Availability Metrics

Imagine you are in the middle of a stock transaction on the Internet and then the application hangs for more than half an hour . Or, it is Friday morning and you are running your quarter-end financial report, which is due Monday, and your system crashes and stays down for 96 hours. Or, you are doing a business transaction on a well-known web site and the server supporting it crashes, rendering the site unavailable for a prolonged time. Scenarios such as these are far more frequent than desirable and are incurring costs and decreasing productivity.

The following was from a financial news posted on the Netscape Web site not long ago:

Nasdaq will extend trading for one hour today, to 5 P.M. EDT, after a network problem forced the electronic stock market to temporarily suspend trading through the SelectNet and Small Order Execution System, or SOES.

Today marked the second day in a row and the third time this month that the Nasdaq has had problems with its execution system.

In this Internet age of network computing, one of the most critical quality attributes is system and network availability, along with reliability and security. Requirements for high availability by mission-critical operations have existed since society became reliant on computer technologies. In the Internet age, software code is distributed across networks and businesses increasingly share data, a lack of system availability is significantly increasing adverse impacts. In this chapter we discuss the definition and measurements of system availability, possible approaches to collecting customer's outage data, and the ways to use customer's data and availability metrics to drive improvement of the product.