So, how does capacity management change when moving from a server-oriented or client-server environment to a private cloud?
First, it's important to understand that most organizations have many internal customers -- and that moving to a private cloud eliminates the separation of computing resources that client-server computing had always promised. For some of those customers, this may be an uncomfortable arrangement and it's important that capacity management and the regular communication of capacity and performance information be in place to satisfy those customers.
Performance and capacity monitoring must be done at multiple levels. While it's possible (and useful) to monitor what happens inside virtual machines, it's far more crucial to monitor from a more global perspective. How much additional capacity is available for unseen peaks and for quick virtual machine deployments?
Over-procurement and over-allocation of resources to one internal group may not be a major expense. But when resources are over-allocated to dozens of internal groups, the incremental cost to the organization itself could be quite large. Furthermore, over-allocation of physical resources to manage the cloud means that hardware is being purchased too far in advance, likely for more money than if those purchases were delayed until needed. Of course these concepts are nothing new -- it's just that within the cloud those additional purchases may be seen as necessary overhead, when in fact it's just a problem of over-allocation pushed up from individual customers through to the organization itself.
Likewise, under-procurement and under-allocation of resources can cause problems in a cloud environment. One of the selling points of the cloud is rapid deployment and complete flexibility - when a customer needs resource, it can be quickly made available. If a company doesn't invest in having *enough* resources available, then there is little advantage for internal customers to agree to a move to a private cloud. Further, the inability to provide sufficient resources in this model means that many internal customers may not be meeting service level agreements at crucial times.
In my final instalment on Monday I'll talk about implementing a capacity management mindset that specifically deals with some of the challenges of working with a private cloud.
Chief Marketing Officer