The other day Ohm Malik wrote an interesting post about the availability issues with Apple’s MobileMe site. He made quite a few key observations related to monitoring and load testing that I wanted to reiterate here. They are:
- There is no-unified IT plan vis-a-vis applications; each has their own set of servers, IT practices and release scenarios. This is becoming more and more of a problem with adoption SOA architectures, SaaS, and mashup models (see yesterday’s post for more information on monitoring these types of architectures). You now need to work closer with your partners when performing load testing to ensure that all components of your application are thoroughly tested. A load test is of little value if it fails to test 25% of your application/infrastructure. This requires more up-front time planning a load test and if you’re using a vendor for load testing then ensure that they provide services that are consultative and strive to understand your environment and infrastructure.
- There’s no unified monitoring system. Monitoring data from all sources (server performance, external application performance, analytics, etc.) needs to be considered and in an ideal environment would be mashed up to gain new insight into the data. Consistent and fine-grained monitoring intervals are two properties that are important in effectively monitoring a web application.
Even the giants like Apple will feel the pain if they fail to ensure performance and availability of their applications through pre-release load testing of infrastructure and on-going monitoring for performance and continued availability.
The performance for this blog (today’s average load time and uptime):
- Average load time is 1.81seconds.
- Availability (uptime) is 100%.