Is there any way to overide resource share settings on the host?

The 3 setting options of resource shares on each project site which does not offer much flexibility for multiple hosts.

The only solution I can currently see is to make myself into a team and run my hosts as team members, but of course that will divide up my user stats.

Any comments would be appreciated.

Bruce

Is there any way to overide resource share settings on the host?

The 3 setting options of resource shares on each project site which does not offer much flexibility for multiple hosts.

The only solution I can currently see is to make myself into a team and run my hosts as team members, but of course that will divide up my user stats.

Any comments would be appreciated.

Bruce

I don't have a solution for you, but just a warning. Resource share seems to have the lowest importance to the boinc scheduling algorithm. One would think it should be the highest, since it represents the volunteer's desires, but that is not the developer's thinking.

In my experience, I have clients that for one reason or another could not eradicate their long term debts between two projects. (These reasons had to do with the projected time the wu's required are very dissimilar between projects, the due dates of the wu's, the frequency of connect, and the desired cache settings.)

I think I have beat the problem a bit by setting the resource share ratio between my two projects to be 999:1, and keeping a two day cache size. After about 9 months, the LDT now seems to be reversing a bit, towards the ideal 0; but there yet is a long way to go. Now, when seti goes down, like it did for the last couple of weeks, I get a lot of my secondary project, but not otherwise.

I have also refused to be a guinea pig and waste cpu cycles on AP; that may affect your project ratios as well, given that the run time estimates for AP are bogus (when I last checked). So I've turned off AP downloads until the smoke clears in a couple of months. Maybe I'll start running those wu's at xmas time.

I simply Detach, I too wish there were another way, But You'd have to get the Boinc Devs to do work there on this idea.

---

The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's

It would be nice if you could allocate the number of CPU’s to use for each individual project and each computer you have.

I don’t like it when I’m “only” running my secondary projects if I have work available in my primary project.

Like if you have a quad you could say in “preferences” only use 1 CPU on Einstein and 3 on Seti or 2 Seti, 1 Einstein, 1 Rosetta. A person may also want to run many projects and see results from each of these projects on a daily basis and allocate 1 core to each project. The way it is now you change the number of CPU’s used in 1 project, it changes in all projects.

The number of cores in a computer keeps going up every year. It might be time for a change in the way you can allocate your CPU usage throughout projects.

Ga Day

It would be nice if you could allocate the number of CPU’s to use for each individual project and each computer you have.

I don’t like it when I’m “only” running my secondary projects if I have work available in my primary project.

Like if you have a quad you could say in “preferences” only use 1 CPU on Einstein and 3 on Seti or 2 Seti, 1 Einstein, 1 Rosetta. A person may also want to run many projects and see results from each of these projects on a daily basis and allocate 1 core to each project. The way it is now you change the number of CPU’s used in 1 project, it changes in all projects.

The number of cores in a computer keeps going up every year. It might be time for a change in the way you can allocate your CPU usage throughout projects.

Ga Day

Here is the thing: if you are only running your secondary projects, it means one of two things:

1) You have work from the secondary project, and BOINC needs to get it done and back on time.

2) You've "overcrunched" your primary project, and the secondaries are being run to meet your resource share.

Resource share is not honored on an hour-by-hour basis or even a day-by-day basis. What it really does is control how work is requested.

Once it is downloaded, the scheduler concentrates on getting it all done without missing deadlines (and tracking debt, of course).

---

I don't have a solution for you, but just a warning. Resource share seems to have the lowest importance to the boinc scheduling algorithm. One would think it should be the highest, since it represents the volunteer's desires, but that is not the developer's thinking.

I'm just curious: should BOINC miss deadlines in order to strictly adhere to resource shares?

In my experience, I have clients that for one reason or another could not eradicate their long term debts between two projects. (These reasons had to do with the projected time the wu's required are very dissimilar between projects, the due dates of the wu's, the frequency of connect, and the desired cache settings.)

... and I'm sure the developers would like to know more about this, since it should not be possible.

I think I have beat the problem a bit by setting the resource share ratio between my two projects to be 999:1, and keeping a two day cache size. After about 9 months, the LDT now seems to be reversing a bit, towards the ideal 0; but there yet is a long way to go. Now, when seti goes down, like it did for the last couple of weeks, I get a lot of my secondary project, but not otherwise.

Which is what should happen -- slowly.

I have also refused to be a guinea pig and waste cpu cycles on AP; that may affect your project ratios as well, given that the run time estimates for AP are bogus (when I last checked). So I've turned off AP downloads until the smoke clears in a couple of months. Maybe I'll start running those wu's at xmas time.

Meaning of course that none of the beta testers did anything meaningful, in your opinion.

---

There is or was a utility you can use to view and or erase your BOINC debt. It's called BOINC DV. I used it once a while back when I couldn't get any new work from SETI because of my debt ratio being so out of balance. I may still have it on disk here somewhere. I'll look around and see if I can find it if you're interested.

