Skip to main content

Beyond The Data Centre SLA: The End-user View Of Web Applications

Study after study has shown that if you are a Web-based business and your landing pages are slow to load you will lose business. You will also pay a second penalty, losing search rank, making it harder to recover after fixing site problems. Likewise, an overloaded site can quickly turn a marketing success into a PR problem as clothing store Reiss found out.

Most companies know this, so they ensure that they have a solid SLA in place with their data centre provider covering the performance and availability of their cloud or co-location space so that their apps stay up and are on a platform that should deliver the designed responsiveness.

Unfortunately, guaranteeing that the lights are on, the platters are spinning and that the bandwidth is in place is not enough to ensure success for a Web-based business, your customers do not connect at the cage or even at the edge of your provider’s WAN. Instead, your Web apps must traverse thousands of miles of fibre over multiple networks before reaching their destination. The variables these routes impose play a key role in the overall delivered responsiveness of your applications, and need to be monitored and reported on so that action can be taken to ensure that each end-user’s quality of experience (QoE) remains high.

Some End User Experience Monitoring (EUEM) services on the market can analyse performance both from the internet backbone in over 150 major cities, for an overview of performance, or drill down to customer level via tests run on a network of tens of thousands end-user desktop computers located in multiple countries. It is usually recommended that alerts from the network are copied to systems management teams so that they can start investigating issues and recommending ways to resolve them as soon as they arise.

The payback from end-user monitoring can be almost instantaneous: sometimes tests can highlight that a particular code block was causing loading to appear to pause. The diagnostic information provided will enable the customer’s development team to modify the application so that the page now performs better.

As you run this type of analysis over longer time scales you can establish trend data that informs capacity planning and allows for exception monitoring that can aid early fault detection as well as for strategic planning.

A great example of this is when planning to roll out a service to a new market. If, prior to roll out, you use the EUEM network to run test transactions against the application from your target market, you can compare this to the performance norms to identify if there are any particular local bottlenecks that need to be addressed, for example, by considering moving the load from that market to a more local data centre or modifying the application to split the transactions into smaller parts.

With this sort of flexible capability, I believe EUEM should be considered as part of every Web service infrastructure contract as a complement to standard SLAs. Used fully EUEM will help ensure that not only is the site up, but that it is delivering. Without EUEM can you honestly say you know how your customers see your apps?