Whenever I talk with my customers on the phone, at their site, or trade shows, sooner or later the discussion turns to “CORBA overhead” and what are we doing about it? Dig a little deeper and their real concern is some facet of message latency. How many messages can their application send in some period of time? Or the time it takes to send a single message, or the order and reliability of message processing.
Many factors contribute to message latency. The usual suspects being; application architecture, third party libraries, operating systems, and network protocols. As a middleware for distributed inter-process communication, CORBA provides an abstraction layer that protects developers from many of these criminal elements. While this abstraction allows developers to build features instead of infrastructure, it also causes CORBA overhead to get blamed for their bad behavior.
Over the course of my next few posts, I’ll explore each of these shady characters in more detail. Discussing their roles, interactions, and trade-offs, hopefully helping you to identify which suspect to squeeze to improve message latency.