[c-nsp] BFD bug in IOS SXF6

[Jared Mauch]

On Fri, Jan 12, 2007 at 06:24:03PM +0000, kevin gannon wrote:
> Is this not what the QNX based IOS-XR brings us ?

	There is a lot of weight that comes with that as well.

> On 1/12/07, Anton Kapela <tk at 5ninesdata.com> wrote:
> >
> >
> >
> >
> > > What we really need is some kind of supervisor module with a
> > > much faster CPU. Like the rsp720, only supported in the 6500.
> > > Grumble 7600 split grumble.
> >
> > I'm not sure 'more' cpu to run the same code is the answer. I think (and
> > I'm 100% sure Cisco DE's know it, talk about it a lot, and likely cannot
> > garner enough internal interest for) the need is either some form of
> > multi-tasking RTOS instead of the best-effort, cooperative,
> > run-to-completion-with-some-hacks multitasking that IOS offers today, or
> > perhaps dedicated hardware and appropriate pre-emptive ipc/signalling
> > from linecards to rp/sp components.
> >
> > Apparently the cost/benefit tradeoffs aren't in favor of some flavor of
> > RTOS being pushed down to boxes that a) unsophisticated enterprises and
> > b) ISP's are buying (cat6/7600). I would wager, if there was significant
> > demand for features that depended on real-time features from enough
> > customers with enough potential revenue on the line, this would have
> > been solved in some substantive way years, maybe decades ago. Anyway,
> > while c-nsp discussions are fun, a bunch of us bleeding-edge, 'lets make
> > it better' types aren't likely to represent the majority of Cisco's
> > customers. I won't get stable high-rate ospf hellos anytime soon, nor
> > will any international 'backbones' get stable and dependable BFD in the
> > short term: our special needs simply don't matter enough.

	I suspect the next major SX* release will help clean this up
as it's my understanding it will be the modular OS, which allows another
form of scheduling to happen.

	I would not say that isps are buying the '7600', while some
are and have, i suspect the bulk of folks are buying the 6500
and adding in DFCs as necessary as opposed to a large SIP/SPA/OSM
deployment.  (i'm obviously willing to be wrong).

	- jared


-- 
Jared Mauch  | pgp key available via finger from jared at puck.nether.net
clue++;      | http://puck.nether.net/~jared/  My statements are only mine.