You have selected to receive work from other applications if no work is available for the applications you selected

archae86
archae86
Joined: 6 Dec 05
Posts: 3,156
Credit: 7,178,584,931
RAC: 770,055

RE: There are lots of

Message 84261 in response to message 84260

Quote:
There are lots of resends at the moment.

I can give a keyhole view of how that looks on my little flotilla. Though I have a much narrower outlook than does Gary with his large fleet.

Of three hosts configured with R3/R4 capable app_info.xml setups, two have seen no resent R3 work. The third is my Q6600, which has a very big cache as I've been trying to snag sequential sequences of R4 frequencies for data on the cyclic CPU time variation matter.

I'd got no work for 5 days as I'd lowered the cache nominal time greatly to avoid a huge overfetch of R4 while the Task Duration Correction Factor was still at low R3-consistent levels. When I got my R4 bump-up, I opened the cache, and immediately started getting R3 resends from a specific frequency. Not very many, as it thought they would cost over 30 CPU hours each, but all of my download.

Wanting new, higher frequency R4 sequences for Cycle work, I started deleting one or two of these R3's every few hours, thinking I'd see the end of that stream. So far, in well over a day, it has not come, and I finally switched games, and plan just to accept them. So starting at 1187.15, I've aborted at least six of these guys, and currently have eight 1187.nn in queue, reaching up to 1187.50 so far.

Brian Silvers
Brian Silvers
Joined: 26 Aug 05
Posts: 772
Credit: 282,700
RAC: 0

RE: So starting at 1187.15,

Message 84262 in response to message 84261

Quote:
So starting at 1187.15, I've aborted at least six of these guys, and currently have eight 1187.nn in queue, reaching up to 1187.50 so far.

Feel free to abort as many R3 as you wish... My AMD will like them more than it will like R4 tasks... ;-)

Alinator
Alinator
Joined: 8 May 05
Posts: 927
Credit: 9,352,143
RAC: 0

RE: I'd got no work for 5

Message 84263 in response to message 84261

Quote:

I'd got no work for 5 days as I'd lowered the cache nominal time greatly to avoid a huge overfetch of R4 while the Task Duration Correction Factor was still at low R3-consistent levels. When I got my R4 bump-up, I opened the cache, and immediately started getting R3 resends from a specific frequency. Not very many, as it thought they would cost over 30 CPU hours each, but all of my download.

Wanting new, higher frequency R4 sequences for Cycle work, I started deleting one or two of these R3's every few hours, thinking I'd see the end of that stream. So far, in well over a day, it has not come, and I finally switched games, and plan just to accept them. So starting at 1187.15, I've aborted at least six of these guys, and currently have eight 1187.nn in queue, reaching up to 1187.50 so far.

Yep, that is one of the catches that goes with Locality Scheduling.

Once your host gets selected as a candidate for resend cleanup, you're committed to it until that particular set of datapacks is used up.

It can be kind of annoying if you want to move on to other aspects of data collecting and have continuity for that. ;-)

OTOH, it would seem to be a perfect opportunity to make sure AP is completely fixed by cutting R3 out of the mix in your case. :-)

Alinator

gaz
gaz
Joined: 11 Oct 05
Posts: 650
Credit: 1,902,306
RAC: 0

you have lost this old man

you have lost this old man with all the tec stuff will take all the R3's youcan send

cheers
garry

MarkJ
MarkJ
Joined: 28 Feb 08
Posts: 437
Credit: 139,002,861
RAC: 42

RE: There are lots of

Message 84265 in response to message 84260

Quote:
There are lots of resends at the moment. They are common but not readily available. In the last two days I would have snagged about 300 or so.

I just noticed that the Server Status page now lists R2 and R3 work units with no final result at the bottom of the page. Its saying there are zero R2's and 39,311 R3's at the moment. I think I will setup one of the machines with an R3 app_info so we can clear them, once its finished its current load of R4's (which will take about a day or so).

Bluesilvergreen
Bluesilvergreen
Joined: 20 May 06
Posts: 23
Credit: 1,206,151
RAC: 0

I never get any S5R3 wu's

I never get any S5R3 wu's since my cache got empty. I used the dual app_info.xml before my S5R3 wu's got finished. I received a few S5R4 wu's but not a single S5R3.

So, as Gary said:

Quote:
Unfortunately there is quite a trick to encouraging the scheduler into sending them. It only wants to send the work to hosts that already have the requisite large data files onboard. If a previous exchange with the scheduler has marked the data files for deletion, you won't get resends unless you "unmark" the large data files. If they've actually been physically deleted rather than just "marked", it's too late anyway. All this requires surgery on client_state.xml and I'm not about to encourage that.

There are large data files needed. But I haven't got these files anymore that belongs to S5R3 except the power apps and the necessary files for the power apps.

So is there a chance to receive S5R3 wu's or to download these large data files needed for that?
Btw. What files are that?

Is there a big cache needed to receive S5R3 ?

Alinator
Alinator
Joined: 8 May 05
Posts: 927
Credit: 9,352,143
RAC: 0

RE: I never get any S5R3

Message 84267 in response to message 84266

Quote:

I never get any S5R3 wu's since my cache got empty. I used the dual app_info.xml before my S5R3 wu's got finished. I received a few S5R4 wu's but not a single S5R3.

So is there a chance to receive S5R3 wu's or to download these large data files needed for that?
Btw. What files are that?

Is there a big cache needed to receive S5R3 ?

OK, one thing to keep in mind about Locality Scheduling is that its intended purpose is to reduce the bandwidth required on both ends for sending datapacks to the hosts. Datapacks are the files in the EAH project directory that start with h1_... or l1_....

The idea is that you select a group of hosts to run the tasks which need those datapacks, and then keep sending work for them to that group of hosts until there is no more tasks remaining.

There are some 'trip points' set that the project backend monitors to make sure that sufficient progress is made on all the datapacks, and if it detects that some seem to be 'stalling out' for any reason, it will try to find the most suitable hosts it can to add to that group of hosts to get things moving again.

So for example, say there was a replacement host needed for set of datapacks. Now lets say that you and another host come looking for work at about the same time the project has decided to bring in a relief 'batter'. Your host doesn't have any of the datapacks onboard, but the other has at least a few of them. The project would assign the task to the other host because that would require less DL'ing to get it in the game than your's would.

Carrying a larger cache might improve the odds you would get one, but I would not use that a main reason for increasing my cache, but as always, YMMV.

Alinator

Comment viewing options

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