Hi. I have numerous pcs...many are almost identical in CPU speed and hardware configuration.

I notice that I have at least 1 machine that seems to take forever reporting. For example, all my pcs are set to "Connect about every" .1 hours. Yet I have a machine right now that has not contacted in 8 hours...and this has been happening for the past 5 days (I just installed it 5 days ago).

Yes, the machine is happily busy crunching work. I have it set to run at max 90% CPU time when the machine has been free for more than 1 minute. And this machine is pretty fast...duo core 2 processor at 2.33GHz and with 3gig ram.

Any ideas? This problem happened on my Mac for about a month and then the problem disappeared...no changes by me.

I looked at the Messages Tab and noticed that Boinc seems to think the machine is in use every 30mins...as if someone shakes the mouse or types on the keyboard. These messages are popping up a lot from 2am-6am when nobody is using the machine, we have no pets, there are no earthquakes, etc. And this is a clean machine with XP, Norton Internet Security, and Office...literally that is it. Is there a little known trigger that I don't know about such as a software update that triggers Boinc to think the machine is in use? I always thought it was physical characteristics like using the mouse or keyboard.

The above example is surely stealing some minutes here and there, but shouldn't be stopping Boinc from communicating...I see a backlog of about 6 work units that are completed and each unit takes about 2-3 hours so we're talking 12-18 hours of work that has not been automatically reported.

Thanks for any help!

-Eric


2/22/2008 2:24:59 AM||Suspending computation - user is active
2/22/2008 2:26:00 AM||Resuming computation
2/22/2008 2:44:39 AM||Suspending computation - user is active
2/22/2008 2:45:40 AM||Resuming computation
2/22/2008 3:00:31 AM||Suspending computation - user is active
2/22/2008 3:01:32 AM||Resuming computation
2/22/2008 3:05:28 AM||Suspending computation - user is active
2/22/2008 3:06:30 AM||Resuming computation
2/22/2008 3:09:52 AM||Suspending computation - user is active
2/22/2008 3:11:53 AM||Resuming computation
2/22/2008 3:29:16 AM||Suspending computation - user is active
2/22/2008 3:31:40 AM||Resuming computation
2/22/2008 3:33:39 AM||Resuming computation
2/22/2008 3:34:06 AM||Suspending computation - user is active
2/22/2008 3:35:06 AM||Resuming computation
2/22/2008 3:43:05 AM||Suspending computation - user is active
2/22/2008 3:44:06 AM||Resuming computation
2/22/2008 3:58:57 AM||Suspending computation - user is active

---

Hi, There's two different places to set up your "preferences". You can do so from the website, or you can set preferences locally via the boinc manager "advanced" tab. If you set them "locally", then this "over rides the web based setting. If you go to "advanced", "preferences" and "clear" the local prefs, then it'll reset to the web based settings.

It appears one of them is set to "suspend computation" when the user is "active".

Also, "reporting" is a function of your cache preferences. Perhaps that one machine is set to connect less often than .1 hours (I think it's days, not hours, but you might have performed the math and entered 1/240th of a day)

good luck

tony

These preferences are all local...all in the Advanced>Preferense menu of the thick client.

Right...I see that Boinc is suspending...that's my 2nd point...I'm not there to make it suspend so why is it suspending?

But the suspending issue is the least of my worries...can always set that to run 100%.

-Eric

---

I'm not sure I'm answering your question properly.

"user activity" is determined by mouse or keyboard activity and that activates a timer. so for example, if mouse if moved, boinc sees the user as "active" for X minutes. (I think it's 3 minutes)

I'm not sure I'm answering your question properly.

"user activity" is determined by mouse or keyboard activity and that activates a timer. so for example, if mouse if moved, boinc sees the user as "active" for X minutes. (I think it's 3 minutes)

I have one of those standard Microsoft optical laser mice...do you think it is triggering it?

However, more than 15 hours ago I set Boinc to "run always" so again, 15 hours is more than .5 of a day...and I have it set to contact every .1 days.

-Eric

---

Is a mouse moving the mouse? Sorry, could not resist.

ejuel, it sounds like you are expecting BOINC to contact the project 10 times every day... whether it needs to or not. And, even if there are completed tasks to report, BOINC doesn't not feel they NEED to be reported right away. By combining your reporting and requests for new work and contacting the project less frequently, it is helping the productivity on the project servers.

BOINC uses that number to basically understand how much work it should be requesting, so that it avoids contacting MORE then your preference. So, really, you have defined a preference that says "get enough work so that you will not be contacting the projects more then 10 times per day". And if you are running more then one project, this is one of the balancing acts that BOINC plays.

...so now let's get to the point. Why are you concerned about when tasks are reported? There is a setting somewhere that says "report 'em all immediately". But I don't recall where it is, and doing so make more hits on the project's servers.

