The day job has a registration page for people who are interested in purchasing high performance storage, deskside supercomputing, storage clusters, and other HPC like things from us. This registration page asks some very simple things: who are you, where are you, what your shipping/billing address is, what you want your user name to be, and what things you want information on. Registration is, fundamentally, a matter of trust. We guarantee we will not spam people.

… well, I am trying to figure out what I am doing wrong in io-bm. I need a new method to defeat some of the smarter caching bits, my MPI_Send/MPI_Recv pairs are blocking pairs, and this impacted performance. Not only that, the additional traffic over the Infiniband was definitely a cause of contention on the wire. Doing some TB sized writes at a good rate. The “naive” bandwidths (the way IOzone calculates them) are about where we predicted given the measured IB performance.

This is more of an itch I need to scratch. I am a recovering/reformed computational physicist. I really enjoyed doing work in modeling semiconductors, and I had hoped to post-doc modeling dynamics of proteins among other things. Of course, my academic career ran head first into the deluge of physicists from the former soviet union, all with 20+ year seniority, all willing to work for less money. A generation of young physicists were lost to this onslaught; I decided to do something else after finishing up.

From this article linked from /., I found this tidbit Heh. I wonder if anyone else used that term, for accelerators, before they did. I wonder. APUs are taking HPC by storm. This is creative destruction you are witnessing. We moved mostly out of the market being destroyed over the last year or so, and focused upon the market being created. It absolutely blows me over that Vipin’s and my strategy pitch is being played, almost to the letter, by the successful players in this market.

It looks like IBM is going to shelf the Power Cell processor. Cell during its lifetime never really garnered the ubiquity it needed to do what NVidia is doing with GPU. I had guessed on this site previously that Cell needed to get wider distribution to maintain a base. The business model for acceleration is ubiquity, and then its a tools play. Unfortunately IBM never really seemed to commit to the platform.

… until I need to use it. Anaconda is the Redhat/Fedora installer. It purports to be a reasonable installation tool. But it has a number of interesting issues. Some of these issues make installs … well … exciting. I’ve taken to the philosophy of absolute minimum time spent in anaconda. Call this defensive installation. Anaconda will toss fatal errors, with no hope of recovery … unless you want to try and debug some obscure python … Back in the SGI days, I wrote an installer that largely worked around the SGI installer.

Back from SC09, and for a project Vipin asked me to work on quickly, I am cranking out our roadmaps in greater detail. One of the things I’ve been thinking of for a long time is, what comes after JackRabbit and DeltaV? In the case of JackRabbit, even when it is hobbled by a poorly performing IB network (we are still working on why this is the case), we appear to kick some serious tail in the high performance cluster storage space … our worst case result was 4x faster than a competitors best case on the same problem (info was presented at SC09 in a public talk by the user, so I think its ok to talk about, if not, let me know and I’ll elide this).

From HPCwire … Like … totally … Ok … apart from the humor in the quote (and I am hoping that Allan or the writer meant that comment to be interpreted in a semi-humorous manner … its also very possible Allan didn’t say that and the writer took … er … liberties … yeah thats it … and decided to embark on a more, how shall I say this … creative writing effort than more serious journalism … embellish it a bit), there is another thread that is worth discussing.

Minnesota Supercomputing Institute purchased the first siCluster (PR on disk, finishing up and getting out tomorrow), which is a scalable storage cluster product, aimed at providing very scalable performance and capacity. I was worried after the talk I gave at their booth. Their researcher indicated our performance wasn’t good. We had turned off some caching to avoid problems during the acceptance test for HP, the Itasca cluster vendor. The Itasca cluster is quite nice.

Ok … so so many things going on. All at once. I’ve done what … like 3 on camera interviews over the last 48 hours, and have another one coming up. I’ve given a talk, and attended a talk. More on that in the next post. Ok. Vipin needs me to work on something tonight, so I might not get nearly as much sleep as I want. Gotta hit the gym tomorrow too.