---

There is or was a utility you can use to view and or erase your BOINC debt. It's called BOINC DV. I used it once a while back when I couldn't get any new work from SETI because of my debt ratio being so out of balance. I may still have it on disk here somewhere. I'll look around and see if I can find it if you're interested.

Found it for you: http://www.skipsjunk.net/info/boincdv.html

---

There is or was a utility you can use to view and or erase your BOINC debt. It's called BOINC DV. I used it once a while back when I couldn't get any new work from SETI because of my debt ratio being so out of balance. I may still have it on disk here somewhere. I'll look around and see if I can find it if you're interested.

Found it for you: http://www.skipsjunk.net/info/boincdv.html

I would like to point out that use of this utility intentionally breaks your resource share.

BOINC will eventually meet your resource share in almost all cases, but it uses the debts to do so. Resetting the debts to 0 breaks the ability or BOINC to do this.

---

BOINC WIKI

There is or was a utility you can use to view and or erase your BOINC debt. It's called BOINC DV. I used it once a while back when I couldn't get any new work from SETI because of my debt ratio being so out of balance. I may still have it on disk here somewhere. I'll look around and see if I can find it if you're interested.

Found it for you: http://www.skipsjunk.net/info/boincdv.html

I would like to point out that use of this utility intentionally breaks your resource share.

BOINC will eventually meet your resource share in almost all cases, but it uses the debts to do so. Resetting the debts to 0 breaks the ability or BOINC to do this.

So does detaching, So what? Besides I'd rather not use DV, But to each their own brand of poison. ;)

---

The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's

[/quote]Here is the thing: if [/quote]

I hear what your saying Ned…

I’m just say’n….If I have 8 cores on a computer and want to run 1 core on 8 projects, I should be able to.

**((Only thing that may not make this possible is if for some reason this may be harmful to your computer))**

I’m willing to bet this will be implemented in the future….and the future is NOW.

I hear what your saying Ned…

I’m just say’n….If I have 8 cores on a computer and want to run 1 core on 8 projects, I should be able to.

**((Only thing that may not make this possible is if for some reason this may be harmful to your computer))**

I’m willing to bet this will be implemented in the future….and the future is NOW.

Let me cut off the following arguments before they are stated so that I can better understand the request:

It is my computer.
I am donating my resources.
I am paying for my electricity.
I should be able to specify what runs on my computer and how it runs.
Any other argument that shows ownership of the system.

I fully understand who owns and runs the computer. But why should it matter to anyone which project is running as long as BOINC is adhering to your resource shares setting and each project is getting the processor time you've specified?

I don't understand why people want to make the system more complex, which means more option for novices to not understand, which also means harder to troubleshoot and diagnose if a user is having a problem. What ever happened to people wanting BOINC to be simple and easy to use? The more options you throw at it, the more that a user might be intimidated on how to configure BOINC for their system.


You install BOINC. You join a project(s). You set your resource shares. BOINC managers the CPU time for you and gives CPU time as per your preferences and tries to meet all deadlines. Why make it more complex and make the developers have to rewrite the entire scheduler routine simply because they want to be able to specify down to the CPU level based upon project? Wouldn't the developer's time be better spent making BOINC easier to use and more stable for everyone?

Here is the thing: if [/quote]

I hear what your saying Ned…

I’m just say’n….If I have 8 cores on a computer and want to run 1 core on 8 projects, I should be able to.

**((Only thing that may not make this possible is if for some reason this may be harmful to your computer))**

I’m willing to bet this will be implemented in the future….and the future is NOW.
[/quote]
Nope. Meeting deadlines is the most import function of the CPU scheduler. If the CPU scheduler needs to use all the allowed cores for a single project in order to meet deadlines it will.

---

BOINC WIKI

Here is the thing: if

I hear what your saying Ned…

I’m just say’n….If I have 8 cores on a computer and want to run 1 core on 8 projects, I should be able to.

**((Only thing that may not make this possible is if for some reason this may be harmful to your computer))**

I’m willing to bet this will be implemented in the future….and the future is NOW.

I think it would only happen if a volunteer is willing to actually write the code, and it doesn't impact BOINC overall.

But let's say that you have four cores (so I don't have to pick four projects) and you decide 2 cores for SETI, one for Einstein, and one for CPDN.

What happens if SETI (or Einstein) is down for a few days? Do you let the "Einstein" core do SETI, or does it go idle?

Seems that what you're asking is something of a special case: where you have a nice clean 1:1 ratio between cores and projects.

What you're asking looks cool because you look, and you see two SETI, one Einstein, and one CPDN, but what we have does the same thing if you set SETI at 50, and Einstein and CPDN to 25, but it holds those ratios across outages.

It doesn't work if you want to run five projects on four cores, or if you want to allocate 900 to one project, and 50 to each of the other two.

---

There is or was a utility you can use to view and or erase your BOINC debt. It's called BOINC DV. I used it once a while back when I couldn't get any new work from SETI because of my debt ratio being so out of balance. I may still have it on disk here somewhere. I'll look around and see if I can find it if you're interested.

Found it for you: http://www.skipsjunk.net/info/boincdv.html

I would like to point out that use of this utility intentionally breaks your resource share.

