I have three boxes running BOINC. All are dedicated to SETI.

One is a Toshiba laptop (1500MHz), one is a scratch-built desktop (AMD 2800+), and the third is an old 400MHz Compaq.

For a few weeks prior to Thumper's demise I notcied the number of work units in the desktop's queue was getting progressivly smaller and smaller until there were only two or three at a time. Whereas the two laptops (ever so much slower than the desktop box) would have 20 or 30 work units waiting in queue...

'Course when Thumper crashed, the desktop promptly ran out of work while both laptops still have a few units remaining even today...

All three are using the "default" settings. I tried restarting BOINC and I tried restarting the computer.

It seems to me this started right after upgrading to version 5.8.15 so I removed it and reinstalled my previous version (5.4.11) to no avail...

When trying to force an update I'd get the ever famous "Not requesting new work.."

Any ideas?

---

IBA# 11465
http://imagesdesavions.com

Many WUs have different crunch times due to different Angle Ranges.

Longer workunits will likewise fill a cache faster than shorter workunits that require more to fill the same cache.

Most slower machines will be delegated shorter workunits since they will take longer to crunch on those machines.

---

Most slower machines will be delegated shorter workunits since they will take longer to crunch on those machines.

When did that start here? I know Einstein@home was skewing the chances of sending short S5R1 tasks to slower systems, but I hadn’t heard of anything like that being done in this project. In E@h’s case, there were two clearly distinct sizes of S5R1 WU, while here there’s more of a continuum: where is the boundary between ‘shorter’ and ‘longer’?

---

Most slower machines will be delegated shorter workunits since they will take longer to crunch on those machines.

When did that start here? I know Einstein@home was skewing the chances of sending short S5R1 tasks to slower systems, but I hadn’t heard of anything like that being done in this project. In E@h’s case, there were two clearly distinct sizes of S5R1 WU, while here there’s more of a continuum: where is the boundary between ‘shorter’ and ‘longer’?

I noticed that most of my slower machines, or my laptop that's not on very often, are often given "smaller" workunits to start off with - or machines that I re-attach or attach for the first time seem to get "smaller" workunits, and my theory is that these workunts are delegated to test the waters of the new or newly re-attached machine. I noticed that if the machine finishes these smaller units faster, they will be sent larger units while slower machines will keep getting small units.

Unless you think this is another issue of an imperfect caching algorithm, but I'm not so certain I buy into that theory (not that it couldn't be imperfect).

---

I have three boxes running BOINC. All are dedicated to SETI.

One is a Toshiba laptop (1500MHz), one is a scratch-built desktop (AMD 2800+), and the third is an old 400MHz Compaq.

For a few weeks prior to Thumper's demise I notcied the number of work units in the desktop's queue was getting progressivly smaller and smaller until there were only two or three at a time. Whereas the two laptops (ever so much slower than the desktop box) would have 20 or 30 work units waiting in queue...

'Course when Thumper crashed, the desktop promptly ran out of work while both laptops still have a few units remaining even today...

All three are using the "default" settings. I tried restarting BOINC and I tried restarting the computer.

It seems to me this started right after upgrading to version 5.8.15 so I removed it and reinstalled my previous version (5.4.11) to no avail...

When trying to force an update I'd get the ever famous "Not requesting new work.."

Any ideas?

What is the connect ever X days set to? What version of BOINC? What other projects are you attached to? What are the long term debts of all projects? What are the deadlines and remaining run times of all tasks for all projects?

---

BOINC WIKI

What is the connect ever X days set to?

All three use the same default preferences which I have set for connect every 10 and cache 10 days' worth.

What version of BOINC?

It seems to me this started right after upgrading to version 5.8.15 which I installed on 28 March. A couple days later I noticed BOINC "reporting 3 results, not requesting any more work". Hardware all the same, background programs all the same. Benchmarks still blow the other two boxes away.

What other projects are you attached to?

I have three boxes running BOINC. All are dedicated to SETI. No other projects.

What are the long term debts of all projects?

You got me by the short hairs... No idea what that means.

What are the deadlines and remaining run times of all tasks for all projects?

On the machine with the problems? Zero. No tasks remaining.

---

IBA# 11465
http://imagesdesavions.com

Most slower machines will be delegated shorter workunits since they will take longer to crunch on those machines.

Hmmm... The number of work units wouldn't bother me so much except that I somehow knew I'd end up with 'nothing' to do if there was a server problem...

I like having ten days worth of work in queue. After I installed 5.8.15 the projected time for the completion of what I was carrying {Before Thumper Crash} was hardly over 20 hours.

I tried making special preferences for just this machine to force it to look for more work but that did no good. I tried to reset the project to force more work downloaded but that did no good. I even rolled back to 5.4.11 thinking the new version was messed up but that did no good either...

All it would do is report results and hardly ever request new work, even with only 20 hours worth of work in the queue...

---

IBA# 11465
http://imagesdesavions.com

I would suggest reading Work Buffer in the Boinc Wiki for a greater understanding of how the Connect to network about every 'x' days effects the amount of work cached. Pay particular attention to the Recommendation for Single Project Systems section and the table it contains. Seti has WU with deadlines as short as 4 days. It may be counter intuitive, but a setting of 10 days on the Connect to network will actually reduce the amount of work you receive because of the algorithm used to ensure that work doesn't miss deadlines.

