The limits should be set so I can get at least 24 hours of work in a cache.

That is a great idea except that the data base can't handle it.

---

Problem here is the limit of 100 tasks, which doesn't make much sense in my opinion. Would be better if Seti would respect the cache settings in Boinc manager.

Agreed. Every time I mention this here, however, I get jumped on as being a dummy or selfish?

Sigh .... how quickly we forget.

Back in April 2013, right after the main servers were moved to the Campus Data Center (CoLo), Matt posted this message about the move, and why there would still be limits on Tasks. So until Matt or Eric say that the hardware issues have been solved, I would not expect significant changes to the Task allocation limits.

---

Donald
Infernal Optimist / Submariner, retired

Back in April 2013, right after the main servers were moved to the Campus Data Center

Has it really been that long? How time flies.

---

Problem here is the limit of 100 tasks, which doesn't make much sense in my opinion. Would be better if Seti would respect the cache settings in Boinc manager.

Agreed. Every time I mention this here, however, I get jumped on as being a dummy or selfish?

Sigh .... how quickly we forget.

Back in April 2013, right after the main servers were moved to the Campus Data Center (CoLo), Matt posted this message about the move, and why there would still be limits on Tasks. So until Matt or Eric say that the hardware issues have been solved, I would not expect significant changes to the Task allocation limits.

Good information. I haven't watched the SETI moves as some of you have and I'm not an IT tech, but even I can understand that logic. Maybe if we threw some money at the issue it would be solved (tongue in cheek of course). I, for one, am completely satisfied with the performance of my rig and my RAC. If I could throw money at that I wouldn't be sitting at this console. Or would I? Anyhow, tested my hardware, took on Rosetta and Milkyway, and I am a happy cruncher...

Peace and happy crunching to all.

---

I'm almost not sure if throwing more money at the problem would inherently fix the reasons for the in-progress limits. I know the crew have been investigating and diagnosing the problem(s) for quite some time now, and I still don't think there's any real consensus on whether it is hardware or software that is the main limitation.

The hardware seems like it should be able to handle the I/O load just fine, but you never know how efficient the drivers are in certain OS environments. I think it's much like SSDs that claim they can do 90,000 IOPS, but when you benchmark them, you struggle to get more than 5,000. I know when I tested my SSD, I got like 5700 IOPS on 4K random read and the official specs say 90,000. Someone here on the forum said that it comes down to queue depth as well as the width of the queue, so I tried the test again and when I cranked the numbers up to something like a queue depth of 64 and a width of 4 or 8 parallel reads, I got upwards of 75,000 IOPS, but basically, that sort of load to get that kind of performance out of an SSD is a situation that you would only stumble across in a real-world environment maybe once, out of a couple hundred million databases, and even then, that kind of load would only last for a second or two and then come back down to more reasonable levels.

The point is, you just don't know if it is truly a hardware limitation, or if it is a driver limitation. And we kind of already know that the DB itself (software) is well into uncharted territory and the particular engine this DB is built on wasn't ever really intended to be as large as this one is, so.. not even the people who made the DB engine really know how to improve its efficiency on this scale. Plus, it doesn't help, either, that I think the project doesn't still have a support contract with IBM for answering questions about the DB itself, so they're kind of on their own.

I'm sure there are some DBs on the market right now that could absolutely handle this kind of load, but those cost a lot of money.

So.. it could be something money could fix, or it might not be. Maybe Informix isn't the problem and it's just the mixture of the software and the drivers, or maybe the hardware doesn't actually perform as well as it seems that it should on paper. We just don't know enough about what the problem actually is in order to suggest solution paths to look into. And when we're talking this much money, you don't just blindly keep throwing money at it until the problem is solved.. you have to have some reasonable proof that the path you intend on going down is the right one, and I think that's the stage the staff is still sitting at.....all the while working on 15 other things simultaneously on their already-limited available time.

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

As I understand it our major problems are data acquisition and the quality of the data.

---

It must be Friday......... ";D)

---

"Sour Grapes make a bitter Whine." <(0)>

To all believing crunching AP WUs is more productive in RAC I give you this....http://setiathome.berkeley.edu/result.php?resultid=4436065248.

Didn't take long but not much credit either....0.31 ?

---

"Sour Grapes make a bitter Whine." <(0)>

Well, don't expect any AP's. Astropulse science database marvin, have been Disabled for Dog knows how many hours.

