Change the dealnes to match SETI@home

Stolichnaya
Stolichnaya
Joined: 21 Feb 05
Posts: 1
Credit: 118,372
RAC: 0
Topic 188435

I run multiple projects and Einstein has a different deadline from Seti.

Make it 2 weeks.

Keck_Komputers
Keck_Komputers
Joined: 18 Jan 05
Posts: 376
Credit: 5,744,455
RAC: 0

Change the dealnes to match SETI@home

Each project will determine the deadline based on their needs. Different projects have different deadlines. I have at least 5 different deadlines amongst the projects that I am currently running. The shortest is 1 day the longest is 1 year.

BOINC WIKI

BOINCing since 2002/12/8

unkx80
unkx80
Joined: 22 Jan 05
Posts: 3
Credit: 1,101
RAC: 0

More like E@H seriously

More like E@H seriously underestimates the amount of time needed for my computer to complete a WU. As a result, my machine often returns the result a few days after the deadline, and meanwhile the E@H scheduler would have sent out another WU which would not be neccessary otherwise. This can't be good for the project.

Marty
Marty
Joined: 20 Feb 05
Posts: 1
Credit: 2,883,242
RAC: 0

The one week from einstein is

The one week from einstein is too short. I will detach the from the project because of this.

A Seti workunit is done in 4h on my pc and a Einstein unit needs 6.5h. But for Seti i have a two week deadline and for Einstein one week. This sounds not good.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5,206
Credit: 43,309,648,714
RAC: 44,776,587

> More like E@H seriously

Message 7898 in response to message 7896

> More like E@H seriously underestimates the amount of time needed for my
> computer to complete a WU. As a result, my machine often returns the result a
> few days after the deadline, and meanwhile the E@H scheduler would have sent
> out another WU which would not be neccessary otherwise. This can't be good for
> the project.

Actually it's not E@H that causes this problem. E@H doesn't know how long you will have your computer running and how many other projects (if any) you might choose to support. You are the only one that knows those answers so you have to take some responsibility for setting your preferences in a reasonable manner.

It is actually Boinc, under instruction from you, that makes a decision on how much forward work to store up. The instruction comes from your choice of your "connect to network" interval preference in your general preferences. Unfortunately Boinc always "overestimates" so you have to be smart enough to try a low time interval (like say 0.1 to 0.5 days) and then adjust it slowly until you get an optimal value for your cache of work so that there is no risk of it expiring.

Whilst I might agree with you that the deadline is too short, we just need to learn to live with it. The developers have their reasons and they are stated in the FAQ. The developers are very much aware of the sentiments about this and have indicated that they may change the deadline in the future. If we want to be associated with this project, we need to accept the rules. After all we were not forced to join.

Cheers,
Gary.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5,206
Credit: 43,309,648,714
RAC: 44,776,587

> The one week from einstein

Message 7899 in response to message 7897

> The one week from einstein is too short. I will detach the from the project
> because of this.
>
> A Seti workunit is done in 4h on my pc and a Einstein unit needs 6.5h. But for
> Seti i have a two week deadline and for Einstein one week. This sounds not
> good.

Are you saying that you can't find 6.5 hours in a week to complete a work unit? If so then I guess you really aren't interested in this project for its own value and perhaps it might be best for all concerned if you did detach. I'm guessing that you really only want to support Seti and so you have allocated E@H such a low share of resources that the work can't be finished within the time limit. Please don't blame E@H for what is really your own personal choice.

Cheers,
Gary.