Also, note that the setting for Cache enough work for an additional 'x' days is not yet implemented and will have no effect until Boinc Manager V5.10.x used.

edit: The suggested method to not end up with an idle computer during an extended outage is to have a backup project with a very low resource share.

---

Boinc V7.2.42
Win7 i5 3.33G 4GB, GTX470

What is the connect ever X days set to?

All three use the same default preferences which I have set for connect every 10 and cache 10 days' worth.

And this is the problem. S@H has some tasks with 4.5 day deadlines. If you get one of these, no more work from S@H will be downloaded until it is completed. The basic problem is trying to cache 10 days of work with 4.5 day deadlines. S@H needs daily connections because of te fairly short deadlines, and a connect every X no larger than a day and a half or so.

Use a secondary project to get through outages (I have not noticed the outage at al).

---

BOINC WIKI

And this is the problem. S@H has some tasks with 4.5 day deadlines. If you get one of these, no more work from S@H will be downloaded until it is completed. The basic problem is trying to cache 10 days of work with 4.5 day deadlines. S@H needs daily connections because of te fairly short deadlines, and a connect every X no larger than a day and a half or so.

This actually makes sense.

---

S@H needs daily connections because of te fairly short deadlines, and a connect every X no larger than a day and a half or so.

Cool... I'll give that a shot. Thanks.

Use a secondary project to get through outages (I have not noticed the outage at al).

There's more than one SETI project?

-Stephen!

---

IBA# 11465
http://imagesdesavions.com

S@H needs daily connections because of te fairly short deadlines, and a connect every X no larger than a day and a half or so.

Cool... I'll give that a shot. Thanks.

Use a secondary project to get through outages (I have not noticed the outage at al).

There's more than one SETI project?

-Stephen!

No, but there is more than one BOINC project. There are lists here, here, and here.

---

BOINC WIKI

There's more than one SETI project?

No (unless you could the Beta Test project, which runs from the same servers as this one), but there are several other scientific endeavours making use of the BOINC infrastructure. A selection can be found on the Choosing Projects page; to join one, choose “Attach to project” from the Tools menu in BOINC Manager, and paste the URL shown beside its name into the first window in the wizard that comes up. Your accounts in each project are separate, but if you use the same e-mail address to sign up, your preferences and statistics will be synchronized.

---

And this is the problem. S@H has some tasks with 4.5 day deadlines. If you get one of these, no more work from S@H will be downloaded until it is completed. The basic problem is trying to cache 10 days of work with 4.5 day deadlines. S@H needs daily connections because of te fairly short deadlines, and a connect every X no larger than a day and a half or so.

Right now I have one work unit (not including all the ones trying to upload). It has less than ten hours to completion and a report deadline of 11 June. BOINC is still telling me, "not requesting new work"...

?

---

IBA# 11465
http://imagesdesavions.com

And this is the problem. S@H has some tasks with 4.5 day deadlines. If you get one of these, no more work from S@H will be downloaded until it is completed. The basic problem is trying to cache 10 days of work with 4.5 day deadlines. S@H needs daily connections because of te fairly short deadlines, and a connect every X no larger than a day and a half or so.

Right now I have one work unit (not including all the ones trying to upload). It has less than ten hours to completion and a report deadline of 11 June. BOINC is still telling me, "not requesting new work"...

?

What is your connect every X? How many projects are attached? What are the long_term debts (from the client_state.xml) associated with their projects? how much work is left on each task? What are the`deadlines? What are the resource shares for each project? What are the time_stats (from yhe client_state.xml)?

---

BOINC WIKI

What is your connect every X?

I've tried everything from 0.01 to 10 with no change.

How many projects are attached?

One

What are the long_term debts (from the client_state.xml) associated with their projects?

<long_term_debt>0.000000</long_term_debt>

how much work is left on each task?

One remaining task with 00:53:25 remaining

What are the`deadlines?

01/11/2007 09:31:51

What are the resource shares for each project?

Only one project, 100%

What are the time_stats (from yhe client_state.xml)?

[/quote]

<time_stats>
<on_frac>0.999996</on_frac>
<connected_frac>0.999999</connected_frac>
<active_frac>0.000426</active_frac>
<cpu_efficiency>0.951876</cpu_efficiency>
<last_update>1194166889.048828</last_update>
</time_stats>

I am certain we will find it's not a setting or a usage problem. This very same machine, in the very same configuration, with the very same settings used to carry several days worth of work all the time. It stopped doing that and hasn't carried more than twenty hours or so since I updated to 5.8.15.

---

IBA# 11465
http://imagesdesavions.com

<active_frac>0.000426</active_frac>
You computer does not appear to spend enough time crunching to warrant dl an other WU until it is almost finished with the current one.

What are the`deadlines?

01/11/2007 09:31:51

I'm concerned about the deadline. Didn't January 11th pass already?

[Edit] I just realized that could be November first too, depending on which method is used for dates.

---

This is either a settings problem, or a run time issue. I am betting that the RAM usage or CPU usage settings are wrong. How much RAM do you have? What are the RAM usage settings for active and inactive?

Is the deadline January or November?

---

BOINC WIKI

Is the deadline January or November?

How does BOINC calculate active_frac? Could it get messed up by a system-time glitch? But it would seem strange that the other time stats weren’t affected: they seem pretty normal.

If Stephen! is using the screensaver and has work-while-in-use set to “no”, could it be one of those cases where BOINC decides that the screensaver kicking in constitutes user activity?