New Transport for omniORB

[Luis Sandoval]

Hello.

The High-performance Computing and Simulation (HCS) Research Laboratory at
the University of Florida, works on the development of light-weight
communication interfaces for high-performance networks as part of the
SCALE (Scalable Cluster Architectue Latency-hiding Environment) project.
One task that is currently underway is the substitution of TCP/IP, as
omniORB's transport, with our communication interface (called SCALE
Channels). The idea is to exploit high-performance network performance to
the maximum by removing the TCP/IP overhead; this will enable us to
effectively run CORBA over high-speed networks such as SCI, Myrinet,
HiPPI, ATM, etc.

We took the first step by adapting omnithread to our low-latency thread
library (SCALE Threads) obtaining a 40 microsecond improvement in the two
way latency. Now we are ready to take the next step, but we will
appreciate any help, as to what are the key pieces of code that we need to
attack. We have knowledge of the key abstract classes that we need to work
on: ropeFactoryTypes, ropeFactory, incomingRopeFactory,
outgoingRopeFactory, Strand, Rope, and Endpoint. But, any further
indications will be appreciated (modifications to the GIOP client and
server, etc.)

And, off course the final result of our research will be freely available
the same way omniORB is available at the present time.

Thanks in advance,
Luis

PS: for more information on the SCALE research project check our web page
at http://www.hcs.ufl.edu

===================================================================
Luis E. Sandoval
Graduate Research Assistant
High-performance Computing and Simulation (HCS) Research Laboratory
http://www.hcs.ufl.edu/~sandoval	email:sandoval@hcs.ufl.edu 
===================================================================