MISO Soup: Multiple Clients and one Server

Printable View

07-24-2007
doubleanti

MISO Soup: Multiple Clients and one Server

Code:

test

So, in order to speed my research I decided to build a four or five node ethernet 'cluster', and run my software on each. Naturally I've given in and decided to reprogram it using networking code to distribute the processing. The idea of having four 2 GHz boxes parsing English makes my mouth water, doesn't it yours?

Anyway, long story short I've been trying to get a single server / multiple client setup where the clients are processing nodes and the server just sends requests and writes results to a centralized place. I have a global variable for sent/received bytes total on the server and client sides. The problem is when I have more than one client connecting to the server, the total of received and sent bytes on the server is lower than the total fo that sent on the clients. I have a feeling this has something to do with mutual exclusion, and I tried isolating the send/received requests on the server-side, but that didn't seem to work. By the way, this is in Java. So anyway I hope someone can shed some light, Bubba? Sunlight? Kermi? Thanks!

=-{da}-=

Code:

// there ain't no source code it's in the attached file! nice new feature to enforce code tags guys... nice...
07-24-2007
doubleanti

Code:

Heres the code attached.
07-24-2007
doubleanti

Okay, that was stupid I didn't make the mutual exclusion flags on the server static. Duh. So it fixed the counting problem. Aside from that I do have a question. =) Right now my program on a single box runs for about six days. With four boxes, ideally it'd take a day and a half, or maybe two days. However the results are huge, some odd 30 GB of data, which I will have to transfer back over the Ethernet. With a fast switch, it's no problem, however I can just use the windows networking and manually split the job, get the same speed gain and not worry about extra coding. But if I do continue, it would be great to be able to run a server, sit down at anyone's computer, and use their processor performance to aid my research (or, case in point, any of the free on campus computers I happen to sit by). What do you think? Okay it's late good night! Glad to be pulling all nighters again!