rick@hanauma (Richard Ottolini) (09/02/88)
Why do you want to use a RPC with X? We have been using X to chain together up to three specialty computers-- a number cruncher, a data server, and a graphics front end-- without having to resort to RPCs. The MAIN is run on the number cruncher. It is spawned through an xterm on the graphics server and creates an interactive window on the graphics server. Data service is independent of X and provided by standard NFS remote mounts and procedures.
wesommer@athena.mit.edu (William Sommerfeld) (09/04/88)
In article <23412@labrea.Stanford.EDU>, rick@hanauma (Richard Ottolini) writes: > >Why do you want to use a RPC with X? Here's four cases: 1) an interactive multi-player game running on several different machines. If the state of the world can be represented in a few KB, it doesn't make sense to have the central machine responsible for all the graphics, which would result in more than a few KB in X calls. [XTrek does it wrong, but I digress..] 2) (sort of the same as 1) Imagine a situation where you use a lot of computation to generate a (relatively) small amount of data, which can be easily "expanded" into graphics by a client. Running the client on the workstation, making RPC's to a compute server, might be a more efficient use of network bandwidth. 3) The "client" is actually more than one process; the pieces need to talk to each other somehow. RPC is one way to do this.. 4) The computation is actually being done by n different machines in parallel, and the client is handing out pieces to the different servers, getting the results back, and plotting them... for example, the computation of a mandelbrot picture by partitioning it into regions, each one of which is handed to the next free member of a pool of processors (which just might happen to be idle workstations). - Bill --