Features:

WHAT I DID DURING MY VACATION AT SUPERCOMPUTING 2004
by Glen Otero, Contributing Editor

The annual Supercomputing conference is always a great time for me, but this year was especially fun. Despite the funding currently flowing into the HPC and Linux clusters space, my over-exuberance this time around is largely due to working for a company that is doing revolutionary things in HPC. Added bonus: they didn't need me to stay in the booth for very long, so I got to cruise the floor in my usual persona--that of freelance cluster conspirator and seeker of Guinness. This modus operandi allows me to visit with old friends, chat with colleagues, meet new people, and get beyond the hype to survey the true Linux cluster landscape. Here's a journal of my antics and what I learned.

Dear Diary:

Monday 11/8/04

Up at o'dark-thirty to catch a plane to Pittsburgh. Down a quad espresso in the airport, which takes the morning edge off, and I promptly fall asleep upon takeoff. Sleep most of the way to Pittsburgh. Finally make it to the Renaissance hotel downtown, which is very nice. Drop my bags and head to the David L. Lawrence Convention Center. It's 5 pm.

Monday Night-7 pm

Hanging out during the opening gala, completely thrilled that I wasn't responsible for pulling off some incredibly complex demo like last year when I did two of them spanning four different booths (Yikes.) Food lines were too long everywhere and the beer too expensive (and not Guinness). So I hooked up with the folks from ClusterWorld and headed over to Mark's Grille at 9 pm for the Ligon/Eadline Cigar and Cognac Invitational Beowulf Gathering, aka LECCBIG. Indulged myself with exquisite cigars and several pints of Guinness with ClusterWorld editor Doug Eadline, publisher Adam Goodman, and ClusterWorld contributor and cluster maven Jeff Layton. Discussed cluster world domination and the state of cluster software. Thanks go out to Tim Wilcox who was a fantastic host. Thanks for all the great cigars and taking care of the ClusterWorld bill Tim!!

Closed down Mark's Grille and headed back to the Renaissance. Plotted cluster workstation world domination with some Orion colleagues in the lobby and called it a night at 2 am.

Tuesday 11/9/04

Chugged a quad espresso at Starbucks to smooth out the edges. It's 8:45 am.

Attended Wolfram Research's Mathematica workshop at 9 am. Wolfram was showcasing gridMathematica 2 (announced at the show), which allows users to utilize Mathematica's rich technical computing environment to program and run parallel applications over SMP machines, clusters and computational grids.

Spoke with the smart folks at Panasas. I've been working with Panasas since SC2003, and all I can say is, Wow! You may have heard of all the awards Panasas has garnered this year, and the substantial placement of their cluster storage products into the oil and gas industry, but did you know that they've already shipped over 1 petabyte of storage? They're also going to be making a software development kit available for their parallel filesystem that will include hooks to MPI-ROMIO. They've already donated the MPI-ROMIO patch to Argonne and so it will appear in the next release. Panasas founder and CTO, Garth Gibson, has had a key role in the development of version 1.0 of the Object Storage Device (OSD) specification, which is intended to improve scalable, cross platform data sharing. While the uninitiated think storage is just another simple commodity cluster hardware component, Panasas is helping those of us that understand cluster bottlenecks by leading the way with much needed cluster filesystem technology. Billy-Bob's JBOD Outlet they are not. To sample and mash a Beastie Boys track--Check, check, check, check it out! Find out what Panasas is all about!

I never get tired of walking by the Panasas booth, which is right next to Orion's, because IMHO, the most beautiful and interesting woman in all of HPC works at Panasas. OK, I am now officially embarrassed, and it's only Tuesday.

Had a scheduled interview with Scyld and Penguin Computing. I met with Don Becker, CTO and founder of Scyld and builder of the first Beowulf, Virginia Cartwright, VP of Corporate Marketing for Penguin Computing, and Marco Annaratone, VP and GM of Scyld, to discuss the future of Scyld and Penguin Computing. While several HPC cluster companies are getting money right now, Scyld explained to me that they are going to channel resources from their funding round into further innovation of the Scyld cluster OS. I was very glad to hear that Scyld has committed to actual hard-nosed cluster OS innovation, since I can't swing a dead cat without hitting another cute, redundant browser-based cluster management and monitoring tool that I don't need. It was exciting to hear Don talk about having completed software projects at Scyld that we had talked about collaborating on years ago.

Tuesday night

Beowulf Bash at Mahoney's on Tuesday night 6 pm. Grabbed a Guinness and continued discussions with Virginia, Matt Jacobs, VP of Sales at Penguin Computing, Marco, Don, and Enrico Pesatori, Chairman and CEO of Penguin Computing. Enrico and Matt confirmed that Scyld would be adding important features to their unique second generation Beowulf cluster operating system as well as targeting specific enterprise customers and vertical markets.

Crashed the Cray party at the Westin in search of Guinness. Discussions with Roger Smith and Joey Jones from the ERC at Mississippi State about the real problems with clusters (other than users : ) and flame bait on the Beowulf list. I learned about all the HPC work going on at ERC and the rest of Mississippi. Exciting stuff going on in that state.

Left the Cray party and tried to find a bar that was open at 11:15 pm to no avail. Apparently they roll up the sidewalk at 11 pm in Pittsburgh. Sheesh. Even San Diego, which is lame on the party scale, has bars open past 11 pm on a school night. Got back to my hotel in time for the bartender to include me in last call, so I ordered 3 pints of Guinness (yes, for myself). Rehashed cluster workstation world dominance schemes with some Orion colleagues. Called it a night at 2 am.

