How to Keep Your Apps Running Optimally: 10 Best Practices

by Chris Preimesberger

Failed connections are surprisingly common, often leading to lengthy retries and time-outs. Typically, they indicate that either a service is broken, or that your application has been misconfigured. After detecting a failed connection, investigate both to get to the source of the problem.

Hidden and misconfigured dependencies (development rather than production databases, for example) can result in application-performance shutdowns. By knowing where these flaws exist, you can correct them.

Application error messages are indicators of problems with either the application or the service. Don’t ignore these. Whether it’s a simple “404 Page Not Found” or a more complicated issue like a Customer Information Control System (CICS) transaction failing in the mainframe, it should be investigated and, if possible, eliminated.

When a system is saturated beyond capacity, application components may take more than 60 seconds to respond to requests. If requests are “hanging,” review capacity plans and check configurations.

Poorly designed applications can result in unbalanced query structures in which individual requests generate multiple back-end calls that slow throughput to a crawl. IT operations teams should identify these situations and work with development teams to establish appropriate balance between front-end and back-end requests.

Management activities such as software upgrades, patches and antivirus checks can be enough to impact application performance if they run during peak hours. Be sure you have visibility into all the processes that run on a system to ensure that management tasks are appropriately scheduled.

Pre-production scalability tests can flag common architectural problems and allow teams to solve them before they disrupt transaction paths.

Changes within the application and/or infrastructure architecture of a system will cause slowdowns and outages. Before a new launch, teams have to verify that any changes will have no impact on service levels within the impacted component area.

Misconfigured servers can request Domain Name System (DNS) entries that don’t exist, and time-outs won’t occur until five seconds have passed. Requests such as DNS lookups, reverse lookups and start of authority transactions can all be subject to delays, thus impacting performance. Identify DNS configuration errors and avoid the performance penalty.

Contrary to popular belief, high-resource utilization doesn’t necessarily indicate an end-user problem. Focus on transaction response time rather than resource utilization to quickly identify real performance problems.