unkx80
unkx80
Joined: 22 Jan 05
Posts: 3
Credit: 1,101
RAC: 0

> > More like E@H seriously

Message 7900 in response to message 7898

> > More like E@H seriously underestimates the amount of time needed for my
> > computer to complete a WU. As a result, my machine often returns the
> result a
> > few days after the deadline, and meanwhile the E@H scheduler would have
> sent
> > out another WU which would not be neccessary otherwise. This can't be
> good for
> > the project.
>
> Actually it's not E@H that causes this problem. E@H doesn't know how long you
> will have your computer running and how many other projects (if any) you might
> choose to support. You are the only one that knows those answers so you have
> to take some responsibility for setting your preferences in a reasonable
> manner.
>
> It is actually Boinc, under instruction from you, that makes a decision on how
> much forward work to store up. The instruction comes from your choice of your
> "connect to network" interval preference in your general preferences.
> Unfortunately Boinc always "overestimates" so you have to be smart enough to
> try a low time interval (like say 0.1 to 0.5 days) and then adjust it slowly
> until you get an optimal value for your cache of work so that there is no risk
> of it expiring.
>
> Whilst I might agree with you that the deadline is too short, we just need to
> learn to live with it. The developers have their reasons and they are stated
> in the FAQ. The developers are very much aware of the sentiments about this
> and have indicated that they may change the deadline in the future. If
> we want to be associated with this project, we need to accept the rules.
> After all we were not forced to join.

While I kind of agree of what you say, I would say that I find P@H WUs estimated at 8h but I often complete them using about 4h while E@H WUs are estimated at about 11h while my computer usually takes about 22h to complete. By the way, my "connect to network" is set to 0.1 days.

It looks like my computer cannot handle four main projects (S@H, L@H, E@H, P@H) with any project having such short deadlines, although I understand the concern. At the same time, I am a bit puzzled that L@H WUs and E@H WUs have similar processing times but L@H can afford a much longer deadline. Guess I'll have to take E@H out after the computer finishes the current E@H WU.

Sorry and thanks for your reply.

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5,206
Credit: 43,309,648,714
RAC: 44,776,587

> E@H WUs are > estimated at

Message 7901 in response to message 7900

> E@H WUs are
> estimated at about 11h while my computer usually takes about 22h to complete.

Are you running Linux? There are many reports of the linux client being particularly inefficient and slow. In my own case the time estimates have been reasonably accurate, but these are all Windoze boxes.

> By the way, my "connect to network" is set to 0.1 days.

OK, that's perfect. So at any time you only should have a maximum of two on hand, the one that is processing and the next one to process. You shouldn't actually get the second one until the first is more than half finished anyway.

> It looks like my computer cannot handle four main projects (S@H, L@H, E@H,
> P@H) with any project having such short deadlines, although I understand the
> concern. At the same time, I am a bit puzzled that L@H WUs and E@H WUs have
> similar processing times but L@H can afford a much longer deadline. Guess I'll
> have to take E@H out after the computer finishes the current E@H WU.

This is where the calculations for your work cache get screwed up. I've seen reports that say that Boinc doesn't properly take into account the fact that each project has to share the resources and new work is downloaded without properly taking into account the other work on hand. No doubt these problems will be addressed in future versions of Boinc but for the moment you might be wise to drop your most troublesome project.

> Sorry and thanks for your reply.

No worries!! When I replied to your first message I didn't realise how many projects you were trying to support and I thought you must have had too big a "connect" interval. Thanks for taking time to explain.

Cheers,
Gary.

unkx80
unkx80
Joined: 22 Jan 05
Posts: 3
Credit: 1,101
RAC: 0

> Are you running Linux?

Message 7902 in response to message 7901

> Are you running Linux? There are many reports of the linux client being
> particularly inefficient and slow. In my own case the time estimates have
> been reasonably accurate, but these are all Windoze boxes.

Windows XP.

> This is where the calculations for your work cache get screwed up. I've seen
> reports that say that Boinc doesn't properly take into account the fact that
> each project has to share the resources and new work is downloaded without
> properly taking into account the other work on hand. No doubt these problems
> will be addressed in future versions of Boinc but for the moment you might be
> wise to drop your most troublesome project.

I see. The BOINC system is sufficiently complex that whenever I run into problems, sometimes I have no idea what cound have caused it.

> No worries!! When I replied to your first message I didn't realise how many
> projects you were trying to support and I thought you must have had too big a
> "connect" interval. Thanks for taking time to explain.

I suppose I should not have supported so many projects in the first place. And thanks for all the explanation. =)

gravywavy
gravywavy
Joined: 22 Jan 05
Posts: 392
Credit: 68,962
RAC: 0

Please also see a related

Please also see a related idea, to have different cache sizes in different projects, in the thread at http://einsteinathome.org/node/188747

~~gravywavy

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.