The State of Performance Management for Hybrid Clouds

September 2, 2016
194 Views

Performance Management for Hybrid Cloud

IT Administrators have more choices today than ever before on how to run the workloads that are mission-critical for the business — physical, virtual, on-premises, in Clouds, or some combination. That’s why it’s vital to find a performance management toolset to help inform those choices.

Performance Management for Hybrid Cloud

IT Administrators have more choices today than ever before on how to run the workloads that are mission-critical for the business — physical, virtual, on-premises, in Clouds, or some combination. That’s why it’s vital to find a performance management toolset to help inform those choices.

The Move to Virtual

In virtualized infrastructures, a workload may not be getting the resources it needs because of a “noisy neighbor” workload that is overburdening a host, one that may be running potentially dozens of VMs. Another possibility is that a mission-critical workload may be relying on some supporting servers or services that are the culprits of a bottleneck. It is impossible for humans to track the complex interactions within a medium-sized virtualized infrastructure to eliminate bottlenecks, and get the most from their IT investments, without performance management tools. The dilemma is that there are many to choose from and they often sound alike.

The Move to Cloud

In a pure cloud environment, the techniques used to collect metrics from the virtualized data center no longer work. This means that performance management tool vendors have to adapt by making the changes necessary to discover, map, monitor and manage all of the workloads and resources available to run those workloads when they are run in clouds. For instance, in the datacenter a performance management tool typically calls vCenter to get a wealth of metric information on hosts, datastores, and guest VMs – but in AWS or Azure, there is no vCenter server. As a result, newer performance management tools that support clouds often cannot handle data centers and the older, more proven tools used in the data center do not support workloads being run in clouds.

The Hybrid Cloud Challenge

How can a performance management tool support workloads that are run in the datacenter, moved to the Cloud for an anticipated spike in demand, then need to be moved back into the data center once the demand spike is over? Relying on the infrastructure is a problem because every Cloud uses different technologies. Requiring application developers to include a monitoring and control agent when their application is compiled is not a realistic solution, as it does not support existing applications and does not scale.

The Hybrid Cloud Solution

In order to reap the benefits of hybrid clouds, IT Administrators are looking for performance management tools that can manage all workloads, whether they are running in the datacenter or in Clouds – from a single pane of glass. Such tools may use a combination of techniques such as collecting information from the infrastructure layer; vCenter or Hyper-V in the data center or AWS , Azure integration in clouds, as well collecting data from the workload OS, be it Windows or Linux OS, either by using an agent or probing the VM with API so it can report on individual application metrics back to the management application.

Application-level Visibility

In the data center, the IT Administrator looks for overloaded hosts. Once found, the VMs on that host are inspected to see if they are starved for resources or using more resources than is reasonable. In either case, a robust root cause analysis requires visibility inside that VM to the sessions and processes that may be exhausting these resources. Moreover when a problem is detected and root cause found, remediating this problem from within the same toolset in that context would save a lot of time and offer uniformity across the variety of architectures and choices

Conclusion

Look for a performance management solution that can handle both the datacenter and Clouds from a single pane of glass, that provides visibility right down to the application level, and ideally, can fix problems as they arise and are identified.