How many servers are on each IP address?

[JX99]

I'm curious if there's just a single server on each IP address or if there are clusters of multiple servers.

The GPU2 IP of 171.64.122.74 has just been getting hammered all week, since last weekend and its collection server at 171.64.122.76 is overloaded frequently as well.

Do both these IP's listed above, represent a single server or a cluster of servers working in parallel?

I've had to suspend GPU2 folding because my 6.20 console client can't make a connection with either of these IP addresses. Applying the -send all flag to the client doesn't help. The client tries a few attempts on port 80 and 8080 on both IP's a few times, then gives up and closes.

As a suggestion, if these are single servers on each IP address, adding some more machines, to form a cluster is warranted in my opinion, unless Stanford increases the work unit size ASAP, thus reducing the overall load on each IP address.

As a side note, I have no problems uploading Windows SMP client or Windows Console client work units, only GPU2 work units.
Checking the F@H server status page frequently confirms my assessment that any server involved with Nvidia GPU2, is overloaded and additional measures of adding more equipment to handle the load is needed! Increasing the connection limit isn't working or scaling properly, so an IP cluster seems to be the next logical choice!

I'm not an expert by any means and am just offering ideas to remedy the overload situation most every Nvidia GPU2 folder is experiencing. There has to be a way to improve this situation quickly because if it continues, people are going quit the project in frustration.

Cordially, JX99

[John Naylor]

Each IP address represents an individual server, not a cluster. If the Pande Group deems it a problem they bring more servers online, but this takes time as they need to actually have a server available before they can use it (obviously).

[VijayPande]

We've been making a few server code modifications as well as bringing a new server on line. It looks like the situation for vsp07 has gotten a lot better today with those changes and it looks like those fixes can now ride us back to normalcy. As the NVIDIA GPU2 client goes from testing to production (planned for this week), we'll bring more servers on line as needed as well.