carloscastilla - Fotolia
When an application launches, the app owners set IT resource requirements, with the help operations, and revise those demands as monitoring data flows in. Longer-term IT capacity planning is vital to determine when or if operations adds resources to the overall deployment. The answer is not always simply to grow.
In the years predominantly physical server workloads, capacity management was fairly simple, with linear growth curves. You just had to stay ahead that curve. With virtualization and the cloud, abstracted workloads cause data center capacity planning issues on existing and new infrastructure. Allocating VMs is a quick process, and while they are not free, VMs are easily tracked and amenable to chargeback and showback schemes. Trouble appears when you run out virtual infrastructure.
To expand or add to virtual infrastructure comes at a staggeringly high price with new hosts, licensing and storage costs. These investments also tend to have extensive lead times simply due to the cost the infrastructure; quotes and approvals don't happen overnight. IT capacity planning in these circumstances, with the audience that will use it, is difficult. Application owners stunningly excel at padding their IT resource requirements for anticipated future needs. This holdover from physical server days means a capacity planner risks spending a lot money based on application owner data that is structured for the worst possible situation or best possible growth curve. While monitoring and reporting help to show what is in use today and allow for IT capacity planning decisions based on existing data, expectations future usage still normally come from the application owner. The odds are no longer in the capacity manager's favor -- and it gets worse.
The resources freed in a cloud migration
With the increased maturity and prevalence public cloud, the next upgrade to an application can remove it from the data center altogether. This migratory trend plays havoc with IT capacity planning, as the traditional linear curve usage could simply disappear once something moves to the cloud. The app might be gone from the data center -- but it's not forgotten. IT operations still monitors and manages this new resource. Ops also must oversee the excess infrastructure left in the data center when the application vacates the premises.
Does it get retired or moved to a lower tier usage? Can it be reallocated to a growing virtual workload for another user, thereby preventing a new equipment investment? What maintenance does the physical infrastructure and virtualization layer still require? Ask these questions, and determine a plan for the hardware and software in the data center if an application is migrating to the public cloud.
While we in IT pick on the application owners a lot for grandiose capacity plans, the infrastructure admins and managers are not squeaky clean. When faced with continued growth and allocation of resources, IT teams tend to go big to safeguard the infrastructure capacity over the course of several upgrades, rather than incremental updates. Network admins push for more bandwidth. Storage admins want more solid-state drives, even if 90% of the storage in the organization is assigned but unused dead space. They are thinking of what is easiest and best for virtual infrastructure, not sensible IT capacity planning. Cloud doesn't stop this; with no resource restrictions, people effortlessly overprovision.
Guidelines for ops and IT capacity planners
Establish what your organization's cloud strategy is or will be. Don't stop at an overarching strategy; dive into details about applications and application upgrades in the future. The cloud strategy needs input from management and the application owners to determine what is really possible.
Timing is critical. Don't wait to create a strategy until something is ready for an upgrade. Evaluate applications annually, as the cloud and what is being offered in the cloud change.
A new virtualization farm with high-speed connections and expansive storage could be a waste of money if an application moves to the cloud instead of staying on premises. This fluid situation puts operational staff and IT capacity planners in a bind, with no real precedent for them to base decisions on.
Operations needs a multiyear IT capacity plan in place, which is impossible if the decision-makers only look at isolated sections of the data center at upgrade time.
Know where and when an application is due for an upgrade or move to the cloud: IT capacity planning professionals can create a grid or map of where key upgrade or other critical changes occur and use it to correlate the growth curves in the existing infrastructure. While this may not prevent increases in infrastructure, it can shape spending estimates and provide budgeting insight based on what is expected to migrate out in the future. Don't expect to stop investing in infrastructure, just avoid overdoing capital expenditures. With virtualization and public cloud, operational expenditures should increase.
Cloud migration is harder than adoption -- there's no universal right answer
Automate Azure cloud deployments using PowerShell scripts
App scaling in the cloud is a piece cake -- if it's done right