Peter Duffy is the CTO of Sumerian, a UK based capacity planning company that aims to change the way progressive enterprises manage the performance and efficiency of their business-critical IT services.
Overprovisioning is the domain of the overambitious, the uninformed or the anxious. The first is understandable, it’s normal to want to be prepared in the face of unprecedented demand. However, a recent survey found that relying on the cloud and taking an ‘overprovision and forget’ approach to capacity management can directly increase IT costs, an unnecessary risk given the current global economic environment.
It’s an important issue that still needs to be addressed by many, with a recent study from Anthesis group pointing towards a market where overprovisioning is rife. Up to 30 per cent of servers worldwide lie in an unused state.
To get on the front foot and ensure you have the right cloud resources in play, at the right time and at the best cost, good planning is essential. This means having the tools in place that can help pinpoint the precise factors affecting service availability and performance, identifying exactly how much cloud resource is required and at what point, and understanding and choosing the best cloud pricing option.
A demanding environment
Demand on IT infrastructure undergoes seasonal variations in all types of industries. For retailers, there’s Black Friday, for financial services it’s the end of the tax year – the list goes on. Before the advent of cloud services, the only option was to resource based on expected peak demand. For example: “My peak demand is going to be 5000 transactions per hour, I need ten web servers, 20 application servers and 30 database servers to ensure ongoing service.”
In this scenario, the infrastructure is ready to cope with peak demand, but the problem is, this demand only comes around once a year. For the rest of the year, that infrastructure is just sitting there idle. To a certain degree, the cloud helps businesses address this issue by allowing them to purchase capacity as and when it’s needed. However, buying cloud resources on demand can be expensive and can have an unexpected impact on IT budgets if not managed correctly. To avoid this, it’s important to have the capabilities in place to help provision accurately, and to fully understand the impact of different cloud pricing models.
Choosing the right option
In an ideal world, the cloud is available on demand, offering users the option to pay for what they use, as they use it, and to switch services at a whim with no commitment. In reality, this isn’t the case and there are many different offerings and models that organisations need to be aware of.
Amazon, for instance offers spot, on-demand and reserved pricing.
- Spot pricing resources will be available instantly and can be used to react to expected surges in demand, but are subject to rapid changes in availability (Amazon basically uses this pricing model to sell its excess capacity).
- On-demand pricing is the most expensive, but provides huge flexibility as you only pay for what you need at the point in time you use it.
- Reserved pricing is often cheaper and is generally used when you provision based on your expected peak demand, ensuring enough cloud infrastructure is in place at all times. Companies will also often take this option if they fear that they could experience a surge in demand at any time.
From firefighting to fire prevention
All in all, the best pricing structure to go for depends largely on the business and industry, but to realise true cost efficiencies from the cloud, applying advanced capacity planning and predictive analytics is essential. A retailer, for example, will have a baseline of demand throughout the year, and at various points will experience seasonal peaks. If it applies predictive analytics, it can accurately model and identify its peak demand and expected fluctuations in demand. Once it has this picture, it can weigh up its cloud pricing options and plan accordingly for its infrastructure on a seasonal basis. It can also use predictive analytics to accurately scenario model and test out its plans in advance, de-risking and assuring performance before the infrastructure is put in place.
The value of an analytics driven approach
No matter how an organisation structures its IT infrastructure and pricing, taking an analytics approach will ensure:
- Past demand trends are leveraged to gauge what to expect in the future
- The infrastructure is optimised to cope with this demand
- Data driven intelligence is used to accurately assess options.
Cloud is not a one-size-fits-all, it is a varied and complex pricing environment. Organisations that benefit from it and experience true cost savings, will be those that plan ahead and consider all their options, in turn ensuring informed and strategic decision-making.