To BOINC or not to BOINC...

[imported_Thunder]

Well, despite having my machines set to cache 2 days of work and being attached to both SETI@Home and Predictor@Home, my first machine completely ran out of work a couple hours ago. :disgust:

The other 6 are at varying stages of being anywhere from 1 hour to 10 hours or so from doing the same.

I really like the concept of BOINC (many projects, dividing priority according to your preferences) and both the current and future projects I've heard about are 'right up my alley'. However, at the same time, after the experiences in the last week, I'm strongly considering putting it aside until the project reaches a greater level of maturity.

I understand that all new DC projects have growing pains, but is it just me, or does anyone else get the impression that the folks from SETI-BOINC don't seem to care much about passing useful information to the users? The last message was the cryptic "The log files are currently being compressed, after thats done the project should return to normal." which was the first thing posted after nearly 2 days of complete outage. So.... what's that mean? Take a couple hours? A day? 10 minutes? Frankly, at this point, what's normal? Fighting desperately for a couple WU's every day or so?

The guys at Predictor are at least passing on better information, but then they're not exactly living up to what they're saying, so I'm not sure which is worse. (They last said their system should be back to giving out work, but it's not)

I've sort of been 'out of the scene' for a while and I don't know whether this is a level of competence and feedback that I should expect from most DC projects or if this is unique to SETI-BOINC.

What are the other 'early adopters' of BOINC thinking on this?

 

[TAandy]

As far as I can tell there are four projects running on BOINC; SETI, Predictor, Astropulse and (don't laugh) Pirates@home (a beta for Einstein@home), and they all seem to be having problems at the moment (although this could just be at my end).
I'm going to keep going with SETI for a while and hope they get the problems sorted out but it seems like a bit of a shambles right now

 

[pcdeatherage]

A week or so ago, I too thought I'd see what Boinc was about...the
first 3 days went just fine - fine enough to make me want to attach,
if not for a few days, a 2nd machine just for the heck of it but
then the current batch of troubles started...

It took me a day to figure out that I needed to contine running
Seventeen or Bust (SOB) concurrently cuz the troubles continued
and continued and continued...so...don't know about the other projects
but SOB picks up the slack just fine when SETI2 is not working like it
should...

So still convinced that the new SETI will be lightyears better than
the old, I think I'll keep trying for a bit longer - and taking thier word
that the last 3 days of credits are being cached and it's just a matter of playing catch up ...

patrick.


ps: from what I can deduce from the SETI2 message boards, the guys
in charge are truly trying their best - even tho the seemingly lack of
communication is sometimes frustrating.

Be warned, there are a couple of people hanging out there that need
to work on their people skills that don't seem to realize that the
rest of us are just normal everyday people that believed what both
SETI sites told us:

from SETI-I: A new version of SETI@home based on BOINC, is now
available. We'll be transitioning to this new version (details are
here). For now, you can use the new version or stick with SETI@home
Classic.

from SETI-II: We have released a new version of SETI@home based on
BOINC, a general-purpose platform for distributed computing projects.
We encourage you to switch to the new SETI@home.

So if you're running another project, check it out and see how compatible it is with SETI-II, if you're not, I can vouch for a SOB/SETI-II combo...

 

[Dalephi]

So far I have been able to get enough work units. I have submitted about 15 workunits and I still have not received any credit for any of them. I know two other people have to come up with the same result to verify them. The SETI BOINC account page shows me as part of TeAm Anandtech but I am not even listed in last place. Hopefully one of the workunits will be verified soon.

I have done several BETA projects before and this seems to still be at that level. I think once everything is ironed out it will be great.

I will keep two of my machines on SETI BOINC and the rest on SETI 1.

 

[SoulAssassin]

My biggest problem seems to be getting workunits. OK, well I have bigger problems than that but they're outside of the scope of this. But seriously that seems to be the biggest issue. I tried running the service install and it installed it but the service won't start so that's rather frustrating as well.

 

[Rattledagger]

For service-install, atleast on XP, go into the service-control and make sure BOINC is set to run as "localsystem".

Astropulse? last seen around Christmas... except Alpha got a new sparc-version or something roughly a month ago. Later both the beta & alpha-sites has vanished without a trace... maybe to a new hidden location...

As for Seti, they're started handing out wu again, but still you're probably look at "no work available" occassionally...


They've just added these things to BOINC, and hopefully these should speed things up...:

"Database optimization in the transitioner:
add several results in a single SQL insert statement
(using values (...),(...),(...))
instead of separate inserts.
- scheduler: in scan_work_array, do all "cheap" checks
(i.e. those that don't require DB access)
before releasing the semaphore.
Thus if all the results are infeasible,
we'll acquire/release the semaphore once, not 100 times.
- re-enable team name search
- tweaked and enabled Rom's DB accelerator for the scheduler.
When a user is reporting results,
it reads the necessary fields in one gulp.
It updates only the necessary result fields.
It updates the workunit transition times in a single SQL statement.
It encloses this all in a transaction."

 

[imported_Thunder]

How did you learn all that Raddledanger??!? :Q

I've looked all over their site and can't find anything you just mentioned!

BTW, I knew about the thing with the service-install only working on XP. Unfortunately, I have no computers with XP and I'm only changing one to XP this week. At $90 a pop, I won't be doing many of them until I'm dragged kicking and screaming into that OS.

