Ascend Archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: (ASCEND) Virtual IP routing / IP Navigator



At 03:31 PM 7/29/98 -0400, Phillip Vandry wrote:
>Just read some stuff on "IP Navigator"'s Virtual IP Routing capabilities.
>(http://www.ascend.com/2871.html).
>
>The Virtual IP routing (splitting the router into multiple independant
>logical routers) sounds like a very cool, but also very unconventional
>feature.

If progress in routers is unconventional, then yes.

>This functionality certainly has a lot of repercussions for the
>router's architecture. I'm curious as to how some of these are dealt
>with. Does anyone have any experience with this?
>
>I guess this is implemented by assigning each interface on the router
>(perhaps an ISDN connection, an Ethernet interface, or a Frame/ATM
>PVC) to exactly one of the multiple different routing groups.

Not to exactly one. Individual routes can be assigned to a given physical
or virtual circuit. But since they are virtual routes, identical addresses
can co-exist on the same physical network. In other words, a VPN.

>How is route caching implemented? Of course the route cache is only
>valid for one of the virtual routers. Having multiple route caches
>probably isn't feasible because the route cache may be in SRAM which
>is limited in quantity, or the route cache (especially in the case
>of the GRF) may be held right on the interface card which has a
>fixed, relatively small amount of RAM. Hypotheses:
>
>  - Route cache is only used for the 'main' router. Not a real good
>    solution, all the other virtual routers take a big performance hit.
>  - Route cache is tagged with an extra key which identifies the
>    virtual router to which the cache entry belongs and keeps the
>    virtual routers separate
>  - The number of possible virtual routers is seriously limited by
>    available cache RAM.

Route caching is only for "weak" or "legacy" routers that cannot hold the
entire route table on a media card. Like a Cisco router. :)

The GRF and IP Navigator can hold the entire Internet IP forwarding table
locally on each media card. This means they can look up any route in about
1 microsecond. But yes, the total number of routes is limited by memory.
However we have more than enough for our customer needs.

>My guess is the second one?
>
>Another point is routing processes, like OSPF. Incidentally, is the
>implementation of OSPF in IP navigator better than the Max 4k
>implementation seems to be from comments on the list? (I have not
>yet tried Max 4K OSPF myself out of fear!).
>
>Sounds like a big undertaking to implement all the routing protocols
>in such a way that multiple independent copies thereof can be run on
>the same router, one for each virtual router. Has Ascend really
>pulled this off nicely?

It seems to work very well, even at high speeds like OC-48/STM-16. FYI, the
IP  Navigator OSPF was designed and implemented primarily by John Moy who
also invented OSPF.

Matt Holdrege		http://www.ascend.com	matt@ascend.com
++ Ascend Users Mailing List ++
To unsubscribe:	send unsubscribe to ascend-users-request@bungi.com
To get FAQ'd:	<http://www.nealis.net/ascend/faq>


Follow-Ups: References: