A major public sector organization with a mixed virtual and physical estate had hardware approaching end-of-life. Before entering a procurement cycle to replace it, the customer wanted to understand exactly what was actually required to meet its current and future demand, so that it could ensure it only bought what it really needed.
As a strategy, in addition to re-virtualizing its estate and performing a further physical–to–virtual migration, the customer also had a clear objective to reduce the number of hosts required to support its workloads — so that it could realize savings on hardware, software licensing and associated support costs.
In order to proceed, it was essential for the customer to build up an accurate understanding of capacity and utilization across its current compute and storage resources, and accurately forecast the capacity required to support planned future growth.
Although the customer had invested heavily in VMware vCenter Operations Manager, it did not have a vendor independent, unified view across its mixed physical and virtual workloads, datacenters and different environments. Nor did it have the capability to model what a right-sized environment would look like on the new hardware platforms under consideration.
Sumerian’s Capacity Planner CPaaS offered the customer a fast and easy way to establish both these things.
The customer environment was split across two main datacenters in an active/ active configuration, using VMware Distributed Resources Scheduler (DRS). Sumerian worked with the customer to identify the best existing data sources to use from this environment — vCenter and Microsoft System Center Operations Manager (SCOM) — and set up an automated data feed into Sumerian Capacity Planner, providing the customer with a daily update of storage and compute usage/growth.
The customer had various workload rules. For example, workloads for a particular service must be kept together and not split across hosts, certain workloads could not be moved due to licensing restrictions and some workloads must be isolated from the rest of the estate for security and confidentiality reasons. This workload placement information was also uploaded into Sumerian Capacity Planner.
With both data and workload information uploaded, the customer was able to view a comprehensive Baseline of its entire estate. Sumerian Capacity Planner’s powerful Sunburst visualization providing the customer with instant insight across the whole infrastructure, and clearly showing both under- and over-utilized parts of the estate.
Then, using Sumerian Capacity Planner’s Forward Thinking capabilities, the customer was able to model various consolidation scenarios to understand how best to rightsize its estate. It was able to see that although parts of its estate were constrained, there was significant capacity in other areas that could meet the demand without increasing the number of hosts. It also became apparent that there was overcapacity of CPU processing power across the estate and that it could reduce the number of cores by 160 (approximately 20%) while maintaining current service levels.
The customer also used the Forward Thinking capabilities of Sumerian Capacity Planner, to model complex disaster recovery scenarios to understand how its failover strategies impacted on both the current and new infrastructure, including growth profiles. For example, Sumerian Capacity Planner was able to highlight that one of the main datacenters was significantly under-provisioned for memory should one of the DR scenarios occur.
Outcome and results
The results of the capacity analysis and scenario modeling conducted using Sumerian Capacity Planner provided the customer with the quantified, data-driven evidence it needs to feel confident it can significantly reduce its current 60-plus host and physical server estate – by up to 70% – when progressing its hardware refresh. And to ensure those refresh plans provide adequately for all likely DR scenarios. The DR modeling has also had a short term impact, leading to the customer adapting it’s current DR failover plans to only failover critical applications — until the new hardware is deployed.
As a secondary step within the procurement process, the customer will also use the Forward Thinking scenario modeling capabilities of Sumerian Capacity Planner to explore and validate the various vendor responses it receives through it’s procurement process.
This will ensure the customer achieves a rightsized infrastructure, able to accommodate anticipated future growth while at the same time saving significant costs on both capital expenditure for the new servers themselves, and the follow on operational costs such as space, cooling, power, people and software licensing.