Seti@Home optimized science apps and information
Optimized Seti@Home apps => Windows => Topic started by: Geek@Play on 12 Sep 2010, 11:47:01 am
-
With all the problems at SETI I am contemplating dropping that project totally and invest all my efforts into a differnt project. I want my new project to be looking to the sky and the universe.
I am looking at Einstein and Milkyway. Didn't Einstein recently announce a newly discovered pulsar? Please I would like some advise from anyone who knows about these two projects. Project uptime, work availability and so forth. Anything you can add to help me decide between the two.
My intent is to totaly end my work at SETI. Too many continueing problems there and I am afraid that major damage now will make use of the new server to replace failed server and not improve the situation.
Please anyone with advise on these two projects?
-
I can't advise on your proposed choices, but I am sorry to hear that you feel you must make the decision to leave Seti altogether.
We don't know yet if there was any serious damage to the existing servers there, or if such damage is repairable if it did occur. Some of the existing servers are already slated to be taken offline.
I understand as much as anybody the frustration one feels when the project of your choice is down or not running smoothly. That's why I did the donation challenge to fund the new server. Seti is where my heart is. And I truly do feel that they are making progress, not only in terms of getting more of the science done, but in making their systems work more smoothly together. It's not their fault that cantankerous AC system took a powder again. That is part of the infrastructure that the Berkeley campus is responsible for.
Might you consider holding off on making the change until the current storm has passed and the new server comes online? If not, I shall have to respect your decision. It is, after all, a personal choice.
But I really do feel that better times are on the horizon for the Seti project.
-
Just minutes before SETI went down I posted a rant about ghost work units. One computer requested work, got 2 wu and 44 ghost work units. It really pissed me off.
I am currently crunching SETI on 3 remaining computers but have no idea if and when SETI comes back up. Of course nobody does. I guess I am just pissed off in general about the situation at SETI.
-
I read that.........
And posted myself about ghosts and the huge amount of WU storage they must be wasting just waiting to time out and have another shot at being processed.
I have floated the hope that maybe when things get sorted they could restart the resend feature to start ghost killing, but I have not gotten any indication if that is a realistic possibility or not yet. I agree it is a huge problem, and one that need to be addressed sooner rather than later.
-
The only thing that COULD solve this problem is if they talked to BOINC developers and tried to incorporate a daily "check routine" in which the servers request a list of work you have currently in your working queue and match that with that list in their servers that says "you got these WU's to process" ..
If a inconsistency is found then the servers re-issue those WU's that you don't have in your progress queue to others directly within a day..
That is a simple modification that wouldn't put the strains to high on the servers, they ofcourse need to find a simple way of reducing the amount of data transfered to and from the servers and this implemantation is not only meant for seti@home , this is meant for the whole BOINC infrastucture in the way servers communicate with the clients itself..
2 cents..
Regards Vyper
P.S Geek.. Don't go .. Stay! It's not greener on the other side.. Other projects have their issues too.. D.S
-
Does anyone know the extent of the damage that occurred when the AC went off? Is it data or hardware?
Todd
-
Does anyone know the extent of the damage that occurred when the AC went off? Is it data or hardware?
Todd
I know of no reports yet......
But being a weekend, I dunno if they will be getting the AC repairs done before tomorrow. And that has to be fixed before they start trying to restart things and assess what may have occurred. Unless they can boot 1 unit at a time for evaluation without overheating things again.
Or they may be taking their Sunday off and waiting until tomorrow when they will be there all day to work things out.
-
If these AC's cool other computers needed on campus then I would bet they are being worked on. If only seti then willprobably wait till Monday.
-
There is allready a mechanism to take care of the ghost work, if only it was enabled.
When a client computer makes a request to the servers that request contains a list of all the work the computer has on board. The server then would compare that list to the list on the servers for that client computer. Any differences would be sent out again to the client computer.
But again, this is disabled because the surge of ghost work units being sent out overwhelms the system. I believe if they enabled it and just weathered the storm out then it would come down to a more normal level.
