Automating Performance Optimisation in a Highly Virtualised Infrastructure Platform

Sky Betting & Gaming relies on a large infrastructure platform across several datacentres, in addition to services running in the public cloud. We are a business with massive swings in demand with complex applications making up our internal and customer facing products, and large fluctuations in demand every few seconds – such as in the run up to a sports or virtual gaming event starting, or in reaction to a change such as a goal being scored or a match finishing.

Thousands of virtual machines live on top of our hardware, and when operating at such scale it becomes untenable to manually decide exactly which virtual machines should run on what hardware hosts at each minute of the day, and whether the resources configured for each are still correct. As a result we have invested in automation to continually make recommendations and execute automatic changes to optimise our platform and ensure there’s never any lag due to contention for VM resources.

We’re also able to leverage the same technology to be smarter and more cost efficient with our capacity and investment cases, such as detailed forward modelling of capacity additions and scenarios such as the effect of replacing older hardware with newer models – rather than hand cranking spreadsheets and assumptions. This allows us to move forward more quickly and keep ahead of the demands from our increasing base of excited customers!

Together with our partners at Turbonomic (previously known as VMTurbo) we’ve put together a brief video to cover this development which can be seen at the top of this post. The original source can be found at Turbonomic.