Distributed Application Monitoring Challenges

Copyright © Cloud Software Group, Inc. All Rights Reserved

Chapter 1 Distributed Application Monitoring : Distributed Application Monitoring Challenges

Distributed Application Monitoring Challenges

Computer systems today drive many aspects of business, controlling services and processes, transmitting information and identifying patterns and decisions. Increasingly the world is automated - yet who and what monitors the computers doing this work? How can we check their performance from the multitude of system parameters and program outputs?

System monitoring solutions require a number of capabilities to effectively aid the administrators of these systems. The operational alerts they create can also be used themselves as useful information in business applications that might track outputs from computer based measuring systems for example.

Monitoring Requirements

There are several requirements for operational systems monitoring. These include:

•

Minimization of side-effects: Monitoring software should have negligible resource footprints, and should have a minimal effect on the performance of observed systems.

•

Minimization of administration costs: Monitoring software should not be a maintenance burden, but a reliable beneficial tool, for the operations staff.

•

Maximization of coverage: Monitoring software should monitor all the aspects of modern systems than are of interest to operations staff.

Most monitoring systems follow the keep it simple mantra for implementing monitoring. They might collect data via logs into a centralized server where scripts can execute against the monitor data – traditionally log file records – to determine required actions. However, TIBCO Hawk’s designers anticipated some additional requirements:

•

Ability to handle distributed systems and very large networks of computers: with the transition from mainframe to client-server to distributed computing throughout the 1980s to 1990s and 2000s, more fault-tolerant monitoring systems were envisaged that would not require administration of centralized servers and specialized fault-tolerance mechanisms

•

Ability to grow to new monitoring requirements with minimal system intervention: both new type of measurement and new source of measurements would be required as new types of hardware and software were/are deployed over time.