Network Working Group Juha Heinanen
INTERNET DRAFT Telecom Finland
Expires January 1998 July 1997
Intra-area IP unicast among routers over legacy ATM
<draft-ietf-ion-intra-area-unicast-00.txt>
Status of this Memo
This document is an Internet-Draft. Internet-Drafts are working
documents of the Internet Engineering Task Force (IETF), its areas,
and its working groups. Note that other groups may also distribute
working documents as Internet-Drafts.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as ``work in progress.''
To learn the current status of any Internet-Draft, please check the
``1id-abstracts.txt'' listing contained in the Internet-Drafts Shadow
Directories on ftp.is.co.za (Africa), nic.nordu.net (Europe),
munnari.oz.au (Pacific Rim), ds.internic.net (US East Coast), or
ftp.isi.edu (US West Coast).
Abstract
This document describes how IP unicast can be efficiently implemented
among routers belonging to the same area of a routing domain, where
the connectivity is provided by a legacy ATM network as defined by
the ATM Forum or ITU. This proposal is designed to be complementary
to IP multicast solutions such as the one described in [1].
1. Introduction
This document describes how a set of routers (such as the access/edge
routers of an ISP or enterprise) connected to a legacy ATM network
can in a dynamic and plug-and-play fashion optimize ATM connections
for efficient forwarding of unicast IP packets. The method can be
used in situations where the number of routers is so large that a
full mesh of point-to-point ATM VCs is not practical from technical
or economic reasons. In addition, it can be applied to smaller
router networks to automate the setup of a full mesh of ATM
connectivity between the routers.
Heinanen Intra-area IP unicast [Page 1]
INTERNET DRAFT April, 1997
The set of routers must belong to the same area of a link state
routing protocol, such as OSPF or IS-IS, that floods topology and
reachability information to every router in the area.
This proposal only deals with IP unicast, but it complements and can
be used in conjunction with IP multicast solutions such as the one
described in [1].
2. Router configuration and behavior
As introduced above, this document defines a method of dynamically
managing ATM connectivity among a set of routers that belong to the
same area of a routing domain, where a link state protocol, such as
OSPF or IS-IS, is used to exchange topology and reachability
information.
Before the dynamic management of ATM VCs can begin, the routers of
the area must be manually configured to exchange routing information
among themselves. There must thus be enough initial connectivity so
that at least one IP path exists from each router to each other
router in the area. This initial connectivity is also used to
forward IP packets when dynamic ATM VCs don't exist.
Note that the initial connectivity doesn't necessarily need to be
implemented over ATM and that not all routers of the area need to be
ATM attached. Furthermore, even if a router is ATM attached, it
doesn't need to participate in the dynamic management of ATM VCs.
The ATM routers of an area can thus be upgraded one at a time to
support the method described in this document.
Once an ATM attached router R becomes operational, it may start to
measure, how many bytes it has received during the past M seconds,
whose final hop router S within the area is also ATM attached. If
the number of bytes is less than N, R forwards the packets according
to its routing table. When the number of bytes equals or exceeds N,
and R doesn't yet have a dynamic ATM VC to S, R creates such a VC and
starts to forward S bound packets directly.
The dynamic ATM VC is unidirectional that allows R to manage it
autonomously without coordination with S. Also, in order to keep the
number of routing peers small and in order to avoid frequent changes
in topology information, R doesn't use the dynamic ATM VC for the
exchange of routing information nor does R advertise the dynamic ATM
VC to its routing peers.
ATM traffic categories and traffic parameters of the dynamic VCs are
decided locally by the network administrator. The default traffic
category is UBR with the peak cell rate set to the link rate and the
Heinanen Intra-area IP unicast [Page 2]
INTERNET DRAFT April, 1997
minimal acceptable cell rate set to zero. Since the dynamic ATM VCs
are only used to carry IP packets, it is recommended that the packets
are "VC multiplexed" [2] to the AAL 5 payload without an LLC/SNAP
header. See [3] and Appendix D.3.2 of [4] for details on coding of
the signalling messages.
