Explaining How Resorce Share works

Charles A. Anspaugh
Charles A. Anspaugh
Joined: 13 Dec 05
Posts: 2
Credit: 11,608
RAC: 0
Topic 190482

I have been reading posts for days and i still dont see how the resorce share works. could only people that know for sure answer this post please. lets start simple. i am running einstien and seti, both set to resorce share 100. i would guess it would share 50/50, but einstien has finished 6 wu,seti 0 wu. could i make this easy by setting both resorce shares to 50 to represent 50% or would this not work, for example einstein resorse share 25,Lhc resorse share 25,seti resorse share 50, would equal 100%?

Stick
Stick
Joined: 24 Feb 05
Posts: 790
Credit: 4,489,287
RAC: 176

Explaining How Resorce Share works

Quote:
I have been reading posts for days and i still dont see how the resorce share works. could only people that know for sure answer this post please. lets start simple. i am running einstien and seti, both set to resorce share 100. i would guess it would share 50/50, but einstien has finished 6 wu,seti 0 wu. could i make this easy by setting both resorce shares to 50 to represent 50% or would this not work, for example einstein resorse share 25,Lhc resorse share 25,seti resorse share 50, would equal 100%?

You have probably already read this from the BOINC Wiki - but it is the best explanation of resource share I know.

However, it sounds like your issue goes a little beyond resource share and you have not given us enough information for us to help yet. First, I would also suggest that you read this regarding scheduling (again from the BOINC Wiki).

If you are still stumped after reading these references and if you are still not processing any Seti units, please include copies of your Seti related messages from under your BOINC Manager "Messages" tab in your next post.

Jord
Joined: 26 Jan 05
Posts: 2,952
Credit: 5,779,100
RAC: 26

If you have been running both

If you have been running both Einstein and Seti for some time already, but have been running more Seti lately, then you have some debt on Einstein. So now Einstein is trying to catch up on its debt, refusing you any new Seti units.

That is by design.

If not sure, try to put Einstein to No New Work for now, then Update the SAH project. You'll see it will download new work. Make sure you Allow new work for Einstein afterwards.

John Cropper
John Cropper
Joined: 10 Dec 05
Posts: 16
Credit: 55,345
RAC: 0

SETI's DB is currently

Message 22954 in response to message 22953

SETI's DB is currently 'taking another nap', probably through the New Year (at least)...guess that suggestion will have to wait... :o(


Stewie: So, is there any tread left on the tires? Or at this point would it be like throwing a hot dog down a hallway?

Fox Sunday (US) at 9PM ET/PT

Jord
Joined: 26 Jan 05
Posts: 2,952
Credit: 5,779,100
RAC: 26

It's Seti not being Year 2006

Message 22955 in response to message 22954

It's Seti not being Year 2006 compatible, John. ;)

John Cropper
John Cropper
Joined: 10 Dec 05
Posts: 16
Credit: 55,345
RAC: 0

RE: It's Seti not being

Message 22956 in response to message 22955

Quote:
It's Seti not being Year 2006 compatible, John. ;)

I suspected as much, but given the media's aversion to reporting the Y2K6 story, I didn't expect to hear much about it... ;o)


Stewie: So, is there any tread left on the tires? Or at this point would it be like throwing a hot dog down a hallway?

Fox Sunday (US) at 9PM ET/PT

Charles A. Anspaugh
Charles A. Anspaugh
Joined: 13 Dec 05
Posts: 2
Credit: 11,608
RAC: 0

RE: You have probably

Message 22957 in response to message 22952

Quote:

You have probably already read this from the BOINC Wiki - but it is the best explanation of resource share I know.

However, it sounds like your issue goes a little beyond resource share and you have not given us enough information for us to help yet. First, I would also suggest that you read this regarding scheduling (again from the BOINC Wiki).?

Thank You Stick, these links were pretty helpfull.

here is a cut and paste of what helped me on how resorce share works:

The allocation, or Resource Shares, will always add up to 100% of the available resource, which in this case is the "Idle" CPU Time.

CPU queue overload - Computer is overcommitted
Sort the Work Units by Deadline, with the earliest Deadline first. If at any point in this list, the sum of the remaining processing time is greater than 0.8 * up_frac * time to deadline, the CPU queue is overloaded. This triggers both no work fetch and the Work Scheduler into panic mode. If you don't follow that (I didn't) see example under Computer is overcommitted.

so no matter what your resorce share is set at, if you have a wu near its deadline, boinc will do that wu first.?

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

RE: so no matter what your

Message 22958 in response to message 22957

Quote:

so no matter what your resorce share is set at, if you have a wu near its deadline, boinc will do that wu first.?

close.

If looking ahead, it looks like any wu will be close to its deadline, then no matter what resource share is set boinc runs in deadline order till that wu has been crunched - we call this EDF mode, earliest deadline first

This makes sure it doesn't splat two wu in saving one. Usually ;-)

It keeps track of resource share actually applied in this process, and the projects that had more than their specified share will be held back from asking for more work till the other projects catch up.

So if you had (say) Einstein and CPDN at equal shares, and Einstein ran for two days solid to meet the deadline, then Einstein would not be allowed to ask for more work for another two days to come back to the resource share you'd set.

hope that helps
R~~

~~gravywavy

John McLeod VII
John McLeod VII
Moderator
Joined: 10 Nov 04
Posts: 547
Credit: 632,255
RAC: 0

RE: RE: You have probably

Message 22959 in response to message 22957

Quote:
Quote:

You have probably already read this from the BOINC Wiki - but it is the best explanation of resource share I know.

However, it sounds like your issue goes a little beyond resource share and you have not given us enough information for us to help yet. First, I would also suggest that you read this regarding scheduling (again from the BOINC Wiki).?

Thank You Stick, these links were pretty helpfull.

here is a cut and paste of what helped me on how resorce share works:

The allocation, or Resource Shares, will always add up to 100% of the available resource, which in this case is the "Idle" CPU Time.

CPU queue overload - Computer is overcommitted
Sort the Work Units by Deadline, with the earliest Deadline first. If at any point in this list, the sum of the remaining processing time is greater than 0.8 * up_frac * time to deadline, the CPU queue is overloaded. This triggers both no work fetch and the Work Scheduler into panic mode. If you don't follow that (I didn't) see example under Computer is overcommitted.

so no matter what your resorce share is set at, if you have a wu near its deadline, boinc will do that wu first.?

Nuts. I am going to have to fix that description as it is currently wrong. It was true at one time, but no longer.

The CPU scheduler does a Round Robin simulation with the current work, resource shares and time stats (up frac, on frac, and project Duration Correction Factors) and if any result will not complete within 90% of the time to the deadline, EDF is entered.

Comment viewing options

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