[omniORB] TCP invocation strategy

youzhong liu liu@ecel.ufl.edu
Mon, 2 Aug 1999 22:02:21 -0400 (EDT)


Hi, All,

Could anyone recommend me some references about TCP invocation strategies
for different CORBAs (omniORB, Visibroker, TAO)?

In my oneway latency and throughput testing, there are some sudden
increase in latency or drop down in throughput. For example, for omniORB,
when data size is 4K, the throughput drops to a very low level; for TAO,
the latency increases while the number of calls exceeds 3000 (Request
Train). One possible reason for these phenomenen is some TCP congestion
control algorithms are invoked. I'm wondering whether I can get some
references.

Your help is appreciated!

Youzhong