BOINC will eventually meet your resource share in almost all cases, but it uses the debts to do so. Resetting the debts to 0 breaks the ability or BOINC to do this.

What the person who uses BOINCDV or resets their debt values is saying is "I don't really care that much about adhering to the resource shares."

More than that, it probably means they need to rethink those settings.

---

I know why yer against my suggestion Ozz…….I get PM’s from people too.

Deadline has nothing to do with it, in what I’m talking about John.

It can, and would be very simple to implement a per core, per project CPU allocation, and I KNOW it will be implemented in the future with the increase in the number of cores a computer has.

I’ll give you an example: I don’t want to run Astropulse right now, cause I know my computer will download 20 of them and I’ll be drowning in them for 1 - 2 weeks.

If I could allocate 1 core for Astropulse and 3 for SETI I might do it.

Let ME select how my computer HELPS the cause of various projects…1 core at a time or 8.

Meaning of course that none of the beta testers did anything meaningful, in your opinion.

No Ned, not we think at all. But as any production engineer can tell you the final test, call it the "Gamma" test if you like, is when the product is released.This applies particularly to software. It is not until the product is released that everything is tested till it breaks in all possible circumstances.

With software, the almost infinite varieation in hardware out in the "real world" makes it impossible to Beta test every possible configuration and, it's only after the release is made that you come up against the most acid test of all, the general end user's expectations.

Brodo

I know why yer against my suggestion Ozz…….I get PM’s from people too.

Deadline has nothing to do with it, in what I’m talking about John.

It can, and would be very simple to implement a per core, per project CPU allocation, and I KNOW it will be implemented in the future with the increase in the number of cores a computer has.

I’ll give you an example: I don’t want to run Astropulse right now, cause I know my computer will download 20 of them and I’ll be drowning in them for 1 - 2 weeks.

If I could allocate 1 core for Astropulse and 3 for SETI I might do it.

Let ME select how my computer HELPS the cause of various projects…1 core at a time or 8.

You don't know anything of the sort.

The most important function of the CPU scheduler is to meet deadlines - even if that means temporarily ignoring all other settings about how to divide the resources of the computer.

---

BOINC WIKI

The most important function of the CPU scheduler is to meet deadlines - even if that means temporarily ignoring all other settings about how to divide the resources of the computer.

If you only have 1 core allocated (in preferences), then the deadline should be met. That is what you have allocated…1 core (if this was possible today). The deadline would be based on knowing your only going to use 1 core, therefore it would not download work that 2,4, or 8 cores would need to process.

What could be more difficult than resource share. I see people that post on these boards all the time that are confused. If it was all based on 100% it might not be that bad. But when you can say 500% for project "A" and 50% for project "B" and 8% for project "C" it all can get difficult for the average person that is not a daily participant. How would you know that you can go above 100% in resource share if you where not a regular reader of these boards?

I don't have a solution for you, but just a warning. Resource share seems to have the lowest importance to the boinc scheduling algorithm. One would think it should be the highest, since it represents the volunteer's desires, but that is not the developer's thinking.

I'm just curious: should BOINC miss deadlines in order to strictly adhere to resource shares?

Boinc should not miss deadlines, but should not be too agressive in taking on new work. My experience is that it fills the cache when the LTD is grossly negative already if the 'other' project cannot provide wu's.

In my experience, I have clients that for one reason or another could not eradicate their long term debts between two projects. (These reasons had to do with the projected time the wu's required are very dissimilar between projects, the due dates of the wu's, the frequency of connect, and the desired cache settings.)

... and I'm sure the developers would like to know more about this, since it should not be possible.

Some developer types have explained the issue to me. In my case, with a large cache, Einstein was slipping in wu's eventhough the LTD was grossly negative because their deadlines were shorter than Seti's. The user is the victum in this case and all I could do is reduce the cache size and pervert the resource allocation to resist the boinc algorithm.

I think I have beat the problem a bit by setting the resource share ratio between my two projects to be 999:1, and keeping a two day cache size. After about 9 months, the LDT now seems to be reversing a bit, towards the ideal 0; but there yet is a long way to go. Now, when seti goes down, like it did for the last couple of weeks, I get a lot of my secondary project, but not otherwise.

Which is what should happen -- slowly.

I have also refused to be a guinea pig and waste cpu cycles on AP; that may affect your project ratios as well, given that the run time estimates for AP are bogus (when I last checked). So I've turned off AP downloads until the smoke clears in a couple of months. Maybe I'll start running those wu's at xmas time.

Meaning of course that none of the beta testers did anything meaningful, in your opinion.

Well, I didn't say that the beta testers's efforts were not meaningful, but if you think so I can't argue with you. A user like me isn't privy to the science/programming issues that arose implementing and testing of AP. All we see is that the RAC's are going down, server network/storage chaos when the project was abruptly turned on, and knowledgable people's postings that question whether the actual algorithm is correct yet. So, I think I'm justified in the guinea pig allusion and am very happy to wait awhile while the beta folks debug the production system. As long as there is MB work to do, I doubt there is any harm done to the overall project.