Solving a Data Platform problem with an Infrastructure Answer

Jon Cowling 08-Feb-2016 11:17:05

Solving a Data Platform problem with an Infrastructure Answer

Recently, a new client gave us some interesting feedback on a few Infrastructure Providers he met during his quest to rationalise his Database estate.

The requirement was to tackle SQL Server Sprawl. Whilst they did in fact virtualise, they made the mistake of ‘virtualising the sprawl’ i.e. they had 67 physical SQL servers; now they have 67 virtual SQL Servers!

Granted they solved a hardware and power/cooling problem – but the bigger problem was still apparent. They still had to manage 67 Operating Systems, 67 MS SQL patching, 67 VM’s to backup and 67 SQL servers requiring storage, CPU and IO.

Therefore, he invited dsp and 3 other service providers in to talk with him about specifically tackling the Data Platformproblem.

The first Infrastructure Provider’s answer was to buy more software and hardware, which will cleverly manage their ever expanding SQL server farm.

The second offered virtually the same solution but Cloud based – an Infrastructure as a Service, transferring the management of the Data Platform to them the provider, thus [apparently] removing the headache.

Whilst both (the former to a lesser extent) gave some benefit, they came at a cost and with doubt, as this particular organisation can’t yet go to Cloud due to a complex Application Tier. But the important thing was; neither actually solved the problem! They only threw more infrastructure, cost and complexity at the problem.

We went in with the position of consolidating and modernising the data platform by moving databases and instances to shared SQL Servers, thus changing the 1 Server>to 1 SQL Instance> to 1 Database model.

We initially performed a small consulting exercise to determine what databases was critical and needed attention, how busy the databases were, and what instance consolidation could be accomplished. We then developed a business case showing exactly what can be consolidated and what cost savings could be achieved. Turned out we could reduce the Data Platform from 67 SQL servers to less than 15, and see an ROI in less than 14 months.

The moral of the story is sometimes the problem doesn’t lie in the Infrastructure, but in how you use it and maximise your existing investment. Addressing the Data Platform layer means aligning the infrastructure to how the business needs it to work – not trying to align the business on how the infrastructure works.