Of course the current policy with the 3 day science fair exasperates the problem greatly.
-
If these AC's cool other computers needed on campus then I would bet they are being worked on. If only seti then willprobably wait till Monday.
If I remember correctly unfortunately, Their servers are stuck in a small storage closet. Not much bigger than a broom closet which is why they have so much trouble with the heat. Doesn't take long for things to go south when the air goes out in there. :-(
-
Please anyone with advise on these two projects?
First of all, I'll try not to upset msattler, cause when he gets stirred up, people start throwing away money (trying not to get influenced here :) ).
Basically I would recommend milkyway. I prefer them over einstein because they have various optimized apps, while einstein has one gpu app, which doesn't load that much gpu. The bad for milkyway is they need double precision, but if you look here, http://en.wikipedia.org/wiki/Comparison_of_AMD_graphics_processing_units#Northern_Islands_.28HD_6xxx.29_series all 6xxx will have DP, so it's future proof.
If you like to tinker with app_info, you could go all in.
You could use milkyway as 1st back-up option for seti, with cpu& 0.95 gpu and einstein as 2nd one with cpu & 0.05gpu.
Or, as Raistmer is doing, put also collatz in.
My recommendation is go for projects that support gpus, like seti :) .
-
Even if the project had all new hardware, it wouldn't help much with A/C issues.
Issues that are beyond the control of the project (mostly) and will likely always exist in their current facility.
Beyond that, there are still other underlying bandwidth issues.
Not sure if anything actually can be done to help there either, again because of facilities/location/etc.
I'd mention excess capacity, by us users, as a problem, but that's not going to go over well. ;)
Sadly, taking a serious realistic view of things, it's not easy to have an optimistic view that the project will ever run as smoothly as one may like.
-
Please anyone with advise on these two projects?
First of all, I'll try not to upset msattler, cause when he gets stirred up, people start throwing away money (trying not to get influenced here :) ).
What??? The kittyman get stirred up???
LOL. I guess I do get a bit passionate at times.
-
In the last 10 years plus every spare micro-second I have has gone to SETI and it will continue. Just does not feel right to stop now. If the grass is not greener on the other side as Vyper said, why go.
But I do wish they would do something about all the ghost work units. Then again perhaps I should stop worrying about them.
-
In the last 10 years plus every spare micro-second I have has gone to SETI and it will continue. Just does not feel right to stop now. If the grass is not greener on the other side as Vyper said, why go.
But I do wish they would do something about all the ghost work units. Then again perhaps I should stop worrying about them.
Glad to hear that............hopefully some ghost busting can be done in the not too distant future.
-
There is absolutely no reason to run *just* SETI... maybe you set its priority ridiculously high so that its all that gets any useful work, but during outages (scheduled or unplanned), SETI can and will go down for some time, and some other project would be quite grateful to benefit from those spare cycles that would otherwise be idle.
This is the entire reason SETI went to the BOINC platform in the first place, so PLEASE join at least one other project!
-
My thoughts are that once the new server is up and running, and the other servers have been switched around, the need for the 3 day outage might dwindle, which means that the servers wouldn't be getting hammered so badly when the outage is over. The ghosts realy started being a problem when the servers were stressed to max for several days without a break. If up time was longer, I would gladly cut back on the size of my cache, and I'm sure others would too. Now, with the planned, and unplanned outages, a large cache is needed, which puts greater strain on the servers. I do think this situation will work itself out. I have crunched for Einstein, and although they make minimal use of my 480's, whene ever I have needed a backup project, Einstein loads me with work very quickly. It was exciting that they found that binary system. To me, if I am doing science, then I'm happy.
Steve
-
I doubt that the outages will dwindle since they are still have 10 years of work to process through the NTPKr. This may take a great deal of time to complete and they didnt't leave us an estimate for it
-
Whatever you decide, I hope you wont leave SETI. If not for science, stay for friendship :)