Is it common in the Unix/Linux environment to install compute intensive applications on a Server system and have the client machines download the executables into memory at runtime to run locally? This model seems taxing to the network, and as I understand, has been largely abandoned in the Windows environment in favor of "lite" client side apps or web interfaces which allow the majority of actual computing to stay local on the server. I am curious whether there is a substantial difference in philosophy on the Unix platform, because I am seeing a fairly experienced Unix administrator set up his office that way.

Thanks.

I don't understand the environment you're asking about. What does "client machines download the executables into memory at runtime to run locally" mean? What kind of executables? The only protocol I can think of that more or less matches your description would be running a program on an NFS mounted filesystem. I would avoid that if I could...disks are cheap and each client should it's own copy of frequently used executables. "compute intensive applications" are virtually unheard of...since the arrival of systems with multiple super-scalar risc cpu's. If I had a compute intensive application, I perhaps might try to distribute the compute tasks among several systems. Are you arguing for concentrating a lot of compute intensive tasks on a single system? How many systems? What bandwith is the network? What else is the network doing? I would need a lot more info here.