Summary

In this chapter, we've looked at what issues will affect the process view of the architecture document.

When we use a client-server design, with its centralized data source and business objects to act on the data stored there, we quickly need to find a happy medium between the problem of data corruption by server threads, and unacceptably slow and unscalable server performance.

Specifically we've seen:

• How multi-tasking operating systems introduce multiple processes, and threads within a process, but can also interrupt program execution at any time. The application developer is responsible for making programs thread-safe
  
  
• The need for atomic operations and synchronization to protect shared resources
  
  
• General locking strategies, and those used in the MFC
  
  
• How to avoid deadlock
  
  
• What it takes to be a transaction — ensuring reliability of operations, especially when execution is distributed across multiple objects, machines and datastores
  
  
• The ACID test
  
  
• What help TP systems provide
  
  
• The use of two-phase commit to control diverse resource managers
  
  
• Message queuing and queued transactions
  
  
• Object pooling for scalability
  
  

© 1998 by Wrox Press. All rights reserved.