[nsp] Another April Fools

Zach Wilkinson ztwilkinson at yahoo.com
Tue Apr 1 13:31:25 EST 2003


I have a weakness for marketing hype.

But seriously, the key phrase here is "Free Eval"
which any sales rep would be happy to indulge you with
given the potential PO value.

Hell, my Cisco rep gave me an indefinite eval on a
6506 with only one sup and one ATM/OC12 knowing that I
was only going to buy one.  The 52% discount was also
predetermined.

Other list members:  Can you answer these
questions/concerns?


--- dre <andre at operations.net> wrote:
> On Tue, Apr 01, 2003 at 12:31:47PM -0800, Zach
> Wilkinson wrote:
> > Anyone looking for this level of performance moved
> to
> > Force10 a long time ago.
> 
> How can you back this statement up?  I've long been
> considering
> doing this, but the ability to get internal approval
> to even test
> this hardware is an exercise in madness, let alone
> futility.
> 
> My biggest qualms with the 6500/7600 platform
> including a long
> list, here's some:
> 
> 1) Pinnacle COIL ASIC's are poor, and have numerous
> production
>    affecting bugs.  These bugs are the poorest in
> the history
>    of networking equipment, maybe even including
> Cabletron (but
>    maybe not as bad as PMPE's).
> 2) NetFlow on SUP1A doesn't support sampled, doesn't
> support ASes,
>    and doesn't support interface (in/out) or nexthop
> firelds.  It
>    also requires v7 on the SP and v5/v8 on the RP,
> and accounts
>    the Netflow packets themselves unless you are
> running hybrid
>    and export on the sc0 interface (not an options
> for SupIOS
>    which is the Cisco IP routing platform
> recommendation.
>    Netflow on Sup2 misses flows (i.e. doesn't
> scale), and the
>    Sup720 appears to have similar limitations. 
> However, it will
>    support SNF (sampled), I believe.  It also
> doesn't support full
>    Netflow fields, but does support ASes and
> interface/nexthop.
> 3) Sup2 and Sup720 both have FIB limitations that
> are approaching
>    unrealistic.  Sup2 doesn't support uRPF for
> production needs.
> 4) The BGP policy language is sub-optimal, but this
> is an IOS
>    limitation, not just 6500/7600.  However, no
> Cisco routers
>    will let you debug (or remotely get) BGP updates
> at CPU rate.
> 5) Speaking of CPU rate, same goes for processing of
> packets,
>    especially when combined with ACL's, CAR
> (however, the new CAR
>    features under Sup720 do look nice).  This is a
> long, outstanding
>    issue for Cisco in general, as well.  E.g. ACL
> change affects LMI,
>    or BGP Scanner affects everything.
> 6) Sup2 has only 16 HSRP groups due to a hardware
> limitation
> 7) The whole way it "splits" packets apart is just
> silly and likely
>    a poor architecture decision.  It just seems like
> a hack, something
>    that never should have been done.  Sending the
> header and the payloads
>    in different directions, across separate
> backplanes is by far the
>    most strange L3 switch architecture I've head of
> yet.  This is their
>    "killer" app, or at least Cisco sees it that way.
> 
> What makes Force10 good/better?  Do they solve any
> of these problems?
> There are a few other things I like about 6500/7600,
> but I want to focus
> on the things it can't do.
> 
> dre
> 


=====
Zach Wilkinson
CCNA MCSE-2000
ZWI-ARIN

__________________________________________________
Do you Yahoo!?
Yahoo! Tax Center - File online, calculators, forms, and more
http://platinum.yahoo.com


More information about the cisco-nsp mailing list