F vs. I issue rate on O1AS20-100 production run

archae86
archae86
Joined: 6 Dec 05
Posts: 3145
Credit: 7023354931
RAC: 1816661
Topic 198496

While official comments on the announcement appear to suggest that the choice to send F vs. I work on the new O1 production run is automatic, based on host capability, and it is intended that F work is dispatched to more capable hosts--possibly those with 3M or more of cache, it seems curious that there are separate user preference items for these.

It also seems odd that at the moment I write the server status page indicates 27,976 F tasks in progress, and just 15 I tasks though the ready to send is much more closely balanced (3,694 vs. 1,319).

Is it perhaps the case that some of the beta work got "rebranded" as F work, hence the big lead of F?

Is it likely that the automatic computation of host capability means the F vs. I preference a user might express for a location is ignored? or F will only be sent if both 1. the user expresses that explicit preference and 2. the host passes the capability screen?

I'm really just curious as this all seems a bit odd, and I imagine others would be happy to hear a bit more detail. Nothing here yet qualifies as obviously either a problem or a bug in my eyes.

Christian Beer
Christian Beer
Joined: 9 Feb 05
Posts: 595
Credit: 118578678
RAC: 112021

F vs. I issue rate on O1AS20-100 production run

As Bernd just wrote in Technical News section:

Quote:

This is the first time we have two "locality scheduling" applications on Einstein@Home and there are still a few oddities. Please bear with us while we are ironing out the remaining glitches.

One issue is that the ("locality") scheduler first picks a task to send to a host, then tries to find an application version to process it. If there is no suitable app version for the host the scheduler just plainly gives up instead of picking another task.

I just tweaked the scheduler code a bit, narrowing down the selection of tasks for a given host. This should fix most of the problems.

BM

There was a problem that prevented the I work from being send. This should be fixed now and I already see more in progress tasks.

As for your other questions please have some patience. I'm going to write a post that answers most of those questions later today (probably in the technical News section).

Tom*
Tom*
Joined: 9 Oct 11
Posts: 54
Credit: 274616874
RAC: 999351

Anything here that would

Anything here that would prevent getting I work for my FX-8350? I was getting
some Iwork but few and far between?
There always seems to be hundreds ready to send?

2016-03-13 17:07:15.9388 [PID=10373] Request: [USER#xxxxx] [HOST#12028732] [IP xxx.xxx.xxx.89] client 7.6.22
2016-03-13 17:07:16.0135 [PID=10373] [send] effective_ncpus 4 max_jobs_on_host_cpu 999999 max_jobs_on_host 999999
2016-03-13 17:07:16.0136 [PID=10373] [send] effective_ngpus 1 max_jobs_on_host_gpu 999999
2016-03-13 17:07:16.0136 [PID=10373] [send] Not using matchmaker scheduling; Not using EDF sim
2016-03-13 17:07:16.0136 [PID=10373] [send] CPU: req 207360.00 sec, 4.00 instances; est delay 0.00
2016-03-13 17:07:16.0136 [PID=10373] [send] ATI: req 0.00 sec, 0.00 instances; est delay 0.00
2016-03-13 17:07:16.0136 [PID=10373] [send] work_req_seconds: 207360.00 secs
2016-03-13 17:07:16.0136 [PID=10373] [send] available disk 95.19 GB, work_buf_min 8640
2016-03-13 17:07:16.0136 [PID=10373] [send] active_frac 0.999996 on_frac 0.984573 DCF 1.435204
2016-03-13 17:07:16.0144 [PID=10373] [send] [HOST#12028732] is reliable
2016-03-13 17:07:16.0145 [PID=10373] [send] set_trust: random choice for error rate 0.000010: yes
2016-03-13 17:07:16.0145 [PID=10373] [mixed] sending locality work first (0.7492)
2016-03-13 17:07:16.1084 [PID=10373] [mixed] sending non-locality work second
2016-03-13 17:07:16.1260 [PID=10373] [send] [HOST#12028732] will accept beta work. Scanning for beta work.
2016-03-13 17:07:16.1432 [PID=10373] [debug] [HOST#12028732] MSG(high) No work sent
2016-03-13 17:07:16.1432 [PID=10373] [debug] [HOST#12028732] MSG(high) No work is available for Gravitational Wave search O1 all-sky F
2016-03-13 17:07:16.1432 [PID=10373] [debug] [HOST#12028732] MSG(high) No work is available for Gravitational Wave search O1 all-sky I
2016-03-13 17:07:16.1432 [PID=10373] Sending reply to [HOST#12028732]: 0 results, delay req 60.00
2016-03-13 17:07:16.1433 [PID=10373] Scheduler ran 0.209 seconds

My Intel's have no problem getting F work.

Thanks

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109388696787
RAC: 35914341

RE: Anything here that

Quote:
Anything here that would prevent getting I work for my FX-8350? I was getting
some Iwork but few and far between?
There always seems to be hundreds ready to send?


There is a problem with the I work queue. There appear to be tasks to send but the scheduler has stopped sending them for some unknown reason. I've been informed it will be rectified first thing on Monday.

Cheers,
Gary.

Christian Beer
Christian Beer
Joined: 9 Feb 05
Posts: 595
Credit: 118578678
RAC: 112021

This should be fixed now

This should be fixed now again, sorry for the delay.

Comment viewing options

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