E@H Server Status back online !
6 Feb 2007 13:21:00 UTC
Topic 192394
(moderation:
After 28 days the E@H Server Status Site is back online ! Congratulation to the people , who run the servers. I think, they do a good, stressing job.
But still I believe, more infomation about the situation would be fair in regard to the people behind the computers all over the world. This can be done without unacceptable more workload. The hosts don´t require this information indeed, but the people behind them are humans, and they like to take in some way part in what is going on.
In the near past there have been several remarks in this forum with the same tenor.
Kind regards
Martin
E@H Server Status back online !
)
I agree with you, Martin. We had the same communication lapse during the 'Curse of 32' problem a few months back. It's a great project otherwise!
Curse of 32? I've never heard
)
Curse of 32? I've never heard of that one and now you've made me curious. Unless the Illuminati are in some way involved (and then it would be 23 rather than 32) I have no idea what you're talking about...
RE: Curse of 32? I've never
)
Ah, but 32 is just another cunning disguise …
I believe it was intended as a reference to the 32 WU/CPU quota that was ‘handcuffing’ fast systems running Akos F.’s optimized apps during the S4 run a few months ago.
Oh okay, I seem to remember
)
Oh okay, I seem to remember that now you mention it. Didn't have much to do with it myself because my only "fast" host always got "long" WUs at that time, but I read about it in the message boards. Thanks for explaining!
For anyone who doesn't
)
For anyone who doesn't remember, AKOS s4 longs were about the same size as s5 shorts. Bruce Allen's stated reason for not increasing the quota limit was fear about overloading the servers. Given the late unpleasantness, does anyone here begin to think that maybe, just maybe, he might've known what he was talking about?
RE: Curse of 32? I've never
)
According to your profile, you joined in August 2006. The Curse of 32 refers to an earlier era - the S4 science run - and not the S5 run which was in full swing when you joined. In the earlier S4 run, as others have mentioned, Akos Fekete, a very knowledgeable user was able to "reverse engineer" the code of the science app to make very dramatic improvements in crunch times so that results were being completed 5 - 7 times faster. That was happening almost a year ago. Fast machines were able to completely swallow their daily quota of 32 results per CPU in much less than a day. These machines had a high risk of being idle for a lot of the time if they only supported EAH.
A particular user who penned the opening post in this thread was quite vocal in a quite amusing way about how this problem affected him and what he thought about the project staff and all things BOINC in general. I found it quite amusing to re-read the thread and relive those heady days of the latest AkosF optimisations that were flowing thick and fast.
The first time the term "Curse of 32" was actually used was in this particular post in the same above-linked thread although there were many other subsequent threads where it was also used. It became something of a Tweakster trademark - particularly when he founded a team over at Rosetta called (you guessed it!!) Curse of 32. Sadly, for quite a while his crunching power was lost to the EAH project and Rosetta benefitted. In more recent times I notice he is back with a vengeance, powering up the Top Participants list. He is #22 on RAC and #18 on total credit. I for one am very pleased to see him back.
Cheers,
Gary.
Thanks a lot for explaining
)
Thanks a lot for explaining and for posting those links.
I seem to remember people complaining about finishing their S5 WUs in less than a day, too... but I don't think there were optimized apps involved, just lots of Core 2 Duos and fast Athlons ;-). iirc, that was when the quota was raised to 72 WUs per CPU per day.
RE: but I don't think there
)
In the S4 run, most participants didn't know about or were unwilling to use the "unofficial" optimised apps so that it was really only a select minority who were afflicted by the Curse of 32.
For the S5 run, Akos Fekete was employed by the project to use his expertese in the design of the official app. It's a long story but there were further incremental optimisations that were performed as well, most of which I believe have made it into the official app. There wasn't too much of a problem with daily limits when the fast machines were essentially getting mostly "long" work units. Once those fast machines started getting too many "shorts" there were obviously potential problems again with running out of work amongst other things :).
I'm not really aware of the more recent changes in daily limits as I don't have any machines sufficiently fast and modern to be troubled by the limit, whatever it is :). At the time of the "Curse of 32" era, some of those slow machines did have daily limit problems so I did feel for Tweakster in his anguish. My solution was to use the BOINC Studio client which had the ability to "pretend" that my machines had 4 CPUs so I was able (if needed) to download up to 128 results per machine. That solved the Curse of 32 for me.
Cheers,
Gary.
Gary: Thanks for the
)
Gary: Thanks for the memories. It would appear that any worries about server overload then, were completely dismissed just recently as we transitioned. "Remember December?" What fun.
Warmest Regards-tweakster
RE: After 28 days the E@H
)
That is great, but it would be even greater if the graphs at homepage.hispeed.ch/einstein
revived. Anyone know the owner? Or does that rely on something from Einstein which has not been restarted?