Join BAUT teams for BOINC(Einstein-,SETI-,LHC-,...)

[Ken Vogt]

Hi Fortunate,

The changes from the Sept vintage version 4s to the latest 5.x are really pretty small, and it's mostly in the sign up process (passwords instead of account keys.)

There may have been a few minor tweaks to the GUI, but nothing to make a real difference.

So if you don't plan to join another project, there is no need to upgrade, IMO.

Oh, there are slightly more informative messages sometimes, if the project implements them, but except for LHC being often out of work, your two projects rarely have problems.

Others may have different opinions on the 5.x GUI: since I use BoincView, I don't actually know much about the sequence of BOINC Manager interface changes.

If you do try another project at some point, it probably will be best to upgrade BOINC first: at some point they want to have all account creation through the password system, which requires 5.x

As far as problems with version 5, I've been using the release version 5.2.13 without any problems (Windows, I'm assuming here.) It's been the same since November 29, which is a fairly long time for a BOINC version, and I haven't seen any unusual complaints on the BOINC boards about this one, though you are right, some people reported problems with one or other of the earlier 5.2.x versions.

Whichever way you go, Happy crunching!

[Laser Jock]

Cool, Udo!!

I'll edit above to include the "r1_"

Dr Allen has a post about Albert, which he is updating with answers to crunchers' questions, so you only have to read the first post.

- The new WU have different execution times, typically ranging from about 25% to 100% the previous execution times

- The new WU application incorporates all BOINC graphics and other bug fixes to date

- The new WU application has a slightly re-arranged screensaver, which includes our top wish-list item: a real time clock

So apparently, contrary to what I said above, -oops , WU times will vary, but will often be shorter, as Udo found.

Who gets what WU is random for the moment...

I got some of the new WU's on one of my computers. Wow, they can be really small. They take ~9,400 seconds (~13 claimed credits) whereas the old ones took ~53,000! These new WU's might work well for folks who have dropped out because they have a slower machine.

I also have some of the new ones on my home computer, but I don't know how long they will take (I'm at work now).

[Ken Vogt]

I got some of the new WU's on one of my computers. Wow, they can be really small. They take ~9,400 seconds (~13 claimed credits) whereas the old ones took ~53,000! These new WU's might work well for folks who have dropped out because they have a slower machine.

Glad you got 'em LJ!!

Only problem with use on a slower box might be that some are still normal length, and there's no way (ATM) to choose which size you get...

December 25, 2005
The new "Albert" search application seems to be working very well, so we are beginning to distribute thousands of "Albert" workunits. For more information, please see this thread in the message boards. Happy holidays!!

The indispensable honza posted this screenshot on the E@H board of the new graphics:
http://www.boinc.cz/img/Einstein_Alb...screenshot.png
But of course the graphics still way slow down the crunching -- use sparingly -- never make it your screensaver.

[Laser Jock]

Glad you got 'em LJ!!

Only problem with use on a slower box might be that some are still normal length, and there's no way (ATM) to choose which size you get...

True. My home computer (> 2.5 times faster) crunched one in 14,200 seconds. That's still ~7,000 seconds faster than the old WU's, but they sure to do vary a lot.

[Udo]

Only problem with use on a slower box might be that some are still normal length, and there's no way (ATM) to choose which size you get...

...but the datafile is still used several times for different WUs!
see this host and the WUs that are finished within ~7500 seconds. They are all from datafile 'r1_0186.5_*'. Currently 12 WUs have been taken from this datafile.

As mentioned in an earlier response, my pc at home is still using WUs from a datafile downloaded in September!

Udo

[Wolverine]

Hmm? What happened to the BAUT team page?

[Ken Vogt]

...but the datafile is still used several times for different WUs!
see this host and the WUs that are finished within ~7500 seconds. They are all from datafile 'r1_0186.5_*'. Currently 12 WUs have been taken from this datafile.

As mentioned in an earlier response, my pc at home is still using WUs from a datafile downloaded in September!

Yes Udo, and a very good point!

For each of us, it will no doubt seem that the new WUs all run at the same speed, but since they are all taken from the same master file that is to be expected.

