Checkpointing?

[ftpd]

31-3-2011 17:48:21 RNA World Restarting task cms_GA-p[EZ-Lin64s]_1_Enterobacter-sp.-638_CP000653.cir.EMBL_1301033223_12378_0 using cmsearch2 version 31

I have read somewhere that there is checkpointing when using the new application.
Is this also for this wu? It was processing alsmost 40 hours and start all over now after reboot.

Correct?

Ton (ftpd)

[Torbjörn Klatt]

Unfortunately, the check-pointing is only available for some few work units of the new app yet. As soon as the older work units are finished, the new will all have check-pointing.

Cheers,
Torbjörn

[mxplm]

Right, you do not have checkpointing in every WU. This is because the checkpointing we do is only achieved by grouping several very short tasks into one. This means that all longer tasks are still without checkpoints.

But as far as I know, this will also be the case in the near future, so long running will still reset when rebooting.

[Michael H.W. Weber]

To say it clearly: Except for Linux 32 bit machines for which memory randomization is deactivated in the kernel, RNA World still does not offer a true checkpointing mechanism. What we will have available is a simple bundling of small sub-tasks in a single WU. After finishing each of the bundled taks, RNA World will continue with the next task in that bundle once the machine is shut down, crashes, halts or whatever - so it does not restarte from the very beginning of that task as it was before. Looking at this from the outside appears as if these WUs have checkpointing and indeed they do continue exactly with the next sub-task in that bundle. Nothng more, nothing less.

So, do not come to the false conclusion that now it is safe to restart your machine at any time for any WU we have in the system, because only for the CMSEARCH S application, this bundling applies and evenhere it won't for all of the WUs. You can make out the difference by checking whether the WU name contains an indication ofthe Rfam family or not. Those WUs that do not show this Rfam family name, are the bundled ones containing "pseudo-checkpointing". All the others will restart from the very beginning once the machine is shut down.

Michael.

[ftpd]

Dear all,

Thanks for exploination!

I will not reboot anymore!

Keep on crunching!

Ton (ftpd)

[Norman]

cms_GA-p[EZ-Lin64s]_1_Enterobacter-sp.-638_CP000653.cir.EMBL_1301033223_12378_0

the first numer ( here 1 ) is the number of the wu in a archive