1) Message boards : Number crunching : Because you asked.... (Message 16697)
Posted 13 Apr 2007 by Profile Michael Roberts
Post:
I've had LHC permanently active along with half a dozen or so other projects for the last few years and I have never had any scheduling problems. The only long-term concession I have made to the intermittent work from LHC is to make its priority about 10 times that of the other projects. If the work becomes more regular I'll reduce it again.
2) Message boards : Number crunching : New computer database entry created on each connect (Message 16687)
Posted 9 Apr 2007 by Profile Michael Roberts
Post:
Is there a strong reason why, when merging hosts, they are merged to the highest number in the set instead of the lowest? I imagine that most of the time, particularly in this case when nobody ordered the higher numbers, retaining the lowest would be more user-friendly particularly if the earliest creation date were also retained.
3) Message boards : Number crunching : Faster than a speeding hadron..... (Message 14056)
Posted 19 Jun 2006 by Profile Michael Roberts
Post:
...some jerks dont care what the PROJECT is trying to do...

I would like to explain to bowlingguy300 why I am not "a jerk" as he appears to think (at least, I am not for the reason he proffers...)

Assuming that the aim is to finish each run in minimum time, the intention of spreading work evenly among contributors by taking as little work as possible as an individual is in fact a pessimal strategy, since it will tend to ensure that large caches can be filled late, thus finishing late and delaying the completion of each run to the maximum extent.

A certain proportion of large caches is a normal part of the system. Those who have only intermittent access to the internet invest a great deal of enthusiasm, organisational ability and determination in the project and I salute their contribution.

Since we are a large community of independent personalities, we will have different strategies. Some of us will understand that the work will be finished sooner if as many machines as possible are working in parallel, and every machine will be occupied for roughly the same time. Since we know that some machines quite properly take the maximum determined by the deadlines, the optimal strategy for the project as a whole is for everyone to do the same. So that is what I am doing. It happens that the optimal strategy maximises both the benefit to the project and the contribution of the individual. I have no problem with that.

Naturally not everyone will follow a single strategy - we know that machines will run with all sorts of cache sizes and project shares - and even if they did, there will be a background of repeated workunits - erroneous results, machines break down, electricity bills arrive, whatever... meaning that even if everyone follows a correct optimal strategy there will be a certain number of workunits in progress for the duration of the deadline after most units have finished. So either way, trying to make everybody do the same is pretty futile.

From the point of view of running a cost-effective server it may even be desirable for the clients to be following a broad mix of strategies since that will tend to avoid peaks.

As many others have already pointed out, the project engineers understand this perfectly well and will time each run and the deadlines they specify according to their experience to fit in with their other work - and, I imagine, their holidays.

So please - relax, be happy, stop insulting people, and request as much work from the project as you want to within the constraints defined by the project team.
4) Message boards : Number crunching : Units not validating (Message 12869)
Posted 25 Feb 2006 by Profile Michael Roberts
Post:
One point about your workunit, Ray, is the cpu times for "successful" results - ranging from 1.95s to over 19000s. It is difficult to suppose that all these results can be consistent with each other. Here is a similar one that I have just noticed



©2024 CERN