We'll just have to wait for more experience to see if times vary for different master files.

...Or take it on "faith" (unacceptable here! ) that the new WUs really do vary between 25% and 100% of the old ones.

Hmm? What happened to the BAUT team page?

azazul posted on his forum (also now down) that he was going to grad school in physics; San Angelo University perhaps not offering a PhD program , I surmise he might be moving. His recent E@H RAC
on BOINCstats seems to show all his computers are down, activity from about Dec 17 no doubt being pendings coming in.

I assume it's a big operation moving and reassembling a big farm like his, his own webserver, etc, etc. I'm sure he and the pages will be back with us soon.
____

Meanwhile, with moi as a poor substitute, I don't think it's been mentioned that Udo was the first to go over 200,000 total credits and Binary has become the 8th person with over 100,000 (all according to BOINCstats.) And he will soon be the first BAUT member over 100,000 in seti@home.

Lance and Cyndi will be next in a week or so; though perhaps that is really just 50,000 apiece?

Anyhow, congratulations to all!

[Wolverine]

Ah, thanks Ken. I suspected something along those lines but hadn't seen the mention.

[Udo]

For each of us, it will no doubt seem that the new WUs all run at the same speed, but since they are all taken from the same master file that is to be expected.

We'll just have to wait for more experience to see if times vary for different master files.

...Or take it on "faith" (unacceptable here! ) that the new WUs really do vary between 25% and 100% of the old ones.

Ken, as far as I understood Bruce Allen in his message thread (and own observations in my results) the different datafiles are for different frequencies. Different datafiles will have different computation times, but all WUs from the same datafile should have the same computation times (on the same CPU of course).

Again my new host:
r1_0930.0 - 26.000/26.700 seconds
r1_0186.5 - 7.500/7.600 seconds
r1_1137.5 - 24.600/25.200 seconds

The two different times (26.000/26.700 for example) are an intersting fact I noticed on the new server.
Currently there is no application running on this server except Einstein@home. Therefore all computation times should be nearly the same (on an other server with a single AMD CPU the difference between the computation times is smaller than 1%).
In this case the difference is caused by the Intel Hyperthreading architecture and the dual CPU core. Windows (and Boinc) 'see' 4 CPUs of which I have limited Boinc to 3 CPUs (for reasons which are to complex to explain in this forum).
An 'Einstein' (or 'Albert' to be more precise) application running on one CPU 'alone' is faster than 2 applications running on the other CPU with their corresponding (hyperthreaded) CPU... time is spent for flushing/reloading CPU cache and other CPU equipment which is available only once!

Udo

[HenrikOlsen]

An 'Einstein' (or 'Albert' to be more precise) application running on one CPU 'alone' is faster than 2 applications running on the other CPU with their corresponding (hyperthreaded) CPU... time is spent for flushing/reloading CPU cache and other CPU equipment which is available only once!

Udo

It's still faster to run two at the same time on a hyperthreaded cpu that run them sequentially, so from a total computation point of view it makes sense to run them together.

[Ken Vogt]

Udo: I believe we are trying to say the same thing! Sorry if my post was confusing.

We agree that WUs from the same file will complete in very nearly the same CPU time.

What I meant about "have to wait" was this:

Once I get a second Albert master file, will all of its WUs run at a different speed from all those in the first?

If so, and if which type master file one receives is random, it would not be possible to recommend Albert for people with slower PCs, because if they happen to get a slow master file, they will miss a lot of deadlines, or gum up their other projects by E@H going into Earliest Deadline First mode a lot.

Of course, the project knows the speed of your box, and so E@H could implement a policy of distributing fast WUs to slow machines and vice versa, but I don't know that they have done this.

Henrik: Welcome to the thread!

I'm sure you are correct in this. I believe the folks at CPDN said that running 2 WUs simultaneously on an Intel HT is about 15% faster than running two consecutively.

[azazul]

azazul posted on his forum (also now down) that he was going to grad school in physics; San Angelo University perhaps not offering a PhD program , I surmise he might be moving. His recent E@H RAC
on BOINCstats seems to show all his computers are down, activity from about Dec 17 no doubt being pendings coming in.

