Capacity planning for cloud migrations
The first step to successfully transitioning to a cloud environment involves collecting the right capacity metrics. Next, the enterprise must aggregate, model and visualise this data to see how it would fit into the new cloud infrastructure. Getting the balance right is important. Being too generous with cloud servers may avoid performance issues but may undermine one of the key objectives of cost-cutting.
On the other hand, underprovisioning may cause interruptions to user experience, which could be detrimental to the reputation of the business. Building upon existing resource allocation baselines, ‘What if’ scenario modelling can be utilised to decide which elements of the existing IT estate to migrate to the cloud. Furthermore, banks can consider different scenarios based on varying levels of risk tolerance and cost-cutting intentions. Trial and error in the planning stage can crucially prevent hiccups in the implementation. Capacity still needs to be managed in the cloud. While for some, the cloud projects a future of bottomless capacity, capacity planning will remain just as important. The cloud still has limits, it still costs money and outages will still occur if cloud capacity is not configured correctly.
The ultimate goal of capacity planning in the cloud is to automatically scale compute power using predictive algorithms.
While most of the old practices still apply in the cloud, capacity planning priorities are slightly different. For organisations with a high level of cloud maturity and penetration, the focus will shift towards automating how the cloud environment scales in response to shifts in demand. For example, a retail bank would want to automatically scale up its compute power during Black Friday until Cyber Monday to cope. The ultimate goal of capacity planning in the cloud is to automatically scale compute power using predictive algorithms. with the spike in transaction volume. Instead of hardcoding a rule to determine the extent to which the infrastructure scales, the ultimate goal is to automatically scale using predictive algorithms.
Whereas in a traditional data centre capacity planning may have recommended hardware procurement, if there is a sharp increase in users for an application in the cloud, the compute or storage capacity can automatically change the level required. As soon as the capacity is not needed, it can be automatically scaled back.