As for your machine being "in use" at 2AM when you are sleeping, yes I would guess the optical mouse thought it moved. Or perhaps it actually did if it has a cord and the cord slipped further off the desk or whatever. By indicating BOINC can run all the time, this won't slow your crunching. It is just a question of whether it slows your work on your computer. In most cases, you won't even notice BOINC is running during the day while you are using your computer. But if that becomes an issue, you could schedule it to run specific hours of the day when you are not using the computer.

---

Rosetta Moderator: Mod.Sense

[quote]ejuel, it sounds like you are expecting BOINC to contact the project 10 times every day... whether it needs to or not. And, even if there are completed tasks to report, BOINC doesn't not feel they NEED to be reported right away. By combining your reporting and requests for new work and contacting the project less frequently, it is helping the productivity on the project servers.

BOINC uses that number to basically understand how much work it should be requesting, so that it avoids contacting MORE then your preference. So, really, you have defined a preference that says "get enough work so that you will not be contacting the projects more then 10 times per day". And if you are running more then one project, this is one of the balancing acts that BOINC plays.

...so now let's get to the point. Why are you concerned about when tasks are reported? There is a setting somewhere that says "report 'em all immediately". But I don't recall where it is, and doing so make more hits on the project's servers.
quote]

Ok...well, every time in install Boinc on a new pc, it's local Preferences are set to .1 day...so I don't mess with it. This default has been happening on the past 4-5 installs over the past 2months. Even if it was not set to default, I always thought that the RAC was calculated on how much work is done PER DAY. I saw a REAL LONG post somewhere about how RAC was calculated and that was too confusing...if I am wrong about RAC, can someone sum up how RAC is calculated in say, less than 100 words? :)

As for the mouse thing, yea, I know it doesn't hurt the pc to let Boinc run 24x7 at 80% CPU strength or more...I was just trying to set this 1 pc to only run while someone is away. And for whatever reason, it just isn't working right. I am going to lean towards the mouse triggering the system. Maybe I will just set it to run 24x7 at 90% CPU.


On a total tangent, I have 3 quad-core systems coming online in a few days. :)

-Eric

---

RAC - recent average credit. This is a link to the multi thousands of page Boinc Wiki. Many other answers can be found there using the search box.

In short, RAC is an average of the last several days of work completed. And so whether it gets reported today and makes today's number larger; or tomorrow, the average for the two days is still the same. And RAC goes back a lot further then 2 days.

So, could you isolate just what the difference you are observing is? What behavior you are used to?

---

Rosetta Moderator: Mod.Sense

...so now let's get to the point. Why are you concerned about when tasks are reported? There is a setting somewhere that says "report 'em all immediately". But I don't recall where it is, and doing so make more hits on the project's servers.

I think they took it out of the more recent releases to reduce the load on the servers...

---

Well, the easy one 1st., "Connect about every N days", this is not how often you're going to connect, or how many times per day. It's the longest time you expect your computer will be offline. For permanently connected computers, it's recommended to use zero, or 0.01 days or something, but it's no problem to leave it at the default of 0.1 days (or 0.25 days as atleast WCG is using).

For cache-size, use the "Maintain enough work for an additional N days".

BOINC won't connect to Scheduling-server except if it needs to do it. The reasons are:
1; Asking for more work.
2; If finished result is less than 24 hours from deadline.
3; If finished result atleast 24 hours since uploaded.
4; If less than "connect about every N days" to deadline for finished result.
5; To send trickle-up (CPDN only).
6; User gets impatient and hits "update".

#1 is the most common reason for connection, this normally happens then expected run-time of cached work drops below cache-size, the sum of "connect about every N days" and "Maintain enough work for an additional N days". If multi-project, it can be a long time till next needs to connect. Also, there is instances then work-request is blocked, this can be due to "too large cache-size", stuck download or stuck uploads.

Especially if you runs with a large cache, even a small increase in "duration correction factor", or a decrease in one or more of "on_frac", "active_frac", "cpu_efficiency" can lead to the "5 day cache" suddenly has expected run-time of 6 or 7 days, meaning it can take a couple days before computer thinks it needs to ask for more work...

Due to #3, even if there's no other reason to connect, a result will max wait 24 hours from finished uploaded.



As for the reason of "Suspending computation", it's not in a well-defined pattern to be wrong-reporting due to 90%-throttle atleast.

The mice living it's own life is a possibility, would guess it's fairly common atleast with optical mice...

Hmm, disconnecting the mouse for the night should likely show if this is the reason...

Possibly a client-bug? You can try one of the later alpha-builds like v5.10.43, to see if this fixes the problem...

---

"I make so many mistakes. But then just think of all the mistakes I don't make, although I might."