I apologize for not posting here about the site going down.

Ken is right about me moving, but it is because of things not relating to school. I do plan to go to grad school, but I made that decision too late to get in this spring, so I will be waiting until fall to start (which means moving again). I plan to go to University of Texas Brownsville, depending on acceptance in the university and physics program of course. I should have the webserver back up within a few days.

[Wolverine]

I apologize for not posting here about the site going down.

Ken is right about me moving, but it is because of things not relating to school. I do plan to go to grad school, but I made that decision too late to get in this spring, so I will be waiting until fall to start (which means moving again). I plan to go to University of Texas Brownsville, depending on acceptance in the university and physics program of course. I should have the webserver back up within a few days.

No worries, glad it's on its way back and good luck with your continued education.

[HenrikOlsen]

Henrik: Welcome to the thread!

I'm sure you are correct in this. I believe the folks at CPDN said that running 2 WUs simultaneously on an Intel HT is about 15% faster than running two consecutively.

Thanks.
I should probably have mentioned that I'm basing my statement on a series of experiments on HT performance I ran a while ago when I was involved with debugging and tuning a fairly large cluster.
The result showed that the performance gain from running two processes on one HT cpu varied a lot with the software run, with everything from <5% to >50% depending on how well the software was optimised for the cpu.
The better the software was optimised, the less the gain from running two copies.
15% gain sounds like it's quite optimised to begin with.

[Ken Vogt]

First, Happy Birthday Henrik! :) And a warm and proper welcome to the Rosetta team!!

Yes a 240 node cluster is very impressive indeed!!, and I hasten to say that none of what I know about performance is based on anything like your experience, just what I read on the boards...

With BOINC of course the issue of HT performance is further complicated when you run more than one project, as all the science apps are compiled and optimized differently, by completely different teams.
_____

For very good reasons, this board doesn't permit graphics in signatures, but if you flaunt your BAUT on boards that do allow them, mundayweb now has team signatures, like this:



Unlike BOINCstats boxes, you can select colors and provide your own background image, but it is surprisingly difficult to pick an image and text color that display well together. You can just use the URL from this post, but then you will always be using my personal preferences; best is to register at mundayweb. I will put a stats box post up on azazul's board when it returns if there is interest.

Of course munday has individual stats boxes too, and also a version available as a desktop Yahoo widget, individual only so far.

Notable of course in this latest box is that seti has passed CPDN into second place; that all BAUT teams now have over 1000 credits; and our RAC has fallen quite a bit while azazul has been moving -- it should begin climbing again real soon now. :)

[Laser Jock]

A belated congragulations to Udo for passing 200,000 credits on Einstein@home.

Since the new, smaller WUs came out, I've really started building up the pending credits. I now have 49 WUs with ~1,400 credits pending. Most of the WUs are only for ~15 credits, but that's still a lot of WUs. Has anyone else seen similar?

[azazul]

I am still mostly getting WUs that are about 45 credits myself.

Also, thanks for pointing out Udo reaching 200,000. Congratulations Udo.

[Ken Vogt]

Hi LJ,

This will be off the top of my head, without checking the boards or anything, I've no time ATM:

Like azazul, My new Alberts are around 5 hours, so nothing is unusual.

But if you've got short ones, remember that they still come about 25 WUs per master file, and the other two people in your confirmation group will also get the exact same master file.

So since I'm assuming your talking about your new 4 GHz, you are likely to be way faster than the slowest member of your trio. (Sagan used the same argument for why any civilization we Contact will be way more advanced, AFAIK)

Anyway, suppose for a wild example that one of your trio is a 400 MHz box and WU speed really does scale by clock speed: then you are 10 times faster than him or her, and when you have done the 25 WUs, he/she will only be on # 3; so that one file will cause 22 or so pending WUs for you.

Just my guess....

Anyway, the credit will all eventually come around.

Let me know if I'm crazy.

[Wolverine]

Like azazul, My new Alberts are around 5 hours, so nothing is unusual.

Ditto -- my results have been similarly consistent on the newest units.

