IRIX and Software

irixpgmr wrote: Don't get me wrong. I like einstein@home. I run it on all my machines except my SGI boxes. It just not viable to port at this time.

Understandable - just wanted to point out that the BOINC/SETI@Home currently in use for IRIX is a gcc compile, *not* MIPSpro. So if you're happy with how it works, well that's a start - compile it with gcc

Anyone else having trouble getting work out of the stupid scheduler? Alternatively, anyone experiencing problems with your work units being credited? I've got about 20 units awaiting to be credited dating as far back as the end of May. If this keeps up its not feasible to run the machines if their going to sit idle. I'll go back to classic until they pull the plug entirely.

Yep, SETI is still having problems with their servers.

http://setiweb.ssl.berkeley.edu/tech_news.php

nekonoko wrote: Yep, SETI is still having problems with their servers.

http://setiweb.ssl.berkeley.edu/tech_news.php

Maybe they should switch to Origin's

Any thoughts on getting an R4k optimized SETIi binary compiled? I've been running the client on my Indigo, and it's taking longer to process work units than I think it should ( though I could be wrong... )

Dr. Dave wrote: Any thoughts on getting an R4k optimized SETIi binary compiled? I've been running the client on my Indigo, and it's taking longer to process work units than I think it should ( though I could be wrong... )

Are you talking Classic or BOINC?

the BOINC version... I did a bit of a preliminary calculation, based on measured integer and FP MIPS:



This was done by taking measured float and integer MIPS, multiplying it by the average number of days ( * 24 * 60 * 60 ) and tabulating the results. VERY unscientific, so I make no claims on it's validity. There's a good chance that the client is probably OK anyways - but it's interesting to note that the Indigo runs (if you compare integer speed) 3.71 times slower than the R12k-340, and takes 3.86 times longer to process the data. This data is also probably skewed by turnaround times anyways - as there's probably much interaction with cached data stores etc.

But just wondering if the client was in fact R4k optimized...

I have not tried running BOINC an any R4x00 hardware so I cannot comment , but I did have it running on an O2 R5000@180MHz and can tell you its not worth the effort. Whether BOINC is optimized for R4x00 I also cannot say, but again, probably would not be worth the effort.

kshuff wrote: Whether BOINC is optimized for R4x00 I also cannot say, but again, probably would not be worth the effort.

I don't think the IRIX BOINC client is optimized for anything

That's right. Nevertheless boinc4.58 works fine on my Octane R12000@400. And don't forget that even R12000@400 Mhz is a little bit old fashioned compared with the power of actual x86 and G5-cpus.
But I am proud to say, that my "mighty machine" is still able to take part in the seti-search. Therefore thanks to all guys who made it work.

My dual-350 Octane is really quite competitive, work-unit wise.

Our O2K churns 24/7 on 7 of its 8 300s/250s and it's easily keeping up with some of the higher clocked x86 beige beasties that the code was optmised for (speculation, but it makes me feel better)

Hey, don't knock your R5K 180.... the more SGIs we can get churning on this the better. We are team nekochan (SGI to the MAX) after all Besides we're about to crack the top 500 and we don't want Joerg to get all the glory

I hear ya bro...

Hi all, I just revive this thread instead of starting a new one.

I just installed Boinc with Seti as explained by nekonoko earlier in this thread, but I have some small problems on my Origin3400:

1) I often get this



How does Boinc detect wether a user is active? My machine currently has 16 R12K CPUs, which where at 99.5% idle while above message popped up.

2) The second problem is, that my client seems to post too many request too get any workunits at all. I always have entries like this:



It seems that my machine is asking for work two times in a row, which is then always answered with the "last RPC too recent: 6 sec".

I have it running for about two hours now, and while my x86-box is getting work-units without any problem, the Origin get's none.

Anybody experienced this problem before? And yes, I checked if there is a second instance of boinc running.

best regards and thanx in advance for any help

tuo

gtogl wrote: 2) The second problem is, that my client seems to post too many request too get any workunits at all. I always have entries like this:

Code: Select all

2006-01-15 23:00:57 [---] Insufficient work; requesting more
2006-01-15 23:00:57 [SETI@home] Requesting 69120.00 seconds of work
2006-01-15 23:00:57 [SETI@home] Sending request to scheduler: http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi
2006-01-15 23:01:00 [SETI@home] Scheduler RPC to http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi succeeded
2006-01-15 23:01:02 [---] Insufficient work; requesting more
2006-01-15 23:01:02 [SETI@home] Requesting 69120.00 seconds of work
2006-01-15 23:01:02 [SETI@home] Sending request to scheduler: http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi
2006-01-15 23:01:06 [SETI@home] Scheduler RPC to http://setiboinc.ssl.berkeley.edu/sah_cgi/cgi succeeded
2006-01-15 23:01:06 [SETI@home] Message from server: Not sending work - last RPC too recent: 6 sec


It seems that my machine is asking for work two times in a row, which is then always answered with the "last RPC too recent: 6 sec".

I have it running for about two hours now, and while my x86-box is getting work-units without any problem, the Origin get's none.

Anybody experienced this problem before? And yes, I checked if there is a second instance of boinc running.

best regards and thanx in advance for any help

tuo

I get the above messages all the time on all my boxes, IRIX and wintel. Is any work running? As for the first message about suspending computation and network activity, I get these mesages every once in a while. I do have problems sometimes with pulling new work down from the server. Some boxes are able to download new workunits and other just defer communication. I think its a problem with Seti itself.

I sometimes get work on my Win-Machine, but it is far from satisfieing. I'd say it is working maybe 2 hours a day, the rest it is either waiting to get new units or to post completed units. And I have already played around with some settings as supposed in the official boinc/seti forum.

In the last 31 days, my winbox made around 7000 "points".

With Seti-Classic, my servers and workstations where working at 100% all the time.

The Origin - until now - didn't get any work units. at this very moment, my winbox also has the status "wartende scheduleranfrage", which I would translate to "pending scheduler-request"...the status it "works in" most of the time.

Is it just me or is Seti quite broken since it moved to Boinc? Is anyone here getting work units on a regular basis to keep their machine at 100% 24/7?

best regards

tuo

I think what you may have to do is that if you're going to run a bunch of units, once you get the machine registered tell it to download 5 or 10 days of work at a time, so even if only 5% of the requests go in, it always has fresh data in the cache.

fixed the problem

I changed my boinc setting so that it runs always (didn't know I have to tell boinc this) and that it should keep a cache of 10 days.

first, my origin-client didn't update the settings, so I added them manually to global_prefs.xml

now everything works like a charm, three cpus currently at 100%, and I red in some forums that it takes some time until boinc uses all...*crossedfingers