Because of Einsteins short deadline and the way they decided to do scheduling. I am not going to be able to run any Einstein units for a long time. The scheduler forced Einstein to run even though the einstein units were not in danger of running pass deadline. Einstein seems to have an exhorbantly high LTD and won't get to run for a long time.
Copyright © 2024 Einstein@Home. All rights reserved.
Can't run Einstein due to new scheduler
)
Because of Einsteins short deadline and the way they decided to do scheduling. I am not going to be able to run any Einstein units for a long time. The scheduler forced Einstein to run even though the einstein units were not in danger of running pass deadline. Einstein seems to have an exhorbantly high LTD and won't get to run for a long time.
I am assuming that you mean negative LT debt. First the debt is in seconds, so it looks rather large when it isn't really.
There are three things that cause the CPU to enter process first mode.
1) A result due within 24 hours.
2) A result due within 2* your connect time (queue size).
3) Order the results by deadline and see if by processing them in deadline order all results will get done within 80% of the deadline. If not, enter earliest first.
There are also three things that will prevent download of work from anywhere.
1) See #3 above.
2) More than a max number of projects - default is 5.
3) Add the processing time left / time to deadline for all results. If the total is greater than 0.8 don't download work.
Work will not be downloaded from a project if its LT debt is negative (it has gotten a boost of CPU time from process earliest first recently).
Work will be downloaded from projects with negative debt if there is a CPU idle.
Work should be downloaded from positive LT debt projects if the queue is not filled.
BOINC WIKI
2) More than a max number of
)
2) More than a max number of projects - default is 5.
Why can't I have more than 5 projects?
Why can't I have more than 5
)
Why can't I have more than 5 projects?
Add a line:
(max_projects_on_client>nn(/max_projects_on_client> in global_prefs.xml in your BOINC directory.
[edit] have to learn how to make the greater than sign appear in messages [/edit]
Happy crunching,
Why can't I have more than 5
)
Why can't I have more than 5 projects?
Add a line:
(max_projects_on_client>nn(/max_projects_on_client> in global_prefs.xml in your BOINC directory.
[edit] have to learn how to make the greater than sign appear in messages [/edit]
Happy crunching,
This: > is: >
This: < is: <
Thank you Paul. You really
)
Thank you Paul. You really are almost infinite source of advice.
Happy crunching,
You can attach to as many as
)
You can attach to as many as you like, but the default is to have only a limited number on the host at once. You can also set this to a lower value and get the work returned for each project faster (I would not suggest less than 2 as that may cause the scheduler to run queueless). I wanted this as a web UI configuration setting.
BOINC WIKI
Because of Einsteins short
)
Because of Einsteins short deadline and the way they decided to do scheduling. I am not going to be able to run any Einstein units for a long time. The scheduler forced Einstein to run even though the einstein units were not in danger of running pass deadline. Einstein seems to have an exhorbantly high LTD and won't get to run for a long time.
I have to agree with this statement. It seems as if the new 4.43 CC favors Wu that are in "danger" of missing there deadlines, by a VERY VERY long shot.
The following apperared after switching to 4.43
23-05-2005 20:21:07||New CPU scheduler policy: earliest deadline first.
Though the 2 e@h wu's that were now selected were due on the 29 and 30 of this month
Have 2 day cache, cause this sounds reasonable to me, also have a 90/10 share with s@h/e@h
Dont like this new scheduler at all.
So I'm sorry to say, that I have now switched e@h off, meaning, that I have clicked the "dont allow new work" for this project.
Will be doing only s@h and p@h wu, cause they have same report deadline, which in my opinion means that they would work much better better together as a combo.
No cp hogging with those 2, just a normal sharing :)
So perhaps team, increase the report deadlines, and it might work better with the official Berkely boinc scheduler, or perhaps you might loose searchers.
Gezz had my say now
See you around guys
Because of Einsteins short
)
Because of Einsteins short deadline and the way they decided to do scheduling. I am not going to be able to run any Einstein units for a long time. The scheduler forced Einstein to run even though the einstein units were not in danger of running pass deadline. Einstein seems to have an exhorbantly high LTD and won't get to run for a long time.
I have to agree with this statement. It seems as if the new 4.43 CC favors Wu that are in "danger" of missing there deadlines, by a VERY VERY long shot.
The following apperared after switching to 4.43
23-05-2005 20:21:07||New CPU scheduler policy: earliest deadline first.
Though the 2 e@h wu's that were now selected were due on the 29 and 30 of this month
Have 2 day cache, cause this sounds reasonable to me, also have a 90/10 share with s@h/e@h
Dont like this new scheduler at all.
So I'm sorry to say, that I have now switched e@h off, meaning, that I have clicked the "dont allow new work" for this project.
Will be doing only s@h and p@h wu, cause they have same report deadline, which in my opinion means that they would work much better better together as a combo.
No cp hogging with those 2, just a normal sharing :)
So perhaps team, increase the report deadlines, and it might work better with the official Berkely boinc scheduler, or perhaps you might loose searchers.
Gezz had my say now
See you around guys
You have to give it some time. After it is done with the short deadline WUs, it will not download any work from the project for a while. The resource sharing is over a longer time than a single day.
BOINC WIKI