About me

My name is Joe. I used to run a small revenue-backed high performance storage, cluster, and accelerated computing company named Scalable Informatics. We did all sorts of neat things like build really fast clusters, very fast and reliable storage systems, acceleration systems and software, tune code, parallelize applications, but you can follow the link if you would like more info. This blog isn’t about the company.
I do what I can to maintain a positive outlook and a sense of humor. It is important not to take yourself or your market so seriously, or at least not all the time.
Please note that if you read the rest of this, that my sense of humor is either good, or badly flawed. I know I need to keep my day job.
My background … A little physics, lots of programming, embedded measurement stuff on fun and noteworthy experiments. Running on supercomputers of all shapes, sizes, architectures from about 1986 onwards (does an IBM 3090 VF180 really count as a supercomputer?). Spent some time working at some small blue computer company at one of their research sites in my home town after completing my undergraduate degree in physics. Went to another school out in the midwest area. Thought they were doing cool things there. Got an MS in physics. Somehow convinced myself I wanted to be a professor. Maybe it was the life style was so cool. I dunno. So onwards and upwards. Find a school and thesis advisor studying stuff that will be useful in the future. Try to get ahead of the wave. So, since Silicon is going to run out of steam, lets study Gallium Arsenide.
If I only knew then what I know now. The running joke about Gallium Arsenide (yeah there is a running joke, and no, you don’t need a PhD to understand it) is that it is the material of the future. And always will be.
Sometime in the early to mid 90s, while running on supers at the PSC, I realized I liked working on the computers more than the physics. This was a great source of professional confusion for me. How could one be a physics professor if one preferred to work on writing faster Symplectic integrators as compared to working on the things that the integrators were used for?
I solved the problem by doing the only sensible thing. After getting to the writing stage (ABD), I joined SGI. Ok, so six weeks after I joined, the stock started its terminal and monotonic decline. I claim no part of that was my fault. Really. I didn’t do it. Nosiree.
I finished up the PhD while at SGI. I was doing something else which was demanding, while I was doing the demanding completion/thesis writing/defense/… Yeah. Not fun.
At SGI I did a few things. Got an “At-a-boy” for developing an automated way to load operating systems over the network. All you had to do was point it to a bootp server and have a working dhcp server, and off it went. Gave the code away for free. In 1996 before this was considered good. SGI liked it so much the rewrote it into Roboinst that still graces Irix today, though I don’t know too many shops still running Irix. What was funny was that at the same time I was working on this, I was finishing my dissertation, and defending my thesis. Owie.
I developed some cool web based molecular visualization tools. You uploaded a molecule file in almost any format, and out popped a VRML file. Which you could manipulate in your browser. 1998-ish. Cool stuff. In 1999-2000 I wrote ct-blast later called SGI GenomeCluster . First commercial code to parallelize NCBI BLAST as far as I know. A few other competitors were announced shortly after we started showing it at GSAC 2000. SGI pulled the rug out from under us by getting out of the cluster business for the first time in 2001. Thats when I left. I thought clusters were going to be the next big thing.
Joined another company. I should have waited a month, taken a layoff package from SGI, and started Scalable back then, but, well, I was chicken. And again, not my fault. SGI stock dipped to about $0.30 within weeks of my leaving them. Really I had nothing to do with this.
Spent 16 months at this other company. Building a product. Building a team. Getting a business and marketing plan together. Getting in front of customers. Getting to the close. Only to watch the … sales management … well … not close. They started having problems, the bubble had burst. I got a package. Small one. Tiny really. Barely enough to start …
… Scalable Informatics.
But this is not about Scalable. This is about my background. At least this document is. The blog is about HPC. All of it, technical to business, marketing, to measurements.
This blog is a way to have a series of conversations with large groups of people, and to encourage discussion. It is not a cathartic release or anything like that. I want to encourage and promote discussions of all aspects of HPC. Not just whats on my mind. I want to hear what is on yours.
I am quite fond of the high performance computing business. It is a business, and lots of other non-technological things happen to impact it. Marketing for one. And business operations. And technological shifts. And … well you get the idea. The point is that while this is about high performance computing, its about many things related to high performance computing as well. Such as running a high performance computing business. Trying to understand new products and where they fit. Trying to raise capital to build new products. Seeing who has raised capital and what they are building.
HPC is interesting as a market. Been around several decades, been growing rapidly. Clusters have been around for more than a decade, though a discernable market for them appears to have gone into high gear around 2000 time frame.
Where is HPC going? Good question. Will anything replace clusters? Probably. Best estimate is APUs, and they aren’t going to replace as much as augment clusters. We have a business plan with some great prototype benchmarks and a number of designs. Will personal supercomputing ever arrive? I have been talking about it since around 1999, so I hope so. What is the airspeed of an unladen swallow? African or European (you have to know these things if you are in HPC).
But seriously folks … The one lesson you need to take away from the history of HPC is that it is always going down market. It is getting cheaper and faster and better. Anyone who wants to build an HPC empire needs to look carefully at the carcasses of the companies that have done this before. These empires are fleeting. They are not long lasting. Once you get into the “protect the turf” mode, you are a “dead company walking”.
This market is changing. Rapidly.
Finally, I do write em as I see em. If you think I see em wrong, tell me. Write a comment back. Nothing wrong with snarkyness or poking or proding. Irreverence, especially if done with style and panache will get you extra points. Points are lost for flame bait, trolling, and similar. Disclaimers at bottom of the pages. If something offends you, let me know. That is not our intent. We want to inform, we want to discuss. We don’t want to annoy.

4 thoughts on “About me”

  1. Yellowdog Linux will run on the card. Its a little bit of an involved process to get it going. You’ll need the serial port up and working, or it will never boot.
    Contact me at the day job and I can put you in touch with the people who supported it. Not sure its really worth it, unless you are in it just to play with it.

  2. Hi Joe, I am from the Garvan in Sydney. We are currently testing our newly deployed NFS over 10GbE. I am very interested in the io-bm.exe you used in many of your tests. I am wondering if it’s released to the public so I can use it as well?

  3. Hi Joe, I’ve found your article at Linux Magazine “MPI in Thirty Minutes”. I’m tried to download the example code, but sadly links are broken. Please could you send me or post the link if still exists?
    Thanks in advance!!

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.