Case Study: Reducing software over-provisioning

Background

The client is a technical services company providing transaction processing services to global clients and looking for opportunities to reduce costs as they implement an updated infrastructure.  The primary purpose of the engagement was to evaluate how their existing server refresh policy might be modified to achieve that goal.

Environment

The client operates three primary and two secondary datacenters across national boundaries. Service offerings are real-time transaction processing and data analytics. Databases are relational and organized as scale-out clusters or running within virtualized servers. We worked with the client to synthesize their data into the following key metrics:

  • The cost structure for the software stack and server environment, presented as a ratio of software-to-server costs, is 9.6. Included software: OS, virtualization middleware, monitoring, and database. The licensing terms included a mixture of subscription and perpetual licenses.
  • Overall server utilization rates were not a factor in the inventory analysis.
  • Client estimated future demand growth rate was 15-20%.
  • Server refresh policy was based on four-years.

Inventory Analysis

The primary objective of the inventory analysis was to evaluate a more flexible, cost-effective, and variable server refresh policy in light of the heavy reliance on subscription licensing. The original purpose of subscription licensing was to provide flexibility to reduce license counts if clients were lost or overall market conditions reduced transaction processing levels. However, this style of licensing also provided the ability to immediately reduce license costs by dynamically adjusting server refresh times and rates to more quickly uptake new, better performance servers.

Ravello simulated a wide variety of use-cases. The simulations involved a range of assumptions regarding the client’s environment (e.g., existing inventory, capacity, cost structure, etc.) and expected future server configurations and relative performance gains. The simulations produced cost-optimized evolutions of future inventory transaction (variable refresh transactions), ongoing inventory balances, and how those levels drove software spending.

Results

For this client’s environment and cost structure, the lowest total cost for both software licenses and faster server turnover (refresh) was projected to be, on average, less than 24 months. The total costs savings, including lower software license fees but higher server refresh rates, was expected in the range of 30-60%, dependent upon actual realized server performance gains. It is important to note, average server turnover of 24 months does not mean all servers are therefore refreshed on a fixed-time schedule of 24 