Why are they only assigning 1 of 4 sub-results of the WU?
20 Apr 2005 17:27:19 UTC
Topic 188999
(moderation:
Hey folks, I was checking my results and I see that there are a lot of work units where only one result of the 4 needed have been assigned. I can't understand why this would be - to reduce the total number of WUs outstanding it would make sense to assign 4 of the results immediately to different clients, rather than spreading them out.
I don't know if this URL will work, but try http://einsteinathome.org/account/tasks for all my results, and http://einsteinathome.org/workunit/845099 as an example of one of the WU's in question - of the four sequentially numbered results that make up that work unit only one has been issued, and it's been over 30 hours since the first was issued to me.
At present, of the 16 results I've completed, two of the WUs have been verified, 5 have all their results issued and are waiting for replies, but 9 of them only have one or two of their results issued. If server capacity is limited right now, why not send out all 4 as quickly as possible, rather than dripping them out over several days???
Why are they only assigning 1 of 4 sub-results of the WU?
)
Einstein has a few flavors of apps. Each WU needs the right one, IIRC it takes 3 days without contact by an appropiate app for the server to issue the app and offending WU to a fast machine (1 day turn around/ high speed connection). It's not a bad system, but every once in a while, something like this will happen.
Disclaimer: I'm pretty sure some of this is incorrect I saw a description of all this once (layed out nice and neat), but for me, the details are pretty fuzzy now.
> Einstein has a few flavors
)
> Einstein has a few flavors of apps. Each WU needs the right one, ...
this is true, but the rest of that post is a bit fuzzy, as the disclaimer admits.
When a machine checks in for a new WU, it gets issued with a WU from the same app as the last WU that that machine had. This is to save the download of a new app if the individual donor machine can do some work with the app it already has.
If you have more than one WU on board right now, you can tell if they use the same app if the WU names math up to the double-underscore. Almost always they will match.
If you are the first machine to be allocated WU from a new app, then there are no other machines available to take WU from that app. As other machines exhaust the apps they are running, those machines will be assigned a WU from 'your' app, and in time a fair sized cadre of machines will build up that all run that app. Once that happens assignment is as GentleGiant suggests, a given WU is assigned four times before the next one is assigned at all.
Notice too that you need more than four machines in the cadre before things run in the ideal way, as the scheduler tries to avoid assigning WU always to the same four machines.
My best guess is that you are a new donor to the project, and the scheduler took it into its head to assign WU from a brand new app. This makes sense as it optimises the amount of work you can do before downloading a new app: if you were a modem user on a paid connection you'd really value that! If you have always-on broadband I guess it makes no real difference to you, but still makes a difference to the network traffic at the project end of the link. I hope I have persuaded you that the effect you describe is a side effect of other good inetentions.
In the short term may well feel irritating, as you will wait a while for credit. When Credit does start to come it will all come quickly as other machines catch up with yours. This may even artificially boost your 'recent average credit' for a week or so...
hope that helps
~~gravywavy