To avoid the risk of capacity exhaust and unplanned infrastructure costs, capacity and demand planning are a critical part of the data center operations. vRealize Operations Manager provides a capacity management solution that continuously monitors and advises on the health of the NFVI environment.
Right-Sizing
Over-provisioning the infrastructure can be avoided by right-sizing the resources. Capacity risk tolerance can be set for a trial environment that is different from the production environment. You can plan for infrastructure procurement or migrations based on available capacity.
As the CSP considers the importance of their environment and workloads, they can set the capacity risk level to Conservative by using the upper bound projection of the forecast for high-performance workloads. In contrast, they can set the capacity risk to Aggressive policy by using the forecast for non-critical workload environment.
Head-room capacity can also be defined on the cluster to ensure that there is space for unexpected spikes in traffic and performance for critical workloads.
Forecasting
Demand planning might be challenging, especially if the CapEx spend should be balanced over time. The forecasting engine can be used to extrapolate existing utilization of workload CPU, memory , and storage across VMs, clusters, and data centers. The just-in-time capacity forecasts can be used to plan for new infrastructure or rebalance with predictable intelligence.
New Service Order
As an operations manager, when a new tenant order needs to be fulfilled you need to make decisions on where to place the new workloads without impacting current operations. Various approaches can be used in the data center clusters including making room for the new order by moving workloads, adding additional resources to the existing workloads, and adding new hosts to the cluster. You can use vRealize Operations What-If analysis tool to emulate such scenarios by using the existing workloads as the target profile. You can model additional CPU, memory, storage to existing workloads to calculate new capacity exhaust. You can emulate new instance of similar workloads to estimate if there is enough capacity in the cluster.
