Daily Quota Exceeded???
15 Nov 2004 3:17:18 UTC
Topic 187091
(moderation:
When did the daily quota go into affect?? I only use my one machine and it does about 32 or so results a day.
Einstein@Home - 2004-11-14 20:01:24 - Message from server: No work available (daily quota exceeded)
Daily Quota Exceeded???
)
> When did the daily quota go into affect?? I only use my one machine and it
> does about 32 or so results a day.
>
> Einstein@Home - 2004-11-14 20:01:24 - Message from server: No work available
> (daily quota exceeded)
>
It seems that e@h is trying to fix machines that trash WU and have implemented a mimimun RPC time and also a WU daily limit.
Like other projects that will need to look at the daily limit in light of some of the faster machines.
I think other projects use 50 WU per day.
73 de Peter VK3AVE
Yes S@h uses the quota of 50
)
Yes S@h uses the quota of 50 per day but one wu on my machina lasts 2.75h so that are 50*2.75=375.5h / 0.75 (for one E@h) = 183,33. So that would be an equal amount ;-).
Greetings from Germany
Basti
Join Ad Astra
> Yes S@h uses the quota of
)
> Yes S@h uses the quota of 50 per day but one wu on my machina lasts 2.75h so
> that are 50*2.75=375.5h / 0.75 (for one E@h) = 183,33. So that would be an
> equal amount ;-).
>
>
>
It would be great if fast machines that have a proven track record of returning good WUs get a higher quotum... if not people probably will devise some kind of a caching mechanism using different host IDs to get WUs.
- Marco
Team Canada
As I understand it, BOINC
)
As I understand it, BOINC knows which wu goes to which computer and to be valid the wu has to come back from the computer is was assigned to. If it doesn't, it is not counted.
> As I understand it, BOINC
)
> As I understand it, BOINC knows which wu goes to which computer and to be
> valid the wu has to come back from the computer is was assigned to. If it
> doesn't, it is not counted.
>
Exactly. So maybe all that you need is have boinc run from more than one folder (c:\boinc1\, c:\boinc2\, etc.) to get unique Host IDs for each client. Very simply: when the boinc client in c:\boinc1\ completed its 50 WUs your control script will start the client with a fresh Host ID in c:\boinc2\ etc.
If I were BOINC/E@H I'd put a limit on
- outstanding WUs (say max 10 or so, report back and you'll be able to get more)
- # of error WUs (say >10 errors on a given day invalidates that client and it will not be able to get any more WUs for that day)
- # of "bogus" results (say >3 fake results [results that aren't the same as the other 2 clients who processed this WU] = ban)
instead of limiting the #WUs to 50.
- Marco
Team Canada
I apologize for not
)
I apologize for not responding sooner to this thread. What happened was this. About four days ago I found that many of the results were being returned with errors by a few users. Some were returning one bad result every minute for an entire day: thousands of bad results in total. This has a bad effect for everyone because it slows down the process of reaching a quorum and validating results. Typically this was happening because of download problems. For example the user no longer had disk space remaining for downloading input files.
So I put limits on the following quantities. These are the current values;
they are slightly larger than the initial values were:
max_wus_to_send: 32
min_sendwork_interval: 600 seconds
daily_result_quota: 64
The first (32) is the maximum number of workunits that you can get in a single 'connection' to the server. The second (600) is the number of seconds after getting work that you have to wait before you can get additional work. The third (64) is the maximum number of results that you can return per day.
Note: these are PER HOST, not per user.
If you are finding that these limits are too small, please post another message here and explain why. I would be happy to increase them somewhat - I just need to understand why this would be a good idea!
Thank you for your patience in this. These sorts of issues are exactly why Einstein@home is currently in a 'testing' not in a 'production' phase. We (the designers and administrators of Einstein@home) need to understand better how to set these kinds of parameters and tune the system so that it works well.
Director, Einstein@Home
I don'[t have this problem
)
I don'[t have this problem because I don't have a dual that is fast enough. My fastest machine, a Pentium 2Ghz, will do a WU in about 45 minutes. This is about 32 per day for a single CPU system. A dual of the same speed would be able to do 64 per day. I am also crunching for enough projects so that I don't get even close to the maximum, however, some of your crunchers may hit the maximum on a regular basis.
One of the S@H crunchers has a machine that has 14 CPUs.
BOINC WIKI
> I don'[t have this problem
)
> I don'[t have this problem because I don't have a dual that is fast enough.
> My fastest machine, a Pentium 2Ghz, will do a WU in about 45 minutes. This is
> about 32 per day for a single CPU system. A dual of the same speed would be
> able to do 64 per day. I am also crunching for enough projects so that I
> don't get even close to the maximum, however, some of your crunchers may hit
> the maximum on a regular basis.
>
> One of the S@H crunchers has a machine that has 14 CPUs.
OK, thank you, I see the point. We really need a maximum workunits PER CPU. I'll see about implementing this or alternatively boost the quotas somewhat.
Director, Einstein@Home
Here's an idea, whether it's
)
Here's an idea, whether it's practical or not I don't know...
How about workunit quota based on the host's benchmark scores? Faster hosts could download more workunits.
What you are really trying to protect against is a host going nuts and spewing out a lot of trash, but there obviously faster machines (not necessarily overclocked) out there which can legitimately handle a lot of work. This would also protect against a slow machine going nuts and running through an arbitrarily high number of workunits.
> Here's an idea, whether
)
> Here's an idea, whether it's practical or not I don't know...
>
> How about workunit quota based on the host's benchmark scores? Faster hosts
> could download more workunits.
It's a good idea. Like you, I don't know if it's practical, but I'll find
out.
> What you are really trying to protect against is a host going nuts and spewing
> out a lot of trash, but there obviously faster machines (not necessarily
> overclocked) out there which can legitimately handle a lot of work. This
> would also protect against a slow machine going nuts and running through an
> arbitrarily high number of workunits.
Yes. It's a good idea.
Director, Einstein@Home