Hi guys and girls, I'm a member of the cpc/bit-tech folding team and I'm trying to get my college to start using F@H on the campus network. Initial responses are good but I need to make a presentation to the head of the network to convince him that it is a worthy cause and that he should allow it on the network. Here's the rub - he needs to know what kind of impact running F@H (running cpu clients @ 25%) on 500+ pc's is going to have on the campus bandwidth. If anyone can point me in the right direction or some one in the know could give me a good estimate it would be a massive help! Thanks guys! Ev1lm1nd666
What client will it be running? What cpu's will it be running on? I remember ps3/landi used to run 300+ machines on his company network and used less than 40GB a month
when you say 25% how do you mean (cpu power man on the systems would not ramp up the cpu speed if the client was set to 25% and you mite not make dead lines) below is how i would do it to prevent slow down on each of the systems the system specs would be useful (CPU only is most likely case that be used for F@H Clients) i would suggest 1 client per dual core (90% cpu limit) and 2-3 per quad core and limit it to 90% per client (depends on what the systems are used for) the 90% limit is there so it F@H responds faster when both or more CPU cores are needed (F@h client backs off faster then it would if set to 100%) the single F@h clients when i say 90% that is for 1 cpu core (so it look like in task manager that be about 45% for dual core and 22% per client for an quad) need to look into how good the cooling inside the pcs are as well options to set small packets, no adv work units and set mem use to 128mb (server should only give small work units out any way)
The machines will be using the standard cpu client with some using the GPU 2 client too. Most of the machines are standard workstation kit, 3Ghz Pentium 4's 2Gb 800Mhz Ram no graphics cards. Some are CAD machines and have low end Tesla cards and tripple core AMD processors. The network admin will be more bothered with the bandwidth usage, cost and usability of the pc's while the clients are running. Thanks for your help guys, this is very much appreciated
What switches / core configuration do you have on the campus network, it should possible to seperate/tag the packs from the FAH client and set them as a lower priority, however this is depends on the hardware used - this would fix the worry about bandwidth during the working day. Thoughts about only running during the evenings (or offpeak) using scheduled tasks, which could be managed via Group Policy or such?
Sorry another thought, assuming that you use a proxy, you should be able to set the priority of the traffic it sends at an application level (however this only effects external bandwidth and not internal traffic) If you can do it via the proxy this would be by far the easiest way to manage it.
the college won't give me any information on the physical infastructure in the server/ routing room, I know there are several rack mount cabinets full to the brim with everything from switches to fibre optic routers. We have a 1gig backbone and I think we have a 16meg connection to the net.
Even with 500+ machines running the classic client, it is very unlikely that more than two or three will be up/downloading simultaneously (unless some event starts all the clients at the same time). Why not just install the client on one machine, and measure the traffic?