Once the dynamic ATM VC from R to S has been created, R starts to
measure the traffic along it. When R detects that during the past K
seconds the number of bytes along the dynamic ATM VC to S has fallen
below L, it deletes the dynamic ATM VC and returns to the initial
mode of operation that was described above.
The values of the constants K, L, M, and N control the rate of
dynamic ATM VC creation and deletion. They are assigned by the
network administrator and may differ from one ATM attached router to
another.
Setting the VC creation and deletion limits N and L to zero, turns
off the measurement process and causes the router to create a dynamic
VC to every other participating router. That can be the default in
small router networks that want to use this method automate the setup
of a full mesh of ATM VCs.
If a router doesn't want to set up any dynamic ATM VCs, the VC
creation limit N is set to a positive value and the measurement
interval M is set to zero. Finally, if a router doesn't want to be a
destination of dynamic ATM VCs, it doesn't make its ATM address
available to the other routers for the purpose of this application.
Note that if a router is not capable in measuring traffic, it can
still participate as a destination of dynamic ATM VCs and can itself
set up dynamic VCs non-selectively to every other router.
3. Address resolution
Since all the routers belong to the same area of a link state routing
domain, they learn each others' router IDs and the IP address
prefixes that are reachable via each router. In order to dynamically
create an ATM VC from one router to another, the source router also
needs to learn the ATM address of the destination router.
A router that wants to participate as a source in the dynamic
management of ATM VCs, makes its ATM address known to the other
routers of the area by including in its link state advertisements a
field that contains an ATM address of the advertising router. In
OSPF, this ATM address information is carried in an Address
Resolution Advertizement [5] with area-local flooding scope and with
a service type. The service type identifies the possible uses of the
Heinanen Intra-area IP unicast [Page 3]
INTERNET DRAFT April, 1997
advertised ATM address which includes the dynamic creation of ATM VCs
according the method described in this document.
A field similar to Opaque LSA could be easily defined for IS-IS.
Futher, it could be possible to use a well-known discretionary non-
transitive attribute of BGP to carry the address resolution
information, but the use of inter-domain routing protocols is outside
the scope of this document.
4. Discussion
The method proposed in this document allows efficient interconnection
of a set of routers over a legacy ATM network. After small amount of
manual configuration, the routers will automatically optimize direct
connectivity among themselves based on dynamic traffic load. Network
administrators can control the number of ATM VCs created by the
method taking into account scalability and cost.
The number of ATM VCs could be further reduced if legacy ATM networks
would support signalling of multipoint-to-point VCs. The method
would also benefit from the capability to renegotiate the traffic
parameters of active ATM VCs. Both of these capabilities are
currently under study in the ATM Forum or the ITU.
5. Security Considerations
Since the method described in this document allows data paths to be
established that bypass the normal hop-by-hop control path, the
location of any access filters should be decided carefully. To
ensure proper enforcement of filter policies, filters should be moved
to the edges of an area so that they me be applied on entry or exit
from the short-cut data path.
References
[1] Farinacci, D., Meyer, D., and Rekhter, Y., "Intra-LIS IP
multicast among routers over ATM".
draft-ietf-ion-intralis-multicast-00.txt, April 1997.
[2] Heinanen, J., "Multiprotocol Encapsulation over ATM Adaptation
Layer 5". RFC 1483, July 1993.
[3] Perez, M, et al., "ATM Signalling Support for IP over ATM". RFC
1755, February 1995.
[4] The ATM Forum, "ATM User-Network Interface Specification Version
3.1". September 1994.
Heinanen Intra-area IP unicast [Page 4]
INTERNET DRAFT April, 1997
[5] Coltun, R., Heinanen, J., Cai Y., "The OSPF Address Resolution
Advertisement Option". Internet Draft, August 1997.
Acknowledgements
I would like to thank Rob Coltun and Lou Berger of Fore Systems for
their comments on earlier versions of this document.
Author Information
Juha Heinanen
Telecom Finland
PO Box 777
33101 Tampere
Finland
Phone +358 400 500 958
Email jh@tele.fi
Heinanen Intra-area IP unicast [Page 5]
