From: Rik Guyler (Rik@Guyler.net)
Date: Tue Dec 17 2002 - 13:20:33 GMT-3
Bob, Cisco never intended a gigastack unit for a datacenter (core)
application. These units are intended solely for a stacked closet
environment. As to why use them, Chuck is right on the nose when he said
price. Gigastacks are about half the price of a fiber GBIC and offer twice
the port density. However, further proof of their misappropriation in the
core is the fact that when switches are cascaded using gigastacks and both
ports are in use, the links between upstream and downstream switches goes to
half-duplex, definitely not what you want in your datacenter. I do agree
with you though - I prefer the chassis-based solutions. The client doesn't
always agree me... ;-}
One thing: adding gigastacks doesn't automatically make your stack a
"cluster". Managing the stack as a single entity is considered a cluster
and is setup using cluster management options in IOS, either by using the
CLI or the Cluster Manager software via the browser. Personally, I never
use this option as I have had bad experiences with it.
Rik
-----Original Message-----
From: Chuck Church [mailto:ccie8776@rochester.rr.com]
Sent: Tuesday, December 17, 2002 10:23 AM
To: bobdu11@cox.net; Larry Letterman; RonaldFugate
Cc: Bob Sinclair; ccielab@groupstudy.com
Subject: Re: Re: Gigastack - What is the point?
Bob,
I think they'll run at 37% even without a trunk. But it's really just a
superficial thing. There's a page about it on Cisco's web site. Since they
do most everything in hardware, you won't see that rise much due to traffic.
Of course things like maintaining IGMP, the mac address table, and CDP are
processor-based. But I've yet to see one run at more than 50%. I'm glad to
see they did fix it in the 2950s and 3550s though. But I agree with Larry.
If it's a data center, I wouldn't be using stackables. If the goal is
keeping tons of users connected to multiple servers 24x7, I'd want the
bigger backplane, layer 3 capability, and redundant sups/power supplies of a
chassis. Just my .02 though...
Chuck Church
CCIE #8776, MCNE, MCSE
----- Original Message -----
From: <bobdu11@cox.net>
To: "Larry Letterman" <lletterm@cisco.com>; "Ronald Fugate"
<RFugate@amdocs.com>
Cc: "'Chuck Church'" <ccie8776@rochester.rr.com>; "Bob Sinclair"
<bsin@cox.net>; <ccielab@groupstudy.com>
Sent: Tuesday, December 17, 2002 5:38 AM
Subject: Re: Re: Gigastack - What is the point?
> Also, take a 35xx series switch, plug it in and connect it to another
switch via trunk and then take a look at the CPU Processor. Even with no
users attached to it and only running spanning tree it sits at around 37%.
The box needs alot of work....We use them but only because the cost savings
against a chasis base switch dictates, but from an engineering standpoint,
can't stand em !!!.....bobdu1
> >
> > From: Larry Letterman <lletterm@cisco.com>
> > Date: 2002/12/17 Tue AM 02:15:24 EST
> > To: Ronald Fugate <RFugate@amdocs.com>
> > CC: "'Chuck Church'" <ccie8776@rochester.rr.com>,
> > Bob Sinclair
> > <bsin@cox.net>, ccielab@groupstudy.com
> > Subject: Re: Gigastack - What is the point?
> >
> > We have a large data center for the engineering of Cisco at the main
> > Campus, where we have close to 1000 servers, and we wont even
> > entertain the thought of stackables in the DC...L2 does not scale
> > anywhere near the range of L3, and the port density of 35XX's does
> > not get close to a chassis based system....
> >
> > Ronald Fugate wrote:
> >
> > >in addition to that:
> > >
> > >In a datacenter that; where hundreds of servers (blade servers,
> > >usually
web environment) are required, the 3548's (or 3550 smi),in a redundant
(layer 2) and teaming nics for end nodes, these switches, with gigastacks,
are usually within a few feet of each other and are great. The gigastacks
offer more flexibility than fiber stacks. The gigastacks can stack to the
switches and leave the other gig slot open for other uses (trunks or gig
access ports whatever).
> > >
> > >In our datacenters the gigastacks were alot more resilient (taking
those unmentioned bumps from engineer running cables).
> > >
> > >And scalability is a big reason.
> > >
> > >
> > >-----Original Message-----
> > >From: Chuck Church [mailto:ccie8776@rochester.rr.com]
> > >Sent: Monday, December 16, 2002 8:59 PM
> > >To: Bob Sinclair; ccielab@groupstudy.com
> > >Subject: Re: Gigastack - What is the point?
> > >
> > >
> > >Bob,
> > >
> > > Price is probably a major reason. Last time I checked, the
Gigastacks
> > >are cheaper than SX gbics. Also, a lot of companies stick with
> > >2900
and
> > >3500s for closets. 4000s and up are considered distribution and
> > >core
level
> > >switches, with a price to match. Price per port is much cheaper
> > >for
2900s
> > >and 3500s than a 4006 with sup 2 and line cards. Since most
> > >networks
tend
> > >to grow rather than shrink, upgradibility is also a factor. Once
you've
> > >maxed out a 4003 or 4006, you've got a big cost to add another
> > >chassis. With stackables, it's much cheaper. Of course there are
> > >networks out
there
> > >that justify a 4000 or higher at the access layer, but those are
special
> > >circumstances.
> > >
> > >Chuck Church
> > >CCIE #8776, MCNE, MCSE
> > >
> > >
> > >----- Original Message -----
> > >From: "Bob Sinclair" <bsin@cox.net>
> > >To: <ccielab@groupstudy.com>
> > >Sent: Monday, December 16, 2002 8:53 PM
> > >Subject: OT: Gigastack - What is the point?
> > >
> > >
> > >>Switch gods:
> > >>
> > >>Any of you folks installed gigastack 35xx or 29xx? I really don't
> > >>see
> > >>
> > >much
> > >
> > >>of an advantage to this technology, so I wonder what I am missing.
Sure,
> > >>you can manage a bunch of switches with one IP address through a
graphical
> > >>interface. BFD.
> > >>
> > >>The fast failover and minimal uplinks would be cool if you could
> > >>stack multiple switches on different floors, but as I read the
> > >>specs, the
> > >>
> > >switches
> > >
> > >>must be within 1 meter of each other. If you need multiples of 48
ports
> > >>
> > >in
> > >
> > >>one closet, why not just use a modular switch?
> > >>
> > >>I have read the docs on CCO, but I don't really see what does this
> > >>technology really buys us, beyond a few corner cases. Any
> > >>feedback or
> > >>
> > >links
> > >
> > >>appreciated.
> > >>
> > >>Bob Sinclair
> > >>CCIE #10427
> > >>.
> > >>
> > >.
> >
>-----------------------------------------------------------------------
>----
----------
> > >
> > >The information contained in this message is proprietary of Amdocs,
> > >
> > >protected from disclosure, and may be privileged.
> > >
> > >The information is intended to be conveyed only to the designated
recipient(s)
> > >
> > >of the message. If the reader of this message is not the intended
recipient,
> > >
> > >you are hereby notified that any dissemination, use, distribution
> > >or
copying of
> > >
> > >this communication is strictly prohibited and may be unlawful.
> > >
> > >If you have received this communication in error, please notify us
immediately
> > >
> > >by replying to the message and deleting it from your computer.
> > >
> > >Thank you.
> > >
> >
>-----------------------------------------------------------------------
>----
----------
> > >.
> > .
> .
.
This archive was generated by hypermail 2.1.4 : Fri Jan 17 2003 - 17:21:47 GMT-3