Friday 3 March 2017

The changing face of Capacity Management (1 of 4)

The view of capacity planning when I started in this profession over 15 years ago is different now than it was then.  Yes, many mainframers have told me that they had virtualization back then and nothing new is really new again.  I'll concede that to a certain degree, but for those of us who came into a client-server world, things have changed dramatically.

Back then, planning for me was the building, calibrating, evaluating, and creating scenarios revolving around analytic models for a single system.  Workloads were the items of importance and quite frequently we looked at technical numbers and pretty much ignored the business.

Things have changed.

Today I spend more time talking about capacity as the overall amount of capacity remaining in a virtualized datacenter rather than the capacity remaining on a server.  I spend more time talking about central storage and SAN performance than I do talking about locally-attached storage.  And for me the workload view has been replaced in many cases with a real-user transactional view we couldn't hope to get back in the days before Y2K.  It's the only way we know whether we're meeting those SLAs.  With services hitting many different tiers (and potentially disappearing into the cloud), that transactional view can actually be the easiest way to measure performance.  How much of that transaction is visible to the planner depends on the cloud implementation and on the service provider.

So, the cloud. Sure, it would be ideal to have knowledge of everything that's happening inside the cloud, but if you're paying for software-as-a-service (SaaS), you probably aren't going to get that from your vendor -- and, to be honest, why would you want that view?  You are paying a premium in order to reduce computing down to the utility level -- flip the switch and the light comes on.  When I turn a light on in my kitchen, I don't think about the wires, the devices, or the electrical grid -- I just want to know that the light will come on.  And I take for granted that there's enough electricity in the system to service all my household needs.

But we pay a price for that, and there's a chance that instead of paying for excess capacity in your data center, you're paying for excess capacity (for yourself and for other clients of the vendor) in the cloud.

In the next few installments, I'll start looking at clouds in more detail.  I'll focus on the different types of clouds, the different types of services that can be purchased from cloud providers, and how that affects how you'll do capacity planning.

Rich Fronheiser
Chief Marketing Officer

No comments:

Post a Comment