Strange Issue

[PY 222]

eOn Administrators,

I just realised today that all of my Linux clients has just gone down almost all at the same time.

Is something going on with the server?

Also, are you guys going to to standardize on the stats points by giving out points in terms of Gigaflops instead of each completed workunit that have different HSizes. This will help even out the stats when we get smaller or larger HSizes.

[and_ped10]

This saturday we updated the server software. This explains client crashes this weekend.

Normally it is not an issue to update the server software.

Take care
Andreas

[graeme]

By go down, I hope you mean that the clients were still running, but were not connecting with the server. Once the server came back up, the clients were, I hope, then able to connect. A long time ago, we saw linux clients terminate when the server was changed, and I hope this is not happening.

As for normalizing the different work unit sizes, I don't know an easy way to do this (counting flops for example). There is a certain amount of randomness in the work unit sizes, and the HSize sets the work for just one part of the calculation. But overall, the average computational effort per work unit will average to the same value for all clients. We only get large biases when we change the calculation, so that all work units become larger or smaller. If there was an easy was to normalize the statistics for these changes, that would be nice, but I'm also not unhappy with the simplicity of the system now.

[PY 222]

I just put a console on my Linux boxes and all clients were terminated. Only my Windows clients stayed up and reconnected and resumed processing after the server came back up.