HEALTH CARE

High volume applications are an exercise in redundancy.  In particular, how can the systems be put together to minimize an outage of the critical application.  When there are 4 million transactions flowing per day, even a two minute outage (or 10,000 transactions) can be painful.  This project was such an exercise.  It not only required a high degree of redundancy but also a high degree of automated health checks that would reconfigure data paths on the fly depending on the results of the health check.

Backstory

• The transaction processing landscape had changed significantly over the last ten years:
  • Transaction volume grew to over 4 million transactions per day.
  • Server and mainframes technology changed significantly.
• The challenge was to advance the core applications while at the same time advancing the network and hardware infrastructures without disrupting the transaction processing environment:
  • Back-end mainframes were replaced.
  • Front-end terminal server technology was replaced.
  • Old TCP enterprise bus technology was replaced.
• Applications residing on the above technologies just kept growing both in terms of volume and complexity.
• The philosophy was to make small but continued upgrades to all aspects of the computing environment with the central theme of not disrupting the transaction processing.

Business Objective

• Implement a fault-tolerant messaging service that interfaced the legacy front-end server technology to the new back-end transaction processing technology

Functional Requirements

• Connect 3,000 TCP sessions to back-end IBM server
• Support sustained 300 transactions per second
• Support bursts to 500 transactions per second
• Ensure ultra-high availability

• Convert all messaging to common format
• Convert Linux TCP messaging to IBM MQSeries® messaging
• Convert WebSphere® Message Broker messages to an internal proprietary message format
• Correlate request and response messages
• Implement transaction security auditing
• Provide continuous end-to-end data path health check
• Load balance transactions based on data path health check

Outcome

By the development of a number of Filters, the Plexus Message Broker was enhanced to convert messages between the Linux systems and the IBM system.  Two instances of the Plexus Message Broker were installed on each of three servers to provide the required redundancy and resiliency. The software supports the average message flow of 200 messages per second with bursts of 500 messages per second.