Blogs

Latency – an application performance killer

By Chris James, Director of EMEA Marketing

The aesthete Stephen Bailey once interviewed the captain of an SR 71 “Blackbird”, the fastest air breathing plane that once flew from New York to London in just under one hour 55 minutes. When it is fired up the two ramjet engines are in sync. If one has an ‘unstart’, the USAF’s parlance for a stall, it takes a few seconds for the on-board computers to react. When asked what this experience was like the captain said “just watered the S*** right out of your eyes”.

So, latency can kill! The above may be an extreme example of latency but a recent study by Business Week found that the average attention span of someone on a web site is 8 seconds – that’s a second less than the memory of a goldfish! If they don’t get what they want in 8 seconds, they move on. Some of Virtual Instruments largest on-line merchant customers count the cost of latency in thousands of dollars per second, heath-care companies really do look at it as a life or death issue.

So what causes application latency?  Our findings are that it is rarely a single ‘smoking gun’. Data centres grow over time and a common remedy for performance issues is to add more hardware. If you are looking to expand capacity quickly and relatively cheaply then you can add in more SAN or NAS storage, but networked storage is notoriously difficult to analyse – until now. Virtual Wisdom, from Virtual Instruments, monitors application performance across the whole data centre, in real time – and by that I mean ‘line speed’. It’s off-line approach to analysis is unique and allows an entire end-to-end view of the data centre, from VM on the servers to LUN or file system on the storage arrays, and everything in between, in a single view.

Virtual Wisdom is used by hundreds of enterprise organisations to de-risk their critical applications and ensure they perform as they should. Application response time SLAs can be set between IT and the business so everyone is confident that customers and end users will not have to wait for service.

By analysing application infrastructure and getting the performance right you can avoid unnecessary stuff coming ‘out of your eyes’.