Missing preferred deadline by 2 hours...

Moderators: Site Moderators, FAHC Science Team

Post Reply
imzjustplayin
Posts: 18
Joined: Thu Mar 13, 2008 7:23 am

Missing preferred deadline by 2 hours...

Post by imzjustplayin »

I was wondering, since I'm going to miss the preferred deadline by 2 hours, does that mean that if start crunching at 3:01am and the result I submit arrives at 5 am two days later, would I still be safe? How exactly does the preferred deadline work? Are ALL WU that exceed the preferred deadline automatically sent out again even if the time when a WU is requested is after the time the "overdue" WU is received?

Say I submit my WU 2 hours past the preferred deadline, for the one hour before the deadline and 2 hours after the deadline, no machine has probed the server for a WU, does this mean if somebody, say 30 minutes after stanford receives my WU, asks the server for a WU, will Stanford send the WU I submitted to Stanford or not? I mean if stanford has a pool of WU that have exceeded the "preferred" and or even "final" deadline, couldn't they send those out first, before they receive my WU 2 hours after the preferred deadline?

The preferred deadline makes sense if they're limited in the # of WU they can send out, machines are probing for a WU and any WU that has exceeded the "preferred deadline" is sent out since the machines are starving for work to be done. However, if there is plenty of WU ready to be sent out, those that have either exceeded the "final deadline" or those that simply have never been sent out before, then shouldn't they just accept the WU that exceeded the preferred deadline?
7im
Posts: 10189
Joined: Thu Nov 29, 2007 4:30 pm
Hardware configuration: Intel i7-4770K @ 4.5 GHz, 16 GB DDR3-2133 Corsair Vengence (black/red), EVGA GTX 760 @ 1200 MHz, on an Asus Maximus VI Hero MB (black/red), in a blacked out Antec P280 Tower, with a Xigmatek Night Hawk (black) HSF, Seasonic 760w Platinum (black case, sleeves, wires), 4 SilenX 120mm Case fans with silicon fan gaskets and silicon mounts (all black), a 512GB Samsung SSD (black), and a 2TB Black Western Digital HD (silver/black).
Location: Arizona
Contact:

Re: Missing preferred deadline by 2 hours...

Post by 7im »

Please read the FAQ pages as suggested earlier. Many of your questions are covered in depth. This next link from the FAQ shows how both of the CPU client deadlines are calculated. A similar but much shorter length is applied to the SMP client.

How do you set the deadlines for the work units?

You can find the deadline information for each work unit on the Project Summary page, linked in the toolbar at the top of this forum page.

You can find the deadline for your specific WU in the unitinfo.txt file in each client. The clock starts ticking when you download the WU.

If you do not upload the WU before the prefered deadline, the WU is then put back in the pool, presumably at the top of the list, and assigned to someone else. How quickly that happens is not published, and doesn't really matter. The WU goes in the list for reassignment. A missed deadline is a missed deadline. The WU is assumed to be lost and sent to someone else for completion.

Because of the serial nature of work units, the 1st generation of a work unit has to be returned before the 2nd generation can be sent out, hence the deadlines to keep the ball rolling when WUs are lost. CPU clients have long deadlines. The SMP client deadlines are very short to promote the very quick return of work units.

If your client uploads a work unit just beyond the prefered deadline, Stanford DOES accept that WU, and gives you credit. And the next generation of work can now be sent out. However, the slow return of that WU has already caused the servers to assign that same WU to someone else, and you have unnecessarily caused that 2nd person to process the same WU instead of processing a new WU. That hurts the project, but because there was still some value to the contribution, you still get credit. Once the final deadline has passed, no credit is given as there is no longer any scientific value.

And the number of "other" available work units has little to do with this process, even when there are almost always more WUs to process. WUs are not assigned to just keep people busy folding work units. As mentioned above, the work is serial in nature, and to move the science forward in the quickest way possible, the WUs are sent out in a specific priority.

That is the way it is. It has worked for a long time, with small adjustments along the way, but this policy it is not up for debate. It has been debated to death in the forums for too many years, and the conclusions are always the same. If there was a better way to do it, when considering ALL of the criteria involved, then the smart people at Stanford, and the smart people in this forum would have already thought of that improvement, and would already be using that improvement if not already working on it.*

*Vijay did announce that a points system adjustment was coming with the v6 client and the use of Passkeys. Other than the FAQ entry on Passkeys and a Blog post, no details large or small have been released yet.
How to provide enough information to get helpful support
Tell me and I forget. Teach me and I remember. Involve me and I learn.
Post Reply