And still down. Over two days now.

Does anyone know what's up with it?

Is it just Grumpy like our Swede? :)

Well, don't expect any AP's. Astropulse science database marvin, have been Disabled for Dog knows how many hours.


And still down. Over two days now.

Does anyone know what's up with it?

It should obvious to all that we are expected to pay a price for the recent abundance of APs. The beatings will continue until morale improves.

---

Well, don't expect any AP's. Astropulse science database marvin, have been Disabled for Dog knows how many hours.

And still down. Over two days now.

Does anyone know what's up with it?

Is it just Grumpy like our Swede? :)

According to the SSP, it's back online again..

P.

---

And now they are splitting a bit, are things to get good soon?

---

It should obvious to all that we are expected to pay a price for the recent abundance of APs. The beatings will continue until morale improves.

Please Sir, may I have another?
:)

---

Some AP have been sent 2:
- HD3450
- HD3650

but they were aborted by my client:
http://setiathome.berkeley.edu/results.php?userid=9989358&offset=0&show_names=0&state=6&appid=

anybody know sthg about it?!
is there AP for a single-point precision cards?

---

non-profit org. Play4Life in Zagreb, Croatia, EU

Some AP have been sent 2:
- HD3450
- HD3650

but they were aborted by my client:
http://setiathome.berkeley.edu/results.php?userid=9989358&offset=0&show_names=0&state=6&appid=

anybody know sthg about it?!
is there AP for a single-point precision cards?

What ATI AP application have you got installed on those two hosts?, the HD3450 and HD3650 don't support OpenCL, you can't run OpenCL apps,
only the Brook+ Hybrid Astropulse app.

Claggy

those device do have Lunatics app installed...so they are Brook+ hybrid ready AP app ready!

but why were they aborted?

---

non-profit org. Play4Life in Zagreb, Croatia, EU

have Lunatics installed for use on 3450 & 3650 cards...4 Brook+ calculations...but got some errors:
http://setiathome.berkeley.edu/results.php?userid=9989358&offset=0&show_names=0&state=6&appid=

all ERRORs are: 201 (0xc9) EXIT_MISSING_COPROC
:/

---

non-profit org. Play4Life in Zagreb, Croatia, EU

have Lunatics installed for use on 3450 & 3650 cards...4 Brook+ calculations...but got some errors:
http://setiathome.berkeley.edu/results.php?userid=9989358&offset=0&show_names=0&state=6&appid=

all ERRORs are: 201 (0xc9) EXIT_MISSING_COPROC
:/

Above link will only work for you when logged in. Please use links with this form when posting :
http://setiathome.berkeley.edu/result.php?resultid=4444476507

Check the first few lines (20-30 after restart of BOINC) of your event log if your GPUs have been detected correct. The error you found points to a problem with BOINC not recognizing the GPUs.

---

_\|/_
U r s

have Lunatics installed for use on 3450 & 3650 cards...4 Brook+ calculations...but got some errors:
http://setiathome.berkeley.edu/results.php?userid=9989358&offset=0&show_names=0&state=6&appid=

all ERRORs are: 201 (0xc9) EXIT_MISSING_COPROC
:/

Above link will only work for you when logged in. Please use links with this form when posting :
http://setiathome.berkeley.edu/result.php?resultid=4444476507

Check the first few lines (20-30 after restart of BOINC) of your event log if your GPUs have been detected correct. The error you found points to a problem with BOINC not recognizing the GPUs.

both GPUs were detected correct! even though my computer r @ home right now...will post u d results! ;)

errors:
http://setiathome.berkeley.edu/result.php?resultid=4444579790
http://setiathome.berkeley.edu/result.php?resultid=4444476507
http://setiathome.berkeley.edu/result.php?resultid=4444476532
http://setiathome.berkeley.edu/result.php?resultid=4440182981
http://setiathome.berkeley.edu/result.php?resultid=4437764137
http://setiathome.berkeley.edu/result.php?resultid=4437649661
those r only 6/13

---

non-profit org. Play4Life in Zagreb, Croatia, EU

both GPUs were detected correct! even though my computer r @ home right now...will post u d results! ;)

Startup messages please, those lines where the coproc features are printed.

And if you can find the lines in stdoutdae.txt where boinc tried to start one of those tasks post them, please.

---

A person who won't read has no advantage over one who can't read. (Mark Twain)