Folding Forum

.
Why is there no more distinction between 'Classic' (type) servers and GPU servers (WS) on this page ?
http://fah-web.stanford.edu/pybeta/serverstat.html

.

In the beginning, there was one type of WU and one type of server code. When SMP and GPU WUs were brought on-line, they needed some changes to the server code to support the unique differences in those assignments, resulting in three branches of the server code, identified as classic, GPU, and SMP. Maintaining three branches of essentially the same code is expensive so as new versions of the server code are developed, they are incorporating the ability to manage all three types of WUs, leading to a universal version of the server code that merges the three branches.

For example the passkey and bonus plan was initially developed on the SMP variation of the server code. With a long-term plan to support passkeys and QRB for GPU and classic projects, does it make more sense to add the server-based QRB features into the other two code branches or does it make more sense to incorporate the ability to manage classic and/or GPU assignments into the SMP server code? (I actually know nothing about how the plan to merge the server code branches is progressing, so this example may not reflect an actual decision that the developers needed/need to make, but it's a logical example that probably explains part of the process, even though I made it up.) The bottom line is that the distinction reported on the old Server Stats page has been a necessary but undesirable concept that has only temporarily been part of FAH.

It was nonetheless handy that the GPU WS' were marked in this way, to easily find them in the long list of servers on the stats listing.
Now you need the IP addresses to determine whether a server is a WS for the GPU client (GPU card) or not ...
Since this is almost static info and wellknown from before, I would have liked it to remain in the listing to help out donors in tracking problems or passing on information about problems and troubles.

.

I agree it was handy.

Looking at it from the perspective of a researcher in the Pande Group, suppose I have mostly SMP projects but wish to get a rapid approximation to what's going to happen when I start a new study. Shouldn't I be allowed to create a GPU project to be followed up with more detailed studies with SMP? Why should somebody have to build an entirely different server just for that one project?

Is the Client column not visible when you view that page?

I still see GPU, Classic, SMP, and PS3 labels for some servers...at least all the active servers...

bruce wrote:I agree it was handy.

Looking at it from the perspective of a researcher in the Pande Group, suppose I have mostly SMP projects but wish to get a rapid approximation to what's going to happen when I start a new study. Shouldn't I be allowed to create a GPU project to be followed up with more detailed studies with SMP? Why should somebody have to build an entirely different server just for that one project?

.
I don't understand what this has to do with changing the marking from GPU (server) to Classic (server) on a - what is in the end - text page with very useful info/data that is also useful for lots of people outside of the PG ??

.

7im wrote:Is the Client column not visible when you view that page?

I still see GPU, Classic, SMP, and PS3 labels for some servers...at least all the active servers...

.
I just checked that (linked) page again; I 'm sure there were only 'Classic' marks there when I posted this thread (a while ago).
Now I do see some servers being marked as GPU (Work) servers, amongst the Classic and SMP (Work) servers.
Don't know who changed it, but thanks anyway.

problem Resolved

.