[Bioclusters] Anyone interested in clustering transmeta cpus?

chris dagdigian dag@fedayi.sonsorol.org
Fri, 24 Aug 2001 17:22:24 -0500 (EST)

Hey folks,

RLX technologies just dropped by to loan us a chassis for the next few
weeks. These are the folks who can fit 24 full blown servers in a 4U rack
by means of a bladed system design. They can do something insane like 336
CPUs in a standard datacenter rack.

Check out the website www.rlxtechnologies.com for more info on their
stuff. They have very impressive CPU density and absolutly beautiful power
consumption and heat output characteristics. 

My chassis only has 12 blades and I'm going to be trying to wrangle it
into a Blackstone-style compute farm running Sun GridEngine over the next
week or so. I've got 2 basic goals in mind: (a) evaluating it as a general
high density platform for use on Blackstone projects and
(b) characterizing how the system performs when running standard
bioinformatics applications off the onboard IDE drives. 

Right now the bladed systems have 2 main drawbacks that I see:

o 512mb SDRAM max on current blades (next product iteration has 1GB DDR)

o 4200RPM laptop IDE disks on each blade may mean relatively slow I/O for
things like blast searching etc. etc.

o No possibility of a PCI slot; this rules out Myrinet and other high
speed interconnect technologies

I took some pictures of the hardware with a digital camera and can throw
them up on the web somewhere if people are interested. I'll have more
detail once I get a chance to get hands-on.