
     entering the Internet.  The use of these machines is
     primarily limited to their local environment.  They may also
     be used as "clients" such as ftp and telnet to access other
     machines.

  *  For security reasons, many large organizations, such as banks,
     government departments, military institution and some
     companies, may only allow a very limited number of their
     machines to have access to the global Internet.  The majority
     of their machines are purely for internal use.

  In the DNA scheme, all machines in a network are assigned a permanent
  internal address and can communicate with any machines within the
  same network.  The allocation of external addresses depends on the
  functions of the machines and as a result it creates three-level
  privileges:

  *  machines which act as servers or used as central computing
     infrastructure are likely to have frequent communications
     with other networks therefore they may require external
     addresses all the time.  These machines are allocated



Wang & Crowcroft                                                [Page 3]

RFC 1335      Two-Tier Address Structure for the Internet       May 1992


     permanent external addresses.

  *  machines which are not allowed to communicate with other
     networks have no external addresses and can only communicate
     with machines within their own network.

  *  the rest of the machines share a number of external
     addresses. The external addresses are allocated by
     the EASS server on request.  These machines can only
     used as clients to call machines in other networks,
     i.e., they can not be called by machines in other networks.

  A network can choose any network number other than its external
  network number as its internal network number.  Different networks
  can use the same network number as their internal number.  We propose
  to reserve one Class A network number as the well-known network
  number for internal use.

The Advantages

  The DNA scheme attempts to tackle the problem from the bottom of the
  Internet, i.e., each individual network, while other schemes
  described in the first section deal with the problem from the top of
  the Internet, i.e., gateways and exterior routing algorithms.  These
  schemes, however, do not need to be consider as mutually exclusive.
  The DNA scheme has several advantages:

  *  The DNA scheme takes an evolutionary approach towards the
     changes.  Different networks can individually choose to
     adopt the scheme at any time only when necessary.
     There is no need for global coordination between different
     networks for their deployment.  The effects of the deployment
     are confined to the network in which the scheme is being
     implemented, and are invisible to exterior routing
     algorithms and external networks.

  *  With the DNA scheme, it is possible for a medium size organization
     to use a Class C network number with 254 external addresses.
     The scheme allows the current Internet to expand to over 2 million
     networks and each network to have more than 16 million hosts.
     This will allow considerable time for a long-term solution to
     be developed and fully tested.

  *  The DNA scheme requires modifications to the host software.
     However, the modifications are needed only in those networks
     which adopt the DNA scheme.   Since all existing Class A and B
     networks usually have sufficient external addresses for all their
     machines, they do not need to adopt the DNA scheme, and therefore



Wang & Crowcroft                                                [Page 4]

RFC 1335      Two-Tier Address Structure for the Internet       May 1992


     need no modifications at all to their software.  The networks
     which need to use the DNA scheme are those new networks which are
     set up after the Class A and B numbers run out and have to
     use a Class C number.

  *  The DNA scheme makes it possible to develop to a new addressing
     scheme without expanding the 32-bit address length to 64-bit.
     With the two-tier address structure, the current 32-bit space
     can accommodate over 4 billion hosts in the global Internet and
     100 million hosts in each individual network.  When we move to a
     classless multi-hierarchic addressing scheme, the use of external
     addresses can be more efficient and less wasteful and the
     32-bit space can be adequate for the external addresses.

  *  When a new addressing scheme has been developed, all current
     Internet addresses have to be changed.  The DNA scheme will make
     such a undertaking much easier and smoother, since only the
     EASS servers and those have permanent external addresses will
     be affected, and communications within the network will not
     be interrupted.

The Modifications

  The major modifications to the host software is in the network
  interface code.  The DNA scheme requires each machine to have at
  least two addresses.  But most of the host software currently does
  not allow us to bind two addresses to one physical interface.  This
  problem can be solved by using two network interfaces on each
  machine.  But this option is too expensive.  Note the two interfaces
  are actually connected to the same physical network.  Therefore, if
  we modify the interface code to allow two logical interfaces to be
  mapped onto one single physical interface, the machine can then use
  both the external address and the internal address with one physical
  interface as if it has two physical interfaces.  In effect, two
  logical IP networks operate over the same physical network.

  The DNA scheme also has implications to the DNS service.  Many
  machines will have two entries in the local name server.  The DNS
  server must examine the source address of the request and decide
  which entry to use.  If the source address matches the well-known
  internal network number, it passes the internal address of the domain
  name.  Otherwise, the name server passes the external address.

  An EASS server is required to manage the sharing of the external
  addresses, i.e., to allocate and de-allocate external addresses to
  the machines which do not have permanent external addresses.  This
  service can be provided by using the "Dynamic Host Configuration
  Protocol (DHCP)" [6].



Wang & Crowcroft                                                [Page 5]

RFC 1335      Two-Tier Address Structure for the Internet       May 1992


  Many hosts do an inverse lookup of incoming connections.  Therefore,
  it is desirable the entry in the DNS server be updated whenever a new
  external address is allocated.  This will also allow an machine which
  currently has a temporary external address to be called by other
  machines.  The updating of the entry in the DNS server can be done
  more easily if the EASS server and DNS server are co-located.

Acknowledgements

  We would like to thank J. K. Reynolds for the network statistics, and
  V. Cerf, C. Topolcic, K. McCloghrie, R. Ullmann and K. Carlberg for
  their useful comments and discussion.

References

  [1]  Chiappa, N., "The IP Addressing Issue", work in progress,
       October 1990.

  [2]  Clark, D., Chapin, L., Cerf, V., Braden, R., and R. Hobby,
       "Towards the Future Architecture", RFC 1287, MIT, BBN, CNRI,
       ISI, UC Davis, December 1991.

  [3]  Solensky, F., and F. Kastenholz, "A Revision to IP Address
       Classifications", work in progress, March 1992.

  [4]  Fuller, V., Li, T., Yu, J., and K. Varadhan, "Supernetting:
       an Address Assignment and Aggregation Strategy", work in
       progress, March 1992.

  [5]  Tsuchiya, P., "The IP Network Address Translator", work in
       progress, March 1991.

  [6]  Droms, R., "Dynamic Host Configuration Protocol", work in
       progress, March 1992.

















Wang & Crowcroft                                                [Page 6]

RFC 1335      Two-Tier Address Structure for the Internet       May 1992


Security Considerations

  Security issues are not discussed in this memo.

Authors' Addresses

  Zheng Wang
  Dept. of Computer Science
  University College London
  London WC1E 6BT, UK

  EMail: [email protected]


  Jon Crowcroft
  Dept. of Computer Science
  University College London
  London WC1E 6BT, UK

  EMail: [email protected]































Wang & Crowcroft                                                [Page 7]


You are viewing proxied material from gopher.rbfh.de. 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.