home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!psinntp!internet!sbi!pivot-sts!captain!talbot
- From: talbot@captain (John Talbot)
- Newsgroups: comp.dcom.sys.cisco
- Subject: Frame Relay
- Message-ID: <1165@pivot-sts.sbi.com>
- Date: 9 Sep 92 17:21:48 GMT
- Sender: news@pivot-sts.sbi.com
- Lines: 85
- Nntp-Posting-Host: captain
-
-
- Subject: Frame Relay
- Newsgroups: comp.dcom.sys.cisco
-
-
-
- OPEN QUESTION TO IMPLEMENTERS OF LARGE
- FRAME RELAY NETWORKS
-
-
- Has anyone noticed the lack of control you
- have in selectively advertising and learning
- routes on frame relay interfaces ? Although
- the usual distribute-lists apply, frame relay
- is unique in that you have >1 router at the
- other end of the link. Now assume you have
- 100-200 routers out there, somewhere, at non-company
- owned sites. Although you own the boxes, you can't
- guarrantee their security at these remote sites.
- Wouldn't it be beneficial to be able
- to have total control of the routing updates at
- the hub router(s) at your site ? This is not possible
- with Cisco's current implementation (not possible with
- other implementations either).
-
- Frame-relay is very "loosey goosey" in how the router
- verifies and controls routing updates. A single misconfigured
- router at a client site could disable your entire network, and
- make it very difficult to locate the problem. These are the
- issues we're currently dealing with.
-
- I'd like to see Cisco enhance the control mechanisms
- to be suitable for a *large* frame-relay environment.
-
- Here's a suggestion. Create a frame-relay version of
- the distribute list feature, this time using extended
- IP access lists. I should preface this discussion with
- the comment that Cisco sees no value in providing routing
- control beyond what is already provided.
-
- Assume topology as below, netmask 255.255.255.0
- My router (my site) is address 129.10.10.10.
- Each client side router has >=1 subnet to tell me about.
- These routers can not be assumed to be secure.
-
- R 129.10.10.1 R 129.10.10.2 R 129.10.10.3
- | | |
- | | |
- | | |
- ---------------------------------------
- ---- ------
- cloud
- ---- ------
- ---------------------------------------
- |
- |
- |
- R 129.10.10.10
- (my router, my datacenter)
-
-
- access-list 101 permit 129.10.10.1 0.0.0.0 129.10.20.0 0.0.0.0
- access-list 101 permit 129.10.10.2 0.0.0.0 129.10.21.0 0.0.0.0
- access-list 101 permit 129.10.10.3 0.0.0.0 129.10.22.0 0.0.0.0
- access-list 101 permit 129.10.10.3 0.0.0.0 129.10.23.0 0.0.0.0
-
- router igrp xxx
- frame-relay distribute-list 101 in [ serial 0 ]
-
- This would restrict which subnets you will learn from which
- routers. This could similarly be achieved with the "distance"
- feature, as mentioned by Bob Albrightson of Cisco. However, this
- distance feature relies on a single access list for each router
- which limits the number of routers to 99 because of the limit of
- the number of standard IP access-lists. An AGS+ with 4,5
- frame-relay links shouldn't be limited to 99 endpoints. Throwing
- more routers and links at the problem is not the answer either.
-
- I believe this feature, although not well received by Cisco, would
- work well, be easily understood, and be consistent with their
- former approaches to routing control.
-
- I'm interested in hearing comments.
-
- Thanks.
-