Denys Rudyi - Fotolia
Editor's note: This article is part two in a two-part series on identifying and preventing cloud sprawl. You can read part one here.
While there are many possible causes of cloud sprawl, two that are especially concerning to enterprise IT teams are hybrid and multicloud deployments.
A common thread in hybrid and multicloud deployments is the division of a cloud resource pool. In either a hybrid or multicloud model, applications can be hosted in several places, which likely differ in terms of their deployment options, base costs and the way in which those costs scale with usage. It's difficult to avoid treating each environment -- whether a public cloud service, a private cloud or other hosting model -- as separate resources. But that can violate the basic principles of economy of scale that justify cloud in the first place.
Resource pools, such as those found in the cloud or in virtualized data centers, lower costs because they avoid dedicated commitments of applications to servers, which can result in underutilization. However, if the resource pool is subdivided into many specialized pieces, based on the type of cloud provider or data center server, for example, then the full resource pool isn't available to any given application and overall efficiency decreases.
Imagine an enterprise has a dozen different virtual machine (VM) types -- with different CPU, memory and so forth -- each supporting its own group of applications. You'd have to plan to supply enough hosting capacity for each of those VM types, including excess capacity for scaling. But all that excess capacity across all those different VM types adds up. If you could limit your VMs to a few types, you'd be able to assign capacity more efficiently.
When you over-segment your cloud resource pool, a small error in application planning can mean that you can't run a given app because resources aren't available. This is a particular problem in hybrid clouds, because congestion in the data center may cause workloads to burst to the public cloud, which can radically increase costs.
To avoid this problem, reduce the number of different configurations to prevent fragmentation in the cloud resource pool. Pick the smallest possible number of server, VM and container configurations so that applications have a broad choice of where they can run. You may find that settling on one resource type or configuration, even if some of the resources are wasted on small, undemanding applications, is still less costly overall.
It's surprising how rarely enterprises take this simple step. Most don't try to harmonize their first public cloud service with their VM configurations in virtual data centers, for example, and even fewer adopt a single resource configuration across multiple cloud providers. Any cloud plan, whether aimed at public, private or hybrid, should start by classifying applications by their resource needs and creating a small number of standard hosting models.
Effective resource planning for hybrid, multicloud
Hybrid and multicloud deployments complicate resource planning because they link one legacy planning process -- data center planning -- with processes that are, at best, different and, at worst, driven by organizations that don't coordinate with the data center or with each other. From a planning perspective, start thinking not just of data center operations, but of IT operations. Broaden the role of the IT organization to harmonize hybrid cloud resource planning.
That won't work, however, if line departments who avoid central IT coordination drive cloud adoption. The biggest shadow IT challenge is getting professional involvement in cloud planning without alienating line organizations. Naming a cloud coordinator to represent all public cloud services within the company is a good approach. This coordinator should view the data center as simply another cloud resource.
Careful cloud service planning is essential, but it's also important to review new cloud contracts to ensure cloud resource pool fragmentation doesn't creep in. This is particularly true where shadow IT or departmental cloud service contracts are let in without central IT review. For both hybrid and multicloud, a dedicated cloud coordinator can balance resource contracts against application needs to ensure money isn't wasted and that any changes in cloud provider pricing or terms are reflected properly in the cloud plan.
Dividing a cloud resource pool is an inherent risk, and one that is heightened when multiple buyer jurisdictions are involved, which is usually the case in multicloud and always the case in hybrid cloud. Careful planning and ongoing service coordination can reduce these risks, but remember that, where possible, using a single cloud provider is one of the best ways to prevent sprawl.
Control and avoid VM sprawl
Get familiar with the multicloud model
Is hybrid cloud right for you?