Implementation advice

[Urbanfreestyle]

Good morning,

So i have just put forward a business case to my boss to use the work servers for F@H.
I am anticipating some questions in return so thought i would seek some advice from you all. The servers are spread out across a few domains and different building locations. none (or very few) are running GPUs so not expecting MASSES of ppd but they all run dual quad or better xeon processors.


1. Is it possible to schedule f@h to run out of hours? i know you can set folding low, med or high but how would that affect the load on the servers if they are being used for other things?
2. How is it best to monitor a lot of machines running f@h? is it possible to use a centralised management console to check the progress / load of each client?
3. Can clients be remotely stopped if there is an issue with the rest of the server and we need to isolate the issues?
4. I hear network traffic is pretty low but some machines are in remote locations so would these have any issues?
How do the clients cope in a Hyper-v environment? i guess they just use the resources available to them so we could dedicate 4 or more cores to the VM and it will use those 4?

Hopefully someone can help me out with this as there are approx 150+ servers sat here not doing much.

[JimboPalmer]

1) You would need to use some other scheduler, it is not a feature of F@H. F@H tries to chose the lowest priority possible so it does not pre-empt other tasks. It uses very little RAM or Disk bandwidth. On a workstation, you can choose On Idle, but that will not function as expected on a server, as it is just looking for the screen saver as an idle indicator.
2) there are third party monitoring applications http://www.hardwarecanucks.com/forum/ha ... post143366

[Urbanfreestyle]

Thanks for the reply, Hopefully they will embrace this idea....

[bruce]

A centralized management console program is included with the distribution package: FAHControl. Not only is it able to monitor remote clients, it can stop/start them as needed to isolate issues. I'm not sure if it has a limit to the number of clients it can monitor.

When run on CPUs, FAH is specifically designed to use only "idle resources" By dint of it running at the lowest possible priority, the OS should schedule all other activities before dispatching work for the CPU(s). [This is not true for GPU processing, but that's not going to be an issue for you.)

We recommend that you get written confirmation of your boss's decision.

[Urbanfreestyle]

well it is going to staff forum, then to big bosses, will get it in writing and i have written an implimentation plan (by that i mean how we will roll it out and run on a few test servers first)

[Joe_H]

For additional controls and management, the folding client has a supported connection for third party use through the local telnet port. Some have used that for scripts that paused and resumed folding at specific times or for when they were playing a game. So for example, it would be possible to create a script that caused folding to pause or finish processing at a certain time, and then later resume folding later in the day.