Has anybody else noticed that very recently the FreeHAL work units have increased their elapsed run times quite substantially -- roughly 35-50%? I sure hope Tobias increases the credits granted per WU in order to compensate.

Oh .... wait ..... never mind .... I see he's taken care of that already

FreeHAL should be renamed to FreeCredits, as they give their 15 credits for just 4 seconds crunching, and 1.5-2h senseless waiting. Projects like that make pages like this here obsolete.

I have to agree with Saenger on this one, from the start FreeHal has been a Farce as far as the Credits go. For a Non Intensive CPU Project to give out the Amount of Credits it does is ridiculous. As much as I like Credits I believe some Degree of Expense or Work should be involved to get it and not just have your Computer sit there Idle all day & all the while collecting credits from FreeHal just because your attached to the Project.

People are getting so attuned to these sort of Projects and feel they should pay as much as the CPU Intensive Projects do and even more. I can remember when BOINC first started out and if a Project or it's Wu's Messed up you got Zero Credits. If a Project was down you got Zero Credits too, now the People & Projects use that as an Excuse to give out extra Credits because they were down. That sort of thinking is Ludicrous really, what ever happened to if a Project was down that you just went & ran another Project until the other one was back up?

I can remember Processing for SETI&Home when it first started up for 45+ Days straight on 6 Box's without ever receiving 1 single Credit or any Extra Credit once we did start to get Credit again. If that were to happen to day People would be ready to Burn the Project/Projects down if they didn't get a Billion Credits as Compensation.

I agree, but it is a BOINC wide problem, whether it is non-cpu intensive WUs, CPU WUs or GPU WUs.

The credits granted vary from project to project, maybe the credit system should be made universal and written into the validator code, based say on Flops processed.

Very good idea, the mastermind behind BOINC, David Anderson, is planning such stuff since quite some time, but it's obviously not on the front page of the "things to do list".
And nevertheless even so he gets ridiculed by some folks who don't care about fairness, who just want as much as possible credits for as less as possible work.

And the variation isn't that much, something about the factor 2 from low pay to good pay, a GPU really does a lot more work than a CPU, it's just not as flexible. FreeHAL doesn't do anything at all for 99.9% of the time, the factor is around 500 in comparison to other projects.

Moreover, FreeHALL is not even a legitimate NCI project. Its WUs are so short that the application intentionally sleeps for some seconds at every loop, otherwise a WU would be over is mere seconds of CPU time.

Granted, this would overwhelm the server, but it's a lame way to address the real problem: if a WU is too little work, either increase the processing on it, thereby increasing the time that it takes to complete one, or increase the size of the WU, however increasing the bandwidth used.

Or take the Pirates approach:
Allow server contact just once per hour and give no more than 25 WUs per contact.

One other problem is, that during the sleep, i.e. 99.9% of the time, if you keep the application in memory, FreeHAL still uses 90MB of RAM per WU for absolutely no use at all.

Only the Linux app uses 90MB of RAM per wu the whole runtime. On Windows the RAM usage per wu is ~60MB for a short period only after the start (< 5min). The rest of the runtime each wu uses only ~5MB of RAM.

It's still a curious thing.

personally i dont care how many points the project gives or how much memory it uses or how much work it really gets done

i want STABLE CREDITS.

over the past few days we have been getting tons of credits one day only to have them all taken back the next day or 2

as far as i can tell over the past week or 2 people running FreeHal have probaly gotten zero credit for there work

Ton's of Credit a lot of them didn't deserve because the Credits are all screwed up ATM. Some People are getting 20 to 30 Times and even more Credit than they normally get Per Day. It's Free Credit so whats the big deal if they take a little back, your Computer didn't do anything to earn it in the First Place ...

I always thought credits for this project rather are based on RAM usage than CPU usage. And indeed, on older comps it would slow them down much more than most CPU-intensive projects due to the massive amount of memory needed. However, I am surprised to learn all this memory used could turn out to be absolutely useless. This would pose serious questions about the way the project was designed and the credits it gave.

