RMI server appears to deadlock but is fine after...

Question

I have a RMI server that is providing services to clients on my network.
When I run tests from the clients (an app that constatnly makes demanding method calls to the RMI server) everything runs fine.
I have one Windows 2000 client and one Slackware Linux 10.1 client, everyone has the same version JVM. They both run the "stress test" at the same time. The Server is a Windows XP MCE on a dual core Processor, the OS has been configured to run both processors.
Both clients run for hours (all day) and so does the server without problem. Except sometimes I come back and it looks as though my clients are expererinceing some type of deadlock and when I check the server it is sitting there not doing anything as if everything is deadlocked.
Now the magic, at the server console I smack a few keys on the keyborad and everything jumps back into business humming along just fine.
What the heck? I have had this happen more than once. My apps have no input facilities whatsoever except CTRL+C which I use to kill them when I'm done with my "stress test".
I have noticed Windows clients can kinda drag everyone else down because of the way it schedules threads.
Any ideas?
Sean

Answer 1

Sometimes thread prioroties can be used effectivly on
Windows if you are having problems with your WINDOWS
machines handling the threads "properly". But not
much use really.
For testing? OK.
I have been developing and testing parallel threaded
applcations and it would not be possible without a
dual-processor machine (properly configured).
Number of processors is irrelevant. Why do you think this makes any difference at all?
You cannot test threaded applications on a single
processor. Single processor runs ONE thread at a
time. Dual-Core or dual processors can run THREADS
in parallel. I've seen it work, beautifully. your
threaded app may run fine on one processor BUT if you
run it on a dual machine it may break. In computer
science parallel computing has been around for a long
time but the majority of people (programmers) have no
idea about it whatsoever and there is a lot of
misconceptions.
I happen to have a Masters in distributed computing. I quite enjoy it. And I still insist that a dual core system is not going to do anything to your threads that a single core system won't do to your threads.
I challenge you to write a program that can tell the number of processors present, without checking the system property. If the OS is properly written and the chipset is working as it should, your program can not tell.
One thing I see repeated is that dual-core machines
have liitle impact on performance because most
applications have only a single thread. Horse
feathers!!! Windows and any other OS is going to run
applications in parallel!! I have noticed a nice
improvemnet in basic operations on a dual-core
(properly configured).
applications don't run in parallel, threads do. adding a 2nd core may make your computer more responsive if the OS preemptive scheduler is poorly written. Otherwise its no different than adding a faster CPU.
This is a wonderful subject (Parallel Computing, the
new PC)and I'm hooked (bad) mainly because this area
is almost unexplored by the masses and there is a lot
of work (read innovation) that needs done in the next
decade on the software end. If you like serious PC
check out Solaris (awsome).I love it. I was working on parallel code in Linux several years back while in school. Its not unexplored really. Search for distributed computing and you will find lots of info. The thing that is new is the attempt to take advantage automagically of multi cpu by programs that were not written to take advantage.
>
P.S. My dual-core 1gig SATA (XPS) Dell machine just
caught up with my PIII single CPU 512mb SCSI U-160
(500sc)machine in total throughput. seat-of-pants
All this swapping has to go.
Sean
Message was edited by:
DataVirtueMy last dual cpu machine was a abit BP6 dual celeron. Its a nice novelty, but buying dual/cpu dual core makes no sense unless the fastest single CPU is not fast enough or the price is right. Dual-cores has come about because the CPU folks are having difficulty making the single CPUs faster. Its a cop-out really. Instead of advancing technology, they add a 2nd CPU and try to act like they did something.
Anyone that chooses a dual-core CPU over a twice as fast single core CPU either does not understand distributed computing, or enjoys the novelty.