[Ken Vogt]

Yeah, Wolv, I hope my next batch is one of the shorter ones, it would be fun to see something different.


I think after sleeping on it that my argument to Laser Jock above really applies to master files of any size. Faster computers just naturally build up huge quantities of pending credit waiting for the slower ones to catch up.

I recall when Evan had the fastest boxes on the team, and we all new little about DC and BOINC (Them wuz the days, eh boys and girls?), he built up huge amount of Pending, and we worried about whether the 7 day limit applies to confirmation.

It does not of course, and so the only thing for Laser Jock and others with blazing machines to watch for is whether any of the Pendings ever become marked Invalid; which can indicate a problem.

I confess I don't look much at my results pages any more (except for the cool graphs in the CPDN spinup): If my RAC stays about the same between resource allocation shifts, I hope things are pretty much OK.

[gopher65]

I use the standard Boinc Manager which you can download from the main BOINC page. After I got over some problems upgrading from 4.X to 5.X (was partially my fault, partially the programs fault) I found that 5.X was much less buggy than 4.X. 4.X had problems reporting workunits correctly, problems contacting the server, and problems displaying messages (for me anyway). These were all fixed in 5.X by the time I upgraded.

[Ken Vogt]

Some bad news from orbit@home:

Our proposal submitted to NASA last summer has not been selected for funding.

The orbit@home project hasn't been able yet to find stable funding for its development, but we will continue to work on it with even more passion, investing as much time as possible on it. All the orbit@home users around the world can be sure that this project will start to crunch at full power at some time in the future, even if this time is getting a bit farther now. And we will also continue to apply for other funding, improving the quality of our proposals to increment the chances of getting funded.

Pasquale

The whole sad thread is here.

May the first rock o@h would have detected fall on the head of a NASA bean counter.


Better news from LHC@home: there is work again, with ~500,000 WU ready to send-- so if you suspended or detached from the project, come on back.

Also, azazul has got the BAUT team page back up, although the stats are taking a bit longer, and his forum as well.

@gopher65: I agree with you on the improvements in version 5.2.x; though I don't remember having the specific problems you mention with 4.x. The better messaging, in terms of understandability, is almost reason enough to use 5.2.x.

[azazul]

The stats pages and my forum are back up. My apologies for it being down for so long.

That doesn't sound very good for orbit@home, but at least they seem to be persistent. I am sure they will eventually find someone to provide funding.

[Wolverine]

No worries, glad to see it back up.

[alainprice]

Just joined LHC@home, and I'm already ranked 2nd in average credits....running 1 machine. Where are all the contributors? Is everyone using Einstein@home?

[azazul]

Just joined LHC@home, and I'm already ranked 2nd in average credits....running 1 machine. Where are all the contributors? Is everyone using Einstein@home?

LHC has been out of ork for about a month or so. I think during this time many switched to other projects. Some will probably slowly come and read Ken's post above about more work units being created and will begin contributing more time to LHC again.

Welcome to the team.

[alainprice]

Yeah, just switched to einstein.

[gopher65]

I have a sloooow computer, so I only run one project (einstein)

[Nereid]

... apologies if this has been answered already.

How is the RAC (recent average credit) calculated? To what extent does it include pending credit?

Do these measures vary by BOINC app, or are the constant across all BOINC apps?

Xiexie dajia.

[azazul]

... apologies if this has been answered already.

I personally don't care, I will answer the same questions over and over if they are asked.

How is the RAC (recent average credit) calculated?

RAC is an estimate for the amount of credit that a user should receive in an average day. If I understand correctly, the amount of credit received in previous days is averaged together, recent days wieghted higher.

To what extent does it include pending credit?

It does not take pending credit into account at all. It only uses the amount of credit awarded on previous days.

Do these measures vary by BOINC app, or are the constant across all BOINC apps?

As far as I know, BOINC uses the number of cycles your computer contributed to a particular work unit to calculate the amount of credit it receives. Therefore, it should be pretty much equal across all of the applications. 2 hours spent on CPDN should be equal to 2 hours on E@H.

Editted for readability