Home

Time your server - response time benchmarking

With the best will in the world, you cannot tell how a server is running just by looking at it. All your monitoring data may look normal: no peaks showing something’s overloaded, no troughs to say that work has stopped. However until you can get data on how the users (whether they are employees of your organisation or visitors to your website) are experiencing the system you cannot say if the server is delivering a high quality of service.

Dalleon Systems’ TIME YOUR SERVER facility gives you objective, numerical information showing the response times that your interactive applications are providing. It does this by running some of the same queries that the users run, with the same kinds of data. It can even work from a remote location, to include network propagation times into the response time.

The process involves identifying the most frequently run requests, inserting realistic query fields and measuring how long the operation takes to complete. By using several different requests from differing applications, forms or web pages, the service can build up a picture to tell you if some areas are running slowly and even help you identify the problem areas. By running the same queries with the same data at set times, we can tell you when the users are most likelyto be dissatisfied with their applications and when low usage means you could reschedule some resource intensive tasks without compromising response times.

You can use this setup to help assure yourself that performace SLAs are being met, and since response time data is retained, trends, patterns and peak times become apparent.

The service needs dedicated PCs to both host the response time measuring package and also to run the application. These can be supplied by us, or we can utilise hardware you already own – possibly by recycling spares that were due for disposal.

As well as deploying a single unit to gather response times from one location – even if it is where all the users are, judicious placement of more monitors can help isolate any problems. For example, siting a monitoring platform in a remote office can give a more realistic measurement of how outside users experience a website, than if response times were measured from the same network segment the server is on. Alternatively, placing a monitor away from internal users, in a server room can tell you the absolute response times without any network overheads. When you compare timing information from various locations, you can quickly work out where any delays come from.

This service captures response time data, but needs something extra to collect, analyse and report results. This can be as simple as a support person with a spreadsheet, or the data can be integrated into a more complete performance management system, that includes features such as automated reporting and alerting if response times are too long.