Running the main server on the generator (UPS connected). Gotta love DTE. Was down this morning, came back, of course lots of work to get done. Then we heard a big bang. Lights flickered and went down. Oddly, one of our circuits is live. The rest aren’t but one is. Found it as the coffemaker light was on. Sheesh. Will probably set up a backup site on the same machine that runs this site.

Been researching this more. The questions I am asking now are, are the MTBF numbers believable? Are their bad batches of NAND chips … SLC, MLC? What failure rates do people see with SLC? We have seen failures in both SLC and MLC units. MLC is generally indicated to be less reliable than SLC. I am specifically looking for failure information. What I am finding is concerning me. Generally, among all controller chips out there, there seem to be a number of people reporting sudden failures in 2-3 month windows.

This is interesting. This is a new line of business for Attachmate, they aren’t in OSes, NOS, and other things directly related to this. I don’t quite understand the rationale behind the acquisition. I need to look at that one more. Best guess is that Attachmate is really a holding company with a loose connection of holdings, with not significant overlap. In conjunction with this, Novell sold 882 patents to Microsoft.

So I wonder if we should be thinking that SSDs shouldn’t fail as often as spinning disk. That is, SSDs don’t have moving parts, and so are much less subject to mechanical wear and tear as they are used. But they do fail. Every brand of SSDs we have used, every one, including Intel, Corsair, RiData, Mushkin … every one, we have seen failures. Some have been absolutely ridiculous in scope (Corsair), some have been mostly due to changes in their mechanical design (RiData) as well as unit failures.

Turns out Centos isn’t quite exactly equivalent to Redhat, as far as DKMS goes. Don’t ask me why, I am having a hard time figuring it out right now. We are trying to not re-invent a wheel and use the Dell developed DKMS system. We want driver rebuilds to trigger on kernel updates when needed. But … while it works correctly with a simple dkms.conf on Centos 5.5, the same thing doesn’t seem to work on RHEL 5.

[updated] Its up on InsideHPC, and showing up on the PR sites now. One of these days I’m gonna learn to do this stuff earlier … … looks like it didn’t make it out some of the PR and news sites (I submitted it late to InsideHPC, and Rich Brueckner was positively inundated with bits, so its possible it may show up later). So the PR is here, and we’ll have the reseller/contact list up shortly.

[update after the fold] Having just been through travel to SC10, and been put through the body scanner, the metal scanner, but thankfully no pat downs, I did see some of the more … aggressive … inspections in the beginning phases. Bruce Schneier has a long post on this, including many many links. I am not sure I want my family to go through this. I am not happy with it, and to be frank, I remain unconvinced that this is better than the null hypothesis, that is, not doing these invasive searches.

Yes, its old news now … announced Tuesday morning, and this is Saturday. Isilon, a maker of scale out NAS units (which hadn’t been spectactularly profitable … they had some issues in the recent past) has growing business in Bio-IT and other areas. They are a player in HPC storage for clusters. EMC hasn’t really been a player in HPC for a while. In the past, some groups have tried to use EMC as the storage provider for HPC, but the economics and performance aren’t a good fit for most of their product line.

I’ve been waiting a while to post this, to see how things develop. Lustre does indeed have a future. The question is, will Oracle cede control over Lustre, or will it be forked by OpenSFS/WhamCloud/Xyratec ? A few short months ago, its future was cloudy at best. Oracle isn’t seemingly interested in HPC, except where it matters for the database side of things. So most things HPC specific (with little possible alternative use cases) have been given the heave-ho.

Reports were in that a Windows based system was able to crack the 1PF barrier, but that the same system running Linux, was faster. Cudos to Microsoft for this … but I have to ask … really … if this statement from Bill Hilf is true: then why is Microsoft competing in the stratospheric regime of performance if its not trying to be there? I see a fundamental disconnect between actions and words.