Why is it so hard to get work?

I've gotten quite a bit of work lately....not a ton but a healthy amnt...

I managed to DL a 12 unit set this weekend. I almost threw a party in honor of the occasion.

Do that every day to get the 5,000 jackpot!

What is really weird is that I have one machine that seems to get work all the time and the rest just get a few units now and again.

I have one machine that has never run out, but that takes 40min to run a WU using Cluster's optimised [CPU] client.
My radeons [on other machines] spend 99% of their time idle.

---

Cheers,

PeterV

.

Why is there throttling at all?
Why is there a shortage of workunit generation? Is there a limit of some kind that prevents new workunits from being created? (such as a lack of server/CPU/memory power-capacity to create said WUs?)

Sorry, I'm just getting a little frustrated seeing that "Milkyway@home: Message from server: No work available" message I keep seeing in my BOINC manager, that's all.

Why is there throttling at all?

There is no throttling. Just ways of accelerating (optimized apps).

Why is there a shortage of workunit generation? Is there a limit of some kind that prevents new workunits from being created? (such as a lack of server/CPU/memory power-capacity to create said WUs?)

Apparently, the number of WUs is not the problem. The problem, I hear, is the bottleneck at the distribution point made by the hammering of gimme, gimme, gimme requests. So they say.

---

I suspect there may be a bit of egg/chicken here. That is, with the release of the optimized .19 application and then the optimized ATI GPU application, the demand for work units skyrocketed. Back in the fall, the optimized CPU application by itself pushed the project hard. Then with the ATI application being deployed, the demands in terms of workunits went over the top. The first constraint appears to be at the distribution point so you see on the server status page 500 to 900 workunits available at any given time. The second and I believe core constraint is at the workunit creation process -- if that wasn't the case you would see an increasing number of workunits as being available on the server status page -- you don't -- it remains at 500 to 900.

In addition the vastly increased work unit productivity of both the CPU and ATI GPU optimized clients, there has also been a significant increase in the number of users here. There were under 6000 users back in September, there are more than 18000 users today. Thus, the release of the optimized clients, plus the tripling of users within about 8 months means there is a HUGE increase in number of work units being requested (script or no script).

It wouldn't surprise me to see the number of workunits processed to have increased by 10 times or more in the past 8 months -- certainly the user level capacity for procesing work has increased by that much.

I doubt the project has seen an infusion of hardware resources to match this, and so we are seeing bottlenecks.

In theory, the setting up of a formal MilkyWay GPU project (first with CUDA support and then critically with an Optimized ATI GPU application) will draw away a large portion of load on the MilkyWay (nominally) CPU project here.

Personally, I sort of wonder if the folks back at the project have adequate work creation resources and for that matter science analysis resources to cope with the work going on now, let alone the traffic for two related and separate projects (Milkyway and Milkyway GPU). The delay between the posting on the home page regarding the GPU project, and any subsequent follow up information suggests they are very much resource constrained at the moment -- I suspect that resource constraint will not be going away any time soon.

Apparently, the number of WUs is not the problem. The problem, I hear, is the bottleneck at the distribution point made by the hammering of gimme, gimme, gimme requests. So they say.

---

I suspect that resource constraint will not be going away any time soon.

Final exams should be in 1-3 weeks...

I suspect that resource constraint will not be going away any time soon.

Final exams should be in 1-3 weeks...

Will that be about the same time that we are able to get more workunits?

I suspect that resource constraint will not be going away any time soon.

Final exams should be in 1-3 weeks...

Will that be about the same time that we are able to get more workunits?

My name is not "Claire Voyant"... ;-)

I suspect that resource constraint will not be going away any time soon.

Final exams should be in 1-3 weeks...

Will that be about the same time that we are able to get more workunits?

My name is not "Claire Voyant"... ;-)

LOL

I suspect that resource constraint will not be going away any time soon.

Final exams should be in 1-3 weeks...

Will that be about the same time that we are able to get more workunits?

My name is not "Claire Voyant"... ;-)

Perhaps we'll get somewhere if cross Claire Voyant's palm with 'Silver' ;-)

I got me a 4870 while I was at the computer shop getting my 4850 back.

Now I have 2 starving beasts to feed. No matter how hard I pound the feeder it's not giving me anything! ;)

How are the top users getting so much work?!

I got me a 4870 while I was at the computer shop getting my 4850 back.

Now I have 2 starving beasts to feed. No matter how hard I pound the feeder it's not giving me anything! ;)

How are the top users getting so much work?!

Some probably have script files that request work every x secs/mins, regardless of whether they get any or not.

---

Never surrender and never give up. In the darkest hour there is always hope.

*.*

---

Well yes it must be some kinda program hammering the server for units reporting they don't have any but since these cards finish them in like 12 seconds or so they must get a huge amount.
The only way to stop them is to release a new application which probably will happen when the new gpu app gets released, then it will take some time before the ati cards

I tell you what.
Those MWGPU WU are really slowing my computer to a crawl..
Seem I get stuck with all the VLAR...
Aw well in the name of science.!!!
@ Misfit you said you "purged" thoose from your system and will not get anymore?
Is there a script to avoid getting those??


Martin

Blah!
Ignore the below comment.
I meant S@H Beta NOT MWGPU.
Wrong forum. Wrong Thread!
Too many forum reading = Brain confusion.

Martin

@ Misfit you said you "purged" thoose from your system and will not get anymore?
Is there a script to avoid getting those??

You will get what they send you. Not a script, but a modified .exe that aborts them. Also I've discovered the slow down in my video and increased crunch time was because of workunits getting suspended in the GPU memory while it crunched other units. That happens when BOINC DLs more work with an earlier deadline and switches tasks.
PM me on SETI if you need specific info. I'm only crunching MB CUDA units with my GPU. No AP and no CPU usage so my app_info will be different than the typical optimized setup.

---

me@rescam.org

Let him download the new installer from kwsn ;)

@ Misfit

Thank you so very much for the help.
You have been great and I am extremely thankful!

I propose a toast to you!

Martin Chartrand