So it's better to invest on machines to sieve while post-processing stage is better placed on the clusters. I though that too after a few researches and calculations and I was waiting for your comments.

A few years ago, ~10 years, the US-Distributed team had the simple way to build a farm of computers by using baskets (I'm missing the name of the thing). They managed to connect two motherboards with only one PSU. The baskets were easily put above each other making like a tower. It was a very simple way to build a farm without having to buy a case box for each machine. If I remember they called each basket a node.

Thank you for all explanations. Carlos.

[QUOTE]A few years ago, ~10 years, the US-Distributed team had the simple way to build a farm of computers by using baskets (I'm missing the name of the thing). They managed to connect two motherboards with only one PSU. The baskets were easily put above each other making like a tower. It was a very simple way to build a farm without having to buy a case box for each machine. If I remember they called each basket a node.[/QUOTE]We remember the system you are talking about but we are unable to find a link to it. We remember it being a very clean and neat design.

Here are some others that are not nearly as cool.

[url]http://bhs.broo.k12.wv.us/homepage/staff/seti/farms.htm[/url]

[QUOTE=Xyzzy;311261]We remember the system you are talking about but we are unable to find a link to it. We remember it being a very clean and neat design.

Here are some others that are not nearly as cool.

[URL]http://bhs.broo.k12.wv.us/homepage/staff/seti/farms.htm[/URL][/QUOTE]

They were cool, weren't they? That was my dream...they had such ideas...
The forum disappeared, I don't know where they are now as a DC team?!?

Question:

We have moved a few boxes to sieving. We have not yet figured out exactly how it works, and we have no idea if they are productive, but we wonder if our resources are better utilized sieving or doing post-processing.

We enjoy the post-processing, probably because most of our computers are loaded with RAM, there is a neat countdown timer and it appears that you always get an answer, but what is more beneficial for the project?

[SIZE=1]Plus, you get your name attached to the result![/SIZE]

:max:

[QUOTE=Xyzzy;311267]Question:

We have moved a few boxes to sieving. We have not yet figured out exactly how it works, and we have no idea if they are productive, but we wonder if our resources are better utilized sieving or doing post-processing.

We enjoy the post-processing, probably because most of our computers are loaded with RAM, there is a neat countdown timer and it appears that you always get an answer, but what is more beneficial for the project?

[SIZE=1]Plus, you get your name attached to the result![/SIZE]

:max:[/QUOTE]

If you want your name attached to the result only do Post-processing for NFS@Home or sieve for M929.
For windows clients it's a waste of time and energy to sieve for NFS@Home when you have 64 bit linux sievers for NFS@Home. But is up to you to decide where to put your machines. I would put them to sieve, at least you will climb at the stats.

[QUOTE=Xyzzy;311261]We remember the system you are talking about but we are unable to find a link to it. We remember it being a very clean and neat design.

Here are some others that are not nearly as cool.

[url]http://bhs.broo.k12.wv.us/homepage/staff/seti/farms.htm[/url][/QUOTE]

[url]http://www.pcmech.com/forum/attachments/general-discussion/940d1028217709-prime-monster-stomp_monster.jpg[/url]

[QUOTE=Xyzzy;311267] we have no idea if they are productive[/QUOTE]
Yes, I see seven computers plowing through sieving tasks. Looks great!

If you don't mind, I can assign the rest of the 8647_61_minus1 tasks to your computers to mop them up more quickly. Then you can start the postprocessing once the pending relations hits 0. I'm doing that for the remaining S3m637 tasks now so it should be ready in a day or so.

As far as sieving vs. postprocessing, both are useful. We wouldn't be able to do anything without our sievers, of course, but relieving the burden of postprocessing, especially the formerly RSALS numbers, is extremely helpful.

Thanks!

[QUOTE]If you don't mind, I can assign the rest of the 8647_61_minus1 tasks to your computers to mop them up more quickly. Then you can start the postprocessing once the pending relations hits 0. I'm doing that for the remaining S3m637 tasks now so it should be ready in a day or so.

As far as sieving vs. postprocessing, both are useful. We wouldn't be able to do anything without our sievers, of course, but relieving the burden of postprocessing, especially the formerly RSALS numbers, is extremely helpful.[/QUOTE]Sounds great!

From our notes it appears we have reserved 8647_61_minus1, np_111, F1629 and S3m637 for post processing. We are very willing to do whatever work is the most productive for the project, so if you feel we should do less, or more, post processing then just let us know what schedule you think is best.

BTW, we have a [URL="http://escatter11.fullerton.edu/nfs/show_host_detail.php?hostid=28464"]computer[/URL] that shows in the BOINC manager that a lot of the calculations are invalid. How do we find out how many are bad and why? The box previously (as of last week) turned in LL-D results that were correct. We have a 32-bit OS on it right now because it, for some reason, will not boot an install CD and we have not had time to track down that issue. We assume we are losing a bit of efficiency running the 32-bit OS but it is a slow computer to begin with.

:spot:

We found the error page. It looks pretty bad!

[url]http://escatter11.fullerton.edu/nfs/results.php?hostid=28464&state=5[/url]

:max:

It's actually ok. The computer was too slow using the 32-bit siever to run the 16e tasks. The heartbeats were taking too long and BOINC thought the siever stopped responding. The 15e tasks are running fine, though. It does occasionally seem to have trouble reading the restart file, but don't worry about that. They are rare enough.

I just instructed the server to send all remaining 8647_61_minus1 tasks to you, so you should see them appearing whenever your computers fetch new work. :cool:

We assume you can see all of the details of all of our computers.

Which strategy, from an OS and architecture POV, would you employ to choose boxes to sieve and boxes to post process?

The i7 seems to really fly on sieving.