posted

1 Comment

The newest release of vRealize Operations (vR Ops) is packed with new features that can benefit any vSphere environment. vSAN is the storage system that turns the SDDC into a reality, and the emphasis on enhancing vR Ops for our vSAN customers is a clear indicator of vSAN’s success at the edge, the core data center, and the cloud.

One of the capabilities introduced in vR Ops 7.5 is support for vSAN in the “Workload Optimization” feature. This allows vSAN powered VMs to have mobility across vSAN clusters through a workload automation policy setting in vR Ops. This cross-cluster mobility VMs will occur as a result of the administrator determining the policy objective: Balance workloads across vSAN clusters, or consolidate VMs to as few clusters as possible.

vR Ops 7.5

Figure 1. Setting the policy objective for workload optimization in vR Ops 7.5

More than just VM movement across Clusters

It is not just the ability to move the compute and storage across clusters, but rather the decision-making process used by vR Ops that makes these intelligent optimizations. This new Workload Optimization feature for vSAN is respectful of the following conditions:

Storage Policies. Will the VM be able to satisfy the conditions of the storage policy applied to the VM(s) to the new cluster? The target cluster could be comprised of a different number of vSAN hosts and other conditions that could influence policy compliance, so this verification ensures that policy compliance can be met prior to any movement of data.

Slack Space. Will there be enough slack space maintained at the target if the VM were to be moved? For this cross-cluster VM mobility, vR Ops checks for more than just free space on the target cluster. It will factor in the recommended slack space to ensure that the VM to be moved does not violate this threshold. Since vR Ops 7.5 understands the storage policy used for the VM, it is fully aware of the total raw capacity that the VM will consume at the target location and includes this in the decision-making process to ensure the recommendations for slack space are honored.

Resynchronization Activity. Is it a good time to move the VM from one cluster to another? Within a vSAN cluster, there may be resynchronization activity occurring at the time the optimization is initiated. Resynchronization activity is vSAN’s way to establish and maintain levels of resilience of object data, as defined by its associated policy. vR Ops is able to detect when and where resynchronization activity is occurring and will prevent workload migrations across clusters from occurring if active resynchronizations are detected. While vSAN 6.7 introduced Adaptive Resync 2.0 to help with the sharing of resources during resynchronizations, this vR Ops check understands that cross-cluster VM migrations are an optimization effort only, and is intelligent enough to look for a time in which resynchronizations are not occurring.

Why is this important?

To understand the real value that this brings to vSAN clusters, we need to differentiate how vSAN clusters are different from traditional vSphere clusters. vSAN treats storage as a cluster resource. Unlike other architectures of the past, having this boundary of storage limited to a cluster mimics other cluster resources like compute and memory. This allows vSAN to provide a truly unique, end-to-end clustering experience, where you can tailor the size of the cluster (compute and storage) to better meet the demands of the organization. Storage as a cluster resource is one of the true gems of designing and operating vSAN powered environments, but for proper movement of a VM across vSAN clusters, both the compute and storage consumed by the VM must be moved to another cluster.

Past versions of vR Ops were able to provide this cross-cluster workload optimization feature, but it was limited to clusters using traditional shared storage, as it used DRS and sDRS to assist with movement of VMs across clusters. Now, this feature is introduced natively to vSAN. Administrators can benefit from the end-to-end clustering that vSAN provides, yet still maintain the mobility of workloads per the requirements of the organization. The workload optimization feature in vR Ops can help those who are considering the design and operational considerations around environments using larger mixed-use vSAN clusters using many hosts, or smaller purpose-built clusters using fewer hosts.

Putting Workload Optimization into Practice

Let’s look at a few examples of how this feature could be applied to vSAN powered environments.

Example #1 – Cluster Balance: Imagine an environment with a large number of hosts and various vSAN cluster sizes. Instead of using extremely large clusters, perhaps the design called for using smaller vSAN cluster sizes to benefit from smaller management and maintenance domains. Choosing the “Balance” option in the vR Ops workload optimization feature will redistribute the VMs across vSAN clusters to even out the use of resources to treat the multiple clusters as an aggregate pool of resources.

vR Ops 7.5

Figure 2. The “Balance” option in the Workload Optimization feature in vR Ops

Example #2 – Cluster Consolidation: Another scenario might be an environment that has a sprawl of similar or dissimilar vSAN clusters in an environment. The Administrator might see an opportunity for consolidation, but isn’t sure on the best way to approach it. This is where vR Ops can make a data-driven decision on how best to consolidate the workloads into as few clusters as possible. vR Ops will help ensure that the operational requirements such as policy and capacity requirements can be met, while running the optimization at a time that is most suitable for the environment.

vR Ops 7.5

Figure 3. The “Consolidate” option in the Workload Optimization feature in vR Ops

A third option comes as a byproduct of these other two objectives demonstrated: Choosing the “Moderate” setting in the middle of the slider bar will strive to reduce workload contention, but will not seek to aggressively move workloads to achieve optimal balance or workload consolidation. Think of this as a similar approach to the default settings found in DRS, where it takes a conservative “as-needed” approach to reduce contention while minimizing overall data movement.

Summary

vSAN powered environments that using vRealize Operations can not only help drive smarter decision making in your vSAN powered environments, but it can also take action for you. To learn more about the features of vR Ops 7.5, see “What’s New in vRealize Operations 7.5 – A Technical Overview.

@vmpete