an hour or so ago the smaller memory units finally came pouring in.

Good! I can't get any on my low mem machines though - the last one said there was 266MB available and the WU required another 533MB! That WU is marked as an 800MB task then...

[EDIT] ah - just got some :D

---

an hour or so ago the smaller memory units finally came pouring in.

Good! I can't get any on my low mem machines though - the last one said there was 266MB available and the WU required another 533MB! That WU is marked as an 800MB task then...

[EDIT] ah - just got some :D

11/30/2007 11:02:42 PM|rosetta@home|Message from server: (there was work but your computer doesn't have enough memory)


Now this computer runs ABC@home. Maybe someday when Rosetta gets its act together we can try again...

here is a news update on progress - thanks for the light hearted exchange from everyone.
first I've unhidden my computers on Rosetta the host ID of the old pc is #678950 for those that like to poke around.

the first wu finished overnight but the comms keeps backing off 24hours so I had to nudge it this morning to get another task. this has finished now so I've successfully completed two 5.85 wus; the time was promising but they didn't earn many credits - is that a sign that they didn't really get very far?

now the server won't give me any more work with the same 800MB memory demand and each time it backs off 24hours unless I do a manual retry of comms. it's failed 5 times so I'll wait a while and see how other peeps are getting on.

it's a bit rich if even 512MB machines with xp won't run - they are hardly toy boxes!
_

I think I'll try malaria next (sans the optimizer application) and then WCG (minus the africaclimate project) which has clear guidance on minimum system needed.

---

Regarding the credits... as you would expect a slower machine gets less credit per hour of CPU time then a faster machine. Your processing probably completed just fine, with valid results, but you didn't complete as many models in the period of time that a faster machine would, so you got less credit. Same credit per model as a faster box, but you completed less models in the period of time.

---

Rosetta Moderator: Mod.Sense

Regarding the credits... as you would expect a slower machine gets less credit per hour of CPU time then a faster machine. Your processing probably completed just fine, with valid results, but you didn't complete as many models in the period of time that a faster machine would, so you got less credit. Same credit per model as a faster box, but you completed less models in the period of time.

thanks: yes, of course; I typically get 15-20cr/wu (apart from the occasional bigger one) but these first two on the old Athlon only got ~4.8cr each. but it's only a statistical sample of two so we'll have to be patient to see how things pan out.
what would help is some way of persuading boinc mgr. not to defer communications for 24hours but a more reasonable hour or half-hour.

I clicked update accidentally on the project and to my surprise a third wu downloaded ok; it's now having to share 50:50 with malariacontrol so it'll be a while before it reports. (the latter project seems to have got going ok on win98 too, though one of the apps. doesn't show progress and cpu time.) /pg

During project outages, BOINC tries repeatedly to contact the project and fails. Each time it increases the wait until it tries again (exponential back-off they call it). This helps avoid having all 100,000 machines demanding work in the first hour the server is back online. There's no perfect answer for how long a time is reasonable. That differs for everyone. But BOINC decided on 24hrs. Now that things are back to normal, the back-off time goes back to normal.

You can always click the "retry communications" to force it to try again, when you know the project is back up and has dug out from it's backlog.

---

Rosetta Moderator: Mod.Sense

Why the sudden need for machines to have 800meg?

Once my machines (All have 512 ram) run out, they'll have to crunch something else.

---

Why the sudden need for machines to have 800meg?

Once my machines (All have 512 ram) run out, they'll have to crunch something else.

So... should i disconnect 512 ram machines on all accounts from rosetta? Never more work?

Btw... if i have 4 core cpu, i need 2,6GB of RAM ONLY for rosetta apps???

ARE YOU CRAZY OR WHAT?

---

So... should i disconnect 512 ram machines on all accounts from rosetta? Never more work?

Btw... if i have 4 core cpu, i need 2,6GB of RAM ONLY for rosetta apps???

ARE YOU CRAZY OR WHAT?

Well, I grabbed some 96-MB-wu's a couple hours ago, so there are other work also, and would expect this to still be the case until something else has been announced.

As for the 800 MB-wu's, they seems to be using 800 MB virtual memory, but seemed to only use 200 MB "real" memory. It's possible they can use more memory later in the wu, but didn't check.

Since the Scheduling-server doesn't take into account #cores then it comes to memory, as long as you meets the 800 MB-requirement, there's no problem to download multiple 800-MB-wu's. Also, if they're only using 200 MB "real" memory, they'll only use 800 MB on a quad-core, and not 3,2 GB. You'll need a large page-file, to handle the virtual-memory-usage.

---

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

Size of pagefile is not matter. I set value 2048MB for pagefile, and the rosetta still not work.

---

what's a pagefile? :-)

I've run the benchmarks again in case the new host isn't properly calibrated.

otherwise it's fetched down unaided a couple of smaller wus without the memory warning and one more with the warning. / pete

I think the WU with MolecularRep in the name are the biggist problem for hosts with a 'low' amount of memory. I noticed these taking up 800+MB actual and 1+GB virtual memory on my box. I have leave in memory when suspended turned on.

I have to correct myself, I got some numbers mixed up. The situation is not as bad as I said. the 800+ MB (~860MB) is the virual memory size (private bytes), memory usage (the working set part of it) is 200+ MB (~220MB).

my molecularrep WU is using 190,000 kb of real memory

the page file is using 1,4 gig, this is with running msn and wmp

I've been getting these messages for weeks:

Mon Mar 9 00:16:07 2009|rosetta@home|Sending scheduler request: Requested by user. Requesting 144244 seconds of work, reporting 0 completed tasks
Mon Mar 9 00:16:12 2009|rosetta@home|Scheduler request succeeded: got 0 new tasks
Mon Mar 9 00:16:12 2009|rosetta@home|Message from server: No work sent
Mon Mar 9 00:16:12 2009|rosetta@home|Message from server: (there was work for other platforms)

---

I've been getting these messages for weeks:

Mon Mar 9 00:16:07 2009|rosetta@home|Sending scheduler request: Requested by user. Requesting 144244 seconds of work, reporting 0 completed tasks
Mon Mar 9 00:16:12 2009|rosetta@home|Scheduler request succeeded: got 0 new tasks
Mon Mar 9 00:16:12 2009|rosetta@home|Message from server: No work sent
Mon Mar 9 00:16:12 2009|rosetta@home|Message from server: (there was work for other platforms)

This is on your PPC isn't it? They only give work intermittently for those machines. See here https://boinc.bakerlab.org/rosetta/forum_thread.php?id=4764

---