Wednesday

Chugged a quad espresso at Starbucks. Edge subsides. It's 9:45 am.

Visited my friends over at the Texas Advanced Computing Center (TACC) booth. TACC has become one of the leading TeraGrid sites under the leadership of Jay Boisseau. Got a very impressive live demo of GridShell by Edward Walker. GridShell is the coolest tool I've seen in a while, and what the rest of us needs to access the TeraGrid, or any computational grid for that matter. GridShell is a shell environment, like tcsh and bash, that has grid services mapped into it so users can access grid resources transparently from the command line. GridShell has no restriction on the types of commands that can be remotely executed, it requires no changes to the application or the underlying distributed/grid infrastructure, and it affords a high level transparency to the distributed resources. GridShell and Grid Portal Toolkit (also from TACC) are tools that I'm definitely going to play with in the near future.

I kept running into my Western Scientific homeboys (we're all located in San Diego) throughout the show. If you can introduce me to a smarter and more fun group of guys, I'll buy you a Guinness (or a lesser beer of your choice).

I almost broke my freakin' neck spinning around to see if Jeff Johnson, Western Scientific's VP of Engineering, was serious when he told me that Terrascale's parallel filesystem software, TerraGrid, is easier to install and maintain than GFS. I used to sit on Sistina's technical advisory board, so that really got my attention. Jeff was serious. He added that the TerraGrid global parallel file system provided a single name space and delivered linear I/O scalability over Gigabit Ethernet, Infiniband, and Myrinet. Unlike many projects I ran across at SC, TerraGrid is deployable now. In fact, Tim Wilcox was walking around the show with Terrascale's software on a USB key, installing it on any server people wanted. Could this be The One? The parallel cluster filesystem that will alleviate my need for NFS on every cluster I build? One can only hope.

Note to self: Download the whitepaper and visit Western Scientific for a demo of TeraFS.

Clearspeed sponsored the most imaginative event during the show by hosting talks by David Turek (who filled in for Mark Seager of LLNL) and noted futurist Ray Kurzweil.

Pretty cool that ClearSpeed's President, Mike Calise, mentioned Orion during his presentation as a company on the cutting edge of HPC. Turek, IBM's VP of Deep Computing, gave a talk that was an excellent synopsis of IBM's Blue Gene/L project and the general low wattage direction HPC is headed in order to overcome the physical restrictions of power and cooling and their associated costs. Mike clearly explained that Clearspeed, IBM's Deep Computing, and Orion share a common low wattage/small footprint supercomputing initiative.

Kurzweil's talk was great. Yes, I'm a fan of his, and yes, I did receive an autographed copy of his book. But I am rather skeptical of some of his predictions and I had some rather important questions about his depiction of nanobots as immune system soldiers. So I didn't really appreciate the two gentlemen that selfishly consumed all of Mr. Kurzweil's time after his talk with childish questions about good versus evil, and citing the recent presidential election as proof. Are you kidding me?! Adding insult to injury was the fact that besides myself, there were some folks from Wolfram that had very real questions that we never got a chance to ask. But it was open bar, so I let it go.

Clearspeed's event didn't have Guiness, so I crashed the SGI Party at the Westin in search of that particular nectar. While I liked the casino idea, I don't think I've participated in a more annoying raffle. It was obnoxious, unimaginative, and interminable. At least one of my colleagues won some champagne and SGI gave away an iPod at the end.

Found some folks I had been looking for all week. They bought me Guinness, so we were able to have insightful discussions on the problems facing further HPC adoption. Kicked around several ideas.

Called it a night at 2 am.

Thursday

In search of coffee...my cruel mistress. Chugged a quad espresso at Starbucks. All is well. It's 8:15 am.

Made a local sales call for Orion in the morning.

Talked to the most beautiful woman in HPC at the Panasas booth. Life is good.

Met with Tim Curns of HPCwire to tell him that I was actually going to write this article. He says OK. Blame him.

Officially cursed SciNet for the horrible wireless coverage during the show. I had to go to the Press room every time I wanted to check my email. Lame.

Tim Wilcox came by the Orion booth and had me drooling with the performance numbers he said we could achieve with the TerraGrid cluster filesystem technology.

Packed up the Orion booth and left in search of Guinness. Skipped the StorCloud party. Had a very tasty burger and some Guinness at Olive or Twist across from the hotel. Caught a plane that had many familiar faces on it back to San Diego at 8:45 pm. Home by midnight.

Life in the fast lane? Hardly. Gonzo journalism? Not even close. But this is as loose and fast as the HPC crowd gets.

Glen Otero received his Ph.D. in Microbiology and Immunology from UCLA in 1995 and immediately escaped to the more temperate climes and better surf in San Diego. After some research on the molecular and cellular biology of HIV and Herpes viruses at the Salk Institute for Biological Sciences, Glen left the wet lab research bench in 1999. Although leaving the research bench, he didn't leave science altogether; traveling all the way across the street to the San Diego Supercomputer Center (SDSC) for a stint at the Protein Data Bank. It was while at SDSC that Glen had his Linux clusters and bioinformatics epiphany. Soon after that illuminating event, Glen founded Linux Prophet, a bioinformatics consultancy specializing in the implementation, design, and deployment of Linux Beowulf clusters in the life sciences. Late in 2002 Linux Prophet evolved into Callident, a Linux cluster software and high performance computing company.

Callident recently partnered with Orion Multisystems, providers of 96-node Deskside and 12-node Desktop Cluster Workstations.