E-Handbook: What's new in private cloud management? Article 4 of 4



Manage private cloud performance and troubleshoot issues

Management tools intended for private clouds are key to troubleshooting performance issues because they provide insight that traditional virtualization management tools might not.

Because private clouds use virtualization, problem resolution works similarly to the way it does in a normal virtual infrastructure. The difference is that the private cloud architecture adds layers of complexity on top of the virtualization stack.

With this type of private cloud architecture, you as the IT administrator have the same responsibilities as a provider would. The tenants of the private cloud are the individual departments in your organization, each with their own policies and administrative tasks.

The key to resolving many private cloud performance issues is to have good management tools that provide insight into the private cloud. The tools that are designed solely for use in virtual server infrastructures are often inadequate for use in private clouds because the tenant-level isolation boundaries might prevent the tools from being able to see the full stack.

Suppose that a tenant within your private cloud creates several virtual servers, as well as a dedicated software-defined network linking those VMs together. Assume that this collection of resources works flawlessly for several months, but one day it begins to unexpectedly suffer from poor performance. Many issues could have caused the problem.

One possibility is that when the tenant created all of the virtual servers, the VMs were all hosted on the same physical server, but one or more of the virtual servers were recently migrated automatically to a different host to balance the demand for physical resources. Prior to this migration, the software-defined network might have been purely virtual; it might never have used any physical network hardware because the virtual servers resided on the same host. When the virtual servers were migrated to a different host, it suddenly became necessary for the tenant's logical network to share network bandwidth with other tenants.

A monitoring tool designed solely for virtual server infrastructures would no doubt be able to alert you to the increase in traffic flowing across the physical network. But because the tenant owns the virtual network in question, the monitoring tool would probably have a tough time identifying the specific tenant resources causing the congestion. You need management tools that let you view the entire private cloud stack and don't treat tenant resources as blind spots.

Avoid potential performance problems

Tenant issues aren't always the cause of private cloud performance problems, but having good visibility into the full private cloud stack is the key to identifying and fixing the problem. Suppose, for instance, that one of the virtualization hosts in your private cloud runs an older version of the hypervisor. Depending on the platform you use and the age of the outdated hypervisor, this type of mismatch can cause a variety of seemingly random problems. It can be surprisingly difficult to determine that an incorrect software version is the root cause of the problem. A good management tool should alert you to architectural inconsistencies in the private cloud infrastructure.

You need management tools that let you view the entire private cloud stack and don't treat tenant resources as blind spots.

One of the keys to keeping private cloud performance problems at bay is to prevent tenants from consuming excessive resources, which could affect other tenants. All the major private cloud management platforms include tools to cap an individual tenant's resource usage. Even so, the private cloud has a finite amount of hardware resources that tenants must share.

It's fairly common for an organization to over provision its hardware resources because it's unlikely that every tenant will max out its resource consumption at the same time. Still, over provisioning only works if you keep track of the demand for resources over time. Neglecting this task leads to performance problems and eventually to resource depletion. As such, you must prioritize resource utilization tracking and capacity planning to avoid private cloud performance issues.

Dig Deeper on Cloud deployment and architecture

Data Center