 

[Rattledagger]

The service-installing is also for NT4 & win2k, but the "feature" of most likely needing to manually change the user to "localsystem" is AFAIK only in XP.

But looks like they only was up for a very short time, since it's now back to "no scheduler responded"...

 

[imported_Thunder]

Nope, I've tried the installer that came with the 3.20 BOINC installer and it gives me errors when trying to install the service from the cli.

And no, it's not any silly administrator privileges thing... I AM THE ADMINISTRATOR. (which makes installing BOINC so easy for me at work. )

 

[Rattledagger]

Well, I'm running the service on NT4 & win2k so it's possible...

But some also seems to have problem, maybe this trick will work?

 

[Dalephi]

I have been running SETI BOINC for about a week. I have submitted about 15 work units. None have been verifed as of yet. TeAm Anandtech is listed as the team that I am a member on. When I look at the members I do not show up on the TeAm. Do I have to wait for the first work unit to be verified before I will show up on the TeAm?

 

[soni]

Dalephi, yes you wont show up on the list, before the first result are verified and credited to you.

 

[Dalephi]

Thanks Soni. This may take a while.

 

[imported_Thunder]

Thanks Rattledagger! I'll give that a whirl when I get back to the office!

 

[CiSmAXz]

An easier way then running it as a service is to run it as a scheulded task. Go to control panel -> scheduled tasks, select the exe, start it when computer starts , check box to open advanced properties when you click finish, settings tab - uncheck stop if runs more than 72 hours. This will run it as soon as your PC starts even without being logged in.

This worked great for classic seti. It is also good because you can schedule it to run whenever you want (like not during the day when users are on their PC's)

I am also very frustrated about not getting any work units for the last few days. There should have been a much better plan in place in order to transition over from Classic SETI.

 

[Spacehead]

Since we have a bunch of SETI/BOINCers all here i'd like to get your input on the stats. Check out my first stats thread here.

I'm thinking of leaving the team stats the way they are there but was contemplating adding to the individual stats part, showing:
Rank/Name...Total Cobblestones...Daily Average...Position Change


Thoughts???
Anything else that could be added/dropped?

What is "recent average credit"(team & individual) & would you like it shown? Or would you like the Daily Average instead?

 

[CiSmAXz]

It looks great so far. I would definitely add some type of individual stats. Total, daily average and position chancge sounds good.

 

[imported_Thunder]

Originally posted by: Rattledagger
Well, I'm running the service on NT4 & win2k so it's possible...

But some also seems to have problem, maybe this trick will work?

Rattledagger,

Since the S@H BOINC site has been down (at least every time I've tried to access it) for nearly a week, I've not been able to read this 'trick'. Any chance you could convey to me what it is or how to do it.

I've tried uninstalling BOINC, then re-installing and making sure not to set the GUI to run automatically, then booting clean again and STILL getting that same dang error every time I try to install the cli as a service.

 

[Rattledagger]

Since I didn't need to do any trick, can't really know if this trick even works, but it was AFAIK doing some changes in the registry, but no idea what changes.
So you'll have to wait till the web is back up. (and hopefully this happens soon...)

Hmm, looks like someone just turned off all machines, since not even "classic" is currently accessible, and the network-graph dropped to 4.16 kbit...

 

[CyGoR]

They say the project is 'up' again, but I can't connect to send/receive anything for BOINC..

August 2, 2004
The project remained down all weekend, but we are back up now. To add insult to injury, the entire lab lost network connectivity to the rest of the world for a few hours this morning. In light of the unexpected length of this last outage, we are looking into mechanisms that allow us to keep certain parts of the project up while others remain down for repair.

 

[networkman]

From what I'm told, getting BOINC'd can be a good deal of fun.

 

[imported_Thunder]

Originally posted by: networkman
From what I'm told, getting BOINC'd can be a good deal of fun.

Oh YES!

...or at least it would be, if all things were working right. Nothing worse than having something go wrong with your BOINCer.

...except possibly that problem of wanting to BOINC, but having nothing available to BOINC!

 

[CiSmAXz]

I am soooooo close to bailing on BOINC and joining another DC project. This sucks....

 

[TAandy]

Signs of life at last, got 18 units
Won't last too long but hopefully it's a start (don't think I'll hold my breath though :roll: )

 

[Rattledagger]

August 3, 2004
While the project is up and functioning normally, the database and web servers are overloaded. Because of this you may have noticed difficulty in transferring work and loading certain web pages. We're tuning these servers and trying to find the bottlenecks now.

August 4, 2004
We discoverd that the sheer number of files in the upload directory was killing NFS performance for the file_upload_handler. We started with a new empty upload directory and the file_upload_handler performed much better. The rate of connection drops on the upload server dropped to zero (most of the time). Of course we need to get all those older results back into play somehow. We are preparing our new SnapAppliance 18000 to host the database. We need to fit it into the power load in our already fairly full data closet, configure the DB volume, run some tests, and then copy over the DB files. We will post the progress as we go.

A method of shattering files among multiple directories was just added by David Anderson, so hopefully shouldn't have any more upload/download-problems.
But till the database is moved to the new box, the most common replies will still be "no schedulers responded" & "no work available".