View Full Version : OProject@home

9th October 2012, 06:59 AM

It's relatively new, but I haven't seen any issues in the 24 hours I've been attached to it.

From their main page:

About OProject@Home

OProject@Home is a research project that uses Internet-connected computers. You can participate by downloading and running a free program on your computer.

The main idea of this project is the analysis of algorithms.
OProject@Home based at OLib Library. The library is open and available in the code.google.com SVN repository.

10th October 2012, 11:50 PM
Work has been steady....

Only issues, more of a convenience than anything, the progress bar stays at zero and some applications don't have estimated times but everything finishes well before the deadlines.... There seems to be somewhat of a delay between updating personal and team statistics.

18th October 2012, 12:21 PM
I've been running it for over a week now with no problems. The admin is very active and responds to questions and suggestions promptly. Seems like a project that should get serious consideration for inclusion in the Vault.

20th October 2012, 12:07 AM
The project's only been running for 2 months. Think it's a bit premature. Adding it to the Vault at this point would most likely overwhelm it and cause "Vault Syndrome". Give it a chance to stabilize and cure a few growing pains.

12th November 2012, 06:16 PM
I agree. Looks promising, but they have, for instance, optimized one of the apps last week to get a 500-fold improvement. A good thing, of course, but it shows that it is a very immature project,so wise to wait a while. I'd say give it another 2 months or so, and let's keep an eye on how things develop.

15th November 2012, 05:12 AM
Is there a Syndrome called Vault Syndrome? I don't think the Vault impacts that much on a project but I wouldn't really know as I don't take much notice these days.

Lets keep an eye on this one.

15th November 2012, 01:18 PM
Is there a Syndrome called Vault Syndrome?

AFAIK coined by fractal here:


Decided to try the project and it's a mess right now. Some of the apps don't work and the one mentioned above has a standard version that does work and a few optimized versions (even more than 500 times faster) that appear to work but give different results. Quite a few users were using the optimized apps via app_info.xml to score big. Looking at the top machines it appears that as of sometime yesterday the WUs from these apps aren't validating anymore. To add to the confusion these bad apps were posted by the project scientist (but had to be compiled by the user). Go figure. This one sure does need some time to mature IMO...

16th November 2012, 09:11 AM
Personally, I think it is too soon to consider this project. I have not run it, just perused their site, read some on the forum, followed some links.

The main deciding factor for me was simple time. The project is 3 months old. The forum's first user joined on 11 Aug 2012. I'm not sure I'd consider a project out of Berkley or Rechenkraft that was only 3 months old!

That said ... this is a project to watch. There are a lot good points to this project ... and one very important point that may be very good or unfortunately not so ... so that first ...

OProject is the brainchild of Łukasz Świerczewski (http://goldbach.pl/~lswierczewski/CV.pdf), project scientist, forum admin, developer, tester (alpha, beta, gamma, etc.) and chief bottle washer. He is a 23 year old IT student at a Polish university. All the potential problems associated with having a project (every last nut and bolt) depend on a single, indispensable resource are a possibility. As are the (far more rarely seen) potential gains of said situation.

I don't read polish, so following his resume/bio was a bit of an adventure ... the bits I think I managed to puzzle out. The upshot was the impression of a (hyper-)focused (on math) individual, a leader, possibly brilliant (there is a language barrier for me), almost driven. But none-the-less he is right at the point where life choices are being made ... and that could impact the project since he is the indispensable resource.

Good points:

extremely responsive admin on the forums (and elsewhere)
extremely impressive array of supported platforms

Windows, Solaris, FreeBSD, OS X, Linux, Android, PS3

results are publicly available in the form of "olib", a " High performance library of mathematical functions and algorithms" (http://code.google.com/p/olib/)
olib is updated regularly (ok, daily in the source trunk, every few weeks real time)
all the bits and pieces presenting a BOINC project (the forums, the home page, getting a new user set up, being responsive, even looking cool and different) look like a far more mature project.
There are several sub-projects.

Bad points:

There are still plenty (way plenty) problems to work out ... yet those problems appear to be being worked on, or rather problems have arisen in the past, been addressed, leave the queue while new problems come in.
It's a really young project! - with a very broad scope. The concept of "Vault Syndrome" could be very real here.

I think I will run a few units, both to check that part out and to support the project. It's worth a deeper look (over a more extended time).

my 2p

18th November 2012, 11:44 AM
Tried running Oproject again, this time allowing all the apps. One app was flagged by Comodo AV as a serious threat. The weird engine app completed and validated after nearly 3 hours runtime and then received no credit because the wingman was using the anonymous platform which ran the same WU in 56 seconds. The project scientist flagged this behavior as a bug 5 weeks ago but hasn't bothered to fix it. This project is even more of a mess than I thought before :-(

Edit: The ALX app was the one flagged by the Comodo Antivirus module.

Edit 2: Thought I'd do a few minutes of detective work to see how anyone can get this to work. Of the 2 who initially reported this project as running with no problems: 1 has no WUs showing (has dropped the project) and the other currently has 17 WUs either with errors or aborted and 4 WUs waiting for validation, all with errors from multiple wingmen. No current WUs at all marked valid. So I surmise that things are going downhill fast at Oproject. Maybe in a year or two...

Edit 3: OK, a little more detective work, wondered how some people are getting credits. Of the 20 top users, 18 have hidden machines, but even these show up as anonymous under top machines. It appears currently that the only app (of 5) that works to validation and scores points is ALX. Of course this is the one that Comodo flags as a serious malware threat. Not sure if it's a false positive, sent it to Comodo for analysis.

I'm Beyond frustration with this project, OK bad pun ;-)

18th November 2012, 03:04 PM
I haven't ran it in awhile since it went down. Moved back towards some Vault projects. When I last ran it, ALX was being detected as a threat on certain tools, Windows and McAfee didn't detect it. ALX was in a testing phase, it ran as non-cpu intensive and didn't actually crunch anything. Haven't bothered looking back into it....
One thing I didn't quite like about it, with as new as it was, the runtimes for apps increased alot rather quickly in the short time I was running it. My time references might be off because of the last PrimeGrid challenge, but tasks went from a few hours to 10+ hours for me. And the ALX app, tasks aren't supposed to take a thread. Sometimes it would use the thread, sometimes it wouldn't. I know some people complained in the forums about not being able to get work while others were getting work from apps not selected, I never experienced that, but I haven't been running it too much recently, either. The last few days, I haven't gotten any work.
And I really didn't like how stats would update frequently for some parts of the site and project but not for others, and I might have projects mixed up again, but I really think it was this one, your results didn't stay visible on the list of tasks for too long....

19th November 2012, 03:31 PM
Having run (just a few) tasks there confirms the remarks of Beyond and RHMash, above. And shows another thing : cheating is going on on an enormous scale. Many hosts show BOINC benchmark values of 1000 million ops for both integer and floating point. On at least one subproject, if you run a WU to completion, and have the bad luck of your wingman being one of those hosts, you get about 1 credit per hour of CPU (the cheater getting this one credit as well, but these hosts appear to finish a WU in seconds that runs hours on a real machine). This has been mentioned in their forum, but no action taken (yet). I understand that there's priorities on getting the project running with code optimisations etc, but if admins, no matter how responsive on other issues, ignore cheating I think is a definite no-go for inclusion in DC-Vault, which is ALL about gathering points.