What is the limit? (was RE: multi-homing fixes)

Roeland Meyer rmeyer at mhsc.com
Thu Aug 30 03:48:49 UTC 2001 

|> From: Vadim Antonov [mailto:avg at exigengroup.com]
|> Sent: Wednesday, August 29, 2001 3:45 PM
|> 
|> On Wed, 29 Aug 2001, Roeland Meyer wrote:
|> 
|> > Sorry, Leo is correct. Technologies he outlined are only 
|> the tip of the
|> > ice-berg of what *isn't* being exploited by the router vendors.
|> 
|> Your average PC doesn't have to be NEBS-compliant, doesn't 
|> have to work
|> more than 24 hours w/o crashing, and doesn't have quite 
|> strict constraints
|> on power & heat dissipation.  It doesn't have to have 
|> redundant power, and
|> its components are readily available and cheap (those are produced in
|> _large_ batches).

Excuse me?!?! What universe are you living in?

I know a few racks full of 1U, 2U, and 4U PC's, with dual power, have been
running 24x7 for over a year, at five 9's, and NONE of them cost more than
$5K. Further, they have over 2TB of RAID. It's a componentized cluster
architecture.

see www.dell.com, www.penguincomputing.com, www.raidzone.com,
www.microway.com, www.aspsys.com, www.hp.com, www.ibm.com,
www.asacomputers.com, www.ixsystems.net, www.enlightcorp.com, and
www.einux.com.

Oh yeah, don't forget to look at www.routerrevolution.com.

Wake up, the coffee's brewing ...

BTW, RAIC = Redundant Array of Inexpensive Computers, usually clustered.

|> And don't get me started on heat and airflow issues :)  Reason #1 why
|> Pluris abandoned the original idea of using commodity CPUs 
|> was heat, not
|> the switching speed.

That was how many years ago? BTW, they gave up because what they hired for
wizards couldn't solve the problem (made religious vs technical decisions).
If it runs hot, either cool it, or get something that runs cooler. Today's
answer is a Transmeta chip, a handful of ASICs, a pair of ECC'd PC133 DIMMs,
and a couple of Sony memory sticks to replace the HDDs. Oh yeah, enough fans
to qualify for an FAA Airworthyness Certificate (only because fans are
cool)<g>.

More information about the NANOG mailing list