I added a server in.... dual E5 2440. Slow as a dog compared to my dual E5 2643 V3 workstation. 6 hours per task versus 30 minutes. No AVX2 on the older E5 2440. Tried the AVX optimizer and the time went from 6 hours per to 7.5. o_O So took that back off as it was actually going to be slower! Going to try it again once it finishes the current group of tasks just to be sure. Seem that if you start it with that and then take it back out it aborts all current tasks too.
 
I added a server in.... dual E5 2440. Slow as a dog compared to my dual E5 2643 V3 workstation. 6 hours per task versus 30 minutes. No AVX2 on the older E5 2440. Tried the AVX optimizer and the time went from 6 hours per to 7.5. o_O So took that back off as it was actually going to be slower! Going to try it again once it finishes the current group of tasks just to be sure. Seem that if you start it with that and then take it back out it aborts all current tasks too.

Try the sse2 one. I'm sure it must have that. On my older systems without avx I ran the sse2 and it took them from 6ish hours to 30 minutes.
 
I added a server in.... dual E5 2440. Slow as a dog compared to my dual E5 2643 V3 workstation. 6 hours per task versus 30 minutes. No AVX2 on the older E5 2440. Tried the AVX optimizer and the time went from 6 hours per to 7.5. o_O So took that back off as it was actually going to be slower! Going to try it again once it finishes the current group of tasks just to be sure. Seem that if you start it with that and then take it back out it aborts all current tasks too.

E5 2440 have AVX :)

EDIT:
never mind, going for the sse2 version is probably the best option for you.
 
Last edited:
Just gave that a go. Maybe it needs to settle down. The tasks had 2.5 hours remaining. Shut it down, copied sse2 over, started it back up. Shows as optimized with the Opti 1.0 on each task now and the time remaining is now right at 4 hours each. :(
 
Just gave that a go. Maybe it needs to settle down. The tasks had 2.5 hours remaining. Shut it down, copied sse2 over, started it back up. Shows as optimized with the Opti 1.0 on each task now and the time remaining is now right at 4 hours each. :(
They won't take that long. The Time Left is inaccurate when using the optimized apps.
 
Just gave that a go. Maybe it needs to settle down. The tasks had 2.5 hours remaining. Shut it down, copied sse2 over, started it back up. Shows as optimized with the Opti 1.0 on each task now and the time remaining is now right at 4 hours each. :(

As RFGuy said, mine show 4 hours remaining and take between 20 and 35 minutes.
 
Yeah, BOINC has to "learn" based on work unit history and initial estimates. So, it may take multiple work units completing to make it accurate.
 
Yup - server is also kicking them through quick now too. 685 pending validation...

When does this sprint end?
 
A long one! Will be interesting to see how it pans out in the end. Ima have to fire a couple of more clients up.
 
One of my systems just will not upload it's 250+ tasks.

Keeps saying it cannot contact the server, all the while the others are completing tasks and uploading them lol
 
Just added in an older server that had a light load (been meaning to find time to move the last functions off it it to another server and decommission it - its pretty much been running just as a print server and hosting a single small database for the time clocks for the past year). Dual E5450 (3ghz). That thing is going to start generating some extra heat in the server room... lol
 
auntjemima are they pending uploading or ready to report?

If ready to report, then click the "report tasks" button in one of the drop down menus. I forget which one now.
 
Threadripper #1 was built today and is already crunching. #2 should arrive tomorrow and I will have it online by the weekend. Shooting for the start of the sprint, but work and family commitments may intervene.

Spent the evening building Threadripper #2 and I now have it up and running and crunching Rake tasks for the sprint. That gives me 100 threads on Rake for the weekend.

Edit: Make that 120 threads. I moved all of my boxes over to Rake now.
 
Last edited:
auntjemima are they pending uploading or ready to report?

If ready to report, then click the "report tasks" button in one of the drop down menus. I forget which one now.

I fixed the issue. I had changed my upload speed to .01kb on that one box when I was trying to get more tasks. Noticed and the problem is solved.
 
Yeah yikes. These wingman projects don't really allow to make up ground either I feel. For every task of a teammate we validate, we validate 3 for an enemy.
 
For those not running optimized apps aka anonymous platform, there is a site in which you can download the apps: https://github.com/sirzooro/RakeSearch/releases/tag/v1.0
I notice there are some new crunchers who have just joined and running the default apps which typically give or take 5 times longer than the optimized apps. For more reading, go to here: http://rake.boincfast.ru/rakesearch/forum_thread.php?id=39#503

1. In the github link, look for instruction set for your CPU e.g. AVX2, AVX, SSE2, etc
2. Download the required zip/tgz file to your local drive.
3. Save it and extract the files. There are two files: app_info.xml and rakesearch application program, typically rakesearch.exe.
4. Abort or finish up all tasks. Exit BOINC.
5. Delete all apps in the rakesearch BOINC project folder. In my Win7, it is located here: C:\ProgramData\BOINC\projects\rake.boincfast.ru_rakesearch
6. Copy the extracted files to where the rakesearch project BOINC project folder is located.
7. Run BOINC mgr. You should see something like this. The application name will be called "Local: RakeSearch (Opti v1.0) 7.02"

upload_2018-7-27_20-34-13.png
 
I'm finally back. Looks like you broke the site - I can't create an account. Not sure I could do much good anyway with the state of the validator, however I will keep trying.
 
From rakesearch admin: http://rake.boincfast.ru/rakesearch/forum_thread.php?id=101&postid=508#508.

We are bombarding the servers with more results than the server can handle. Read the last sentence. Is there a hope?

Now, in the project, more than 1 canonical result generates per second. Concequently, more than 2 results from computers are processed per second. About 3 or 4 hours ago, comparable number of tasks per second was requested from computers. We monitor the logs of project server processes (during a race it is a very interesting exercise!). Validators works normally, but the number of results from computers is really high - about by factor 10 or even 15 more than in usual day of project life. If everyday processing rate will be like this, computing of whole workunits will be completed in 7 or 8 months. By approaching to sprint finish, rate of tasks request (as I think) will go down by several times, I/O load to server will go down also, and validators process all tasks from queue.

Watching!
 
Rakesearch validator update: http://rake.boincfast.ru/rakesearch/forum_thread.php?id=101#513

Not quite clear what he says in the second paragraph. A ray of hope for those with large pending tasks? Let's see what happen tomorrow. Time to go to bed.

xii5ku, thank you for race report! ChristianVirtual, kashi - thank you for iformation about large lag! We know, that many workunits need a validation, but don't know that many of them received a two days ago, because many new results pass validation immediately.

We performed an additional investigation of situation and detect a some unexpected behavior of server processes (may be absolutely correct, but really unexpected because during usual loads we don't saw this) - time of next validation of many (but not all!) results and workunits shifted to 2 August! We set this time to 2018-07-29 00:00:00 and validators already returned to process them. Please check the situation now.

We see a several places for optimization and speedup of server works, but each of them need a server shutdown and this is not a good idea for last 16 hours of race. Formula BOINC sprints and marathons is a really cool things!
 
No. No hope. Hes not gonna shut down the server during the sprint.

But next time it might be better if they can optimize the back end.
 
Back
Top