Post
by Bara » Tue Jul 07, 2009 5:19 am
I wouldn't say useless but extended periods of not running the client can make the work unit so late that it gets reassigned to someone else. I know from reading the PS3 FAQ standard computers processing on their main CPU are given more time since the work they are assigned is tailored to their general utility and flexibility.The example I remember is that what a PS3 can do in 2 calculations might take a standard CPU 20 calculations, but the PS3 is more limited in they type of computations it can do and the regular old Intel or AMD CPU can do a wider variety and complexity of calculations.
Some more info I dug out of the Genral FAQ:
What if I turn off my computer? Does the client save its work (i.e. checkpoint)?
Periodically, the core writes data to your hard disk so that if you stop the client, it can resume processing that WU from some point other than the very beginning. With the Tinker core, this happens at the end of every frame. With the Gromacs core, these checkpoints can happen almost anywhere and they are not tied to the data recorded in the results. Initially, this was set to every 1% of a WU (like 100 frames in Tinker) and then a timed checkpoint was added every 15 minutes, so that on a slow machine, you never loose more that 15 minutes work.
As proteins become more complex and run longer, it is better to have more frames in a WU so that you don't loose so much progress if you have to restart - - hence WUs that have 400 frames instead of 100. That still doesn't take the speed of the machine into account. A fast machine completes a frame in a few minutes while a slow one may take hours, and the donor with the slow machine still doesn't want to lose 99% of those "hours" yet the fast machine doesn't really want the overhead of writing the checkpoints every "few minutes" - - and neither of them wants the upload time associated with results containing many frames.
Starting in the 4.x version of the client, you can set the 15 minute default to another value (3-30 minutes).
Are there any limits to how long my machine can take to finish a work unit (WU)?
Yes. Work Units are serial in nature. When a completed WU is sent back, a new work unit is generated from those results. This must happen many times over within each project (group of work units). A generation 1 work unit must be turned in before a generaton 2 work unit is created and sent out.
To keep these generations moving along, we have to set expiration deadlines in the event a work unit is not uploaded in a timely manner (lost, deleted, whatever). These unfinished work units "expire" and are reassigned to new machines. You will still receive credit for all WUs completed and uploaded prior to the preferred deadline. However, after the preferred deadline, your contribution is not as useful scientifically because another copy of that work unit had to be sent out to another contributor. Even if you eventually complete the work unit, that other contributor still had to process duplicate work to assure the science moves forward. And it would be unfair not to also credit that second contributor.
Even so, full credit is given up until the final deadline. After the final deadline has expired, the client will automaticlly discard the work unit and download new work. If you have trouble completing work units before the preferred deadline, it is recommended to either run the FAH client more hours each day, or to run the client on a faster computer.
As we move to larger and longer WUs, we will extend the expiration time as needed. Deadlines vary on the order of a few days to a several weeks, depending on the nature of the WU. Turn in a work unit just before the deadline is not the goal. It is most helpful to the project to return work units as quickly as possible. And how these deadlines are determined is explained a few answers below.
2xGPU clients (GTX 465/GTS 250); 2xCPU clients
<Aura> Our army of teenage girl artists couldn't draw penises without fits of giggling. Does that count?