I think the servers are all down now for quite a while and maybe these thoughts are already obsolete. Saenger, could you elaborate why this use of memory was totally inefficient? The last applications the project tested seem to point into the same direction, but I simply do not have enough knowledge to judge this.

After all, it's a strange development, at least.

The WU could be ready after e few seconds, if it would be done like any other CPU-intensive WU in BOINC.
Instead it is kept "alive" for up to 2h, without any technical reason, and kept in memory, thus using RAM just waiting there for something to happen. Especially if you have the preference "Keep WU in memory" set to yes, which is a must in some projects like Lattic, this RAM is blocked without any need for it, at least not needed for longer than 4-10 seconds, the time the WU really should last.

Cheers, Sänger.

Let us hope Tobias will get his feet on the ground again and continue with his project, which seemed to be looking bright. Criticism often is needed and certainly was justified in many ways, but I reckon this may have been way to much to handle of late. Especially if science questions get mixed with credit ramble, I wouldn't ever want to be a BOINC project admin for more than a second.

What exactly is the deal with this project? I'm running some NCI workunits, and seem to be getting just ridiculous amounts of credit, so much so that i feel a little guilty about running the darned thing.

I'd expect that nob-CPU units would get far less than this. I may actually stop running it

Why in the world would you want to stop running a good paying project?

Anyway, the NCI WU pays the same as the CI except it takes 4 hours to crunch instead of 1 hour.
I switched to the NCI because the CI were intetrfering with my other projects completion times,
especially Spinhenge.

Each work unit pays 55-65 credits. With two WU per core selected, the credits start to mount up.
On an 8 core computer with 2 CI WU per core, thats 16 WUs per hour x 60credits = 9600 credits per hour.
That'll put some points on the board real quick!

Hope this helps..

Sorceress

Thanks, sorceress!


I have a different take on distributed computing, I guess - it's not about the credits, but I expect the credits to follow the computing work that gets done on them.

I occasionally check stats to see how different projects are doing, but don't really care about the numbers - that's just a fun bit of the game. It's just the way I think!

I also felt the same way when I first started 'crunching' and still do in some ways. But when you have been in DC as long as I have,
the credits begin to gain meaning. I like getting the credits and the recognition and satisfaction that comes with them. Thats also
part of the 'fun'. My take is that I will do the most with what I have, but unlike many who spend thousands on high power computers
to make as many credits as possible, I keep it to the basics of doing it for the science. The credits are nice but they are just one part
of the process.

Sorceress

I would be happy just to be credited for the work that I return. About one third of my returned work simply vanishes without a trace. Posting in those forums about the problem is virtually useless.

According to the news section of the FreeHAL home page, there will be no new work for about 6 months.

The site is down now. Looks like WU's in progress will not be accepted.

I spoke too soon. The FreeHal site is back up and will remain so.

Tobias says the work units will stop in the next few days for 6 months.

Looks like it's done already.

Why the - Ratings skivelitis ??? and here at your Home Teams Base on Top of it ... Personally I'd like to see some sort of implementation that anybody that presses the - Tab has their BOINC Accounts Zeroed out across the Projects for being such Panzies. There - me for that ...

NOTE: I gave you a couple + to help offset the - you got ...

May as well consider the FreeHal Project Dead, I'm not sure if any Project has ever come back after being down for 6 month's and stayed running very long afterward once they did come back.

Too bad

Actually I was planning on detaching all my hosts from FreeHAL soon, but I was also planning on bringing project credits to 1.5 million before doing that ... I was only about ten days away from that goal

I have to agree ... consider the project dead. If Tobias doesn't have the time for the project now because he's busy with his high school graduation year, I can't imagine he'll have more free time when he begins university (assuming he plans to do so).

Collage, Job who knows what else at his age will pop up ??? I'm glad I just got 10 Mil in a few days ago anyway, I actually ran CPU Wu's for a few days to try and get over that mark not even knowing the Project was going to shut down, glad I did now ... I've seen to many Projects go down for an extended period of time, come back and within a short while go down for good ...