Is everything OK since the DC move?

[N0OA]

Hi Guys,

I've noticed a fairly significant drop off in FAH points per WU. See the graph. I haven't made any significant changes in my systems (13 systems with a variety of NVIDIA GPUs in many of them). I have not change FAH versions, re-installed anything, or moved any equipment. Since the data center move, I am back to folding the same number of WU/day (roughly) - so the science is getting done. Howver, the points have fallen off significantly. I was folding about 1.58M WU/day. Not I seem to be setting in at about 1.4M WU/day. That's a pretty significant drop. I'm not doing this for the points... So that's not a concern. I just want to make sure that everything in the environment is working since the DC move.

Thoughts?



N0OA

[folding_hoomer]

Look into your own logs:
Maybe you are getting less Core17-Projects than before . . .

[7im]

Don't know what you mean by DC, but yes, fah servers are meeting WU demands. But like always the mix of work units can change at any time.

DC=Data Center

[bruce]

Look into your own logs:
Maybe you are getting less Core17-Projects than before . . .

That's a reasonable theory but it's hard for anyone to prove/disprove it except by looking at their own logs. (There are no FAH-wide statistics by project or by core published.)

GPUs running core_11/15/16 have been limited to implicit solvent models. Core_17 is more versatile since it adds the possibility of a GPU solving explicit solvent models which tend to involve heavier computing loads due to typically having a higher atom count. It also allows GPUs to solve the same models that could only be solved by CPUs (including smp/bigadv). This allows the application of the equal-pay-for-equal-work policy to be applied which improves the bonus calculations for the more modern GPUs.

[N0OA]

Hi 7im,

Sorry... You are right "DC" was intended to represent "data center". I am used to seeing modest swings in point values as the WU mix changes. But, I hadn't noticed a significant change in the WU mix recently and the drop occurred right after the outage taken to move the systems to the new data center. I hate coincidences If all is working fine with the servers, then I'll just assume it a change in the mix of WUs that I'm getting.

Thanks
<b>N0OA</b>

[Napoleon]

Still not getting any classic non-QRB core_78 work, so I guess there are some minor details to sort out. No big deal, but I for one am running some legacy HW. Wouldn't mind switching them back to folding, sooner rather than later.

[7im]

One, the data center didn't move as far as I know. They just moved some servers around.

Two, I didn't say everything was okay. As noted, some servers have not been returned to service yet, but only affects a small percentage of older hardware.

[N0OA]

Ah... I mis-read what happened on 8/26. I thought there was a change in the data centers and servers were moved to a new location. My mistake.

Since most of my servers are i7 processors and my GPUs are mostly newer NVIDIA cards (GTX580 and later), I don't think I am experiencing any issue due to the servers that support the older platforms still being out of service. Even my old GT218 [NVS 3100M] appears to be folding away again.

I'll run a sample of the WU mix prior to 8/26 and post 8/26 to see if I can find an obvious reason for the difference in point attainment. I was just posting a quick question to see if there was anything obvious that I had missed in the discussions. Once I do the analysis, I'll post the results if there is anything interesting.

Thanks for your help.

N0OA

[HaloJones]

There was a period where Core17 units were scarce so if your GPUS got a bunch of Core15 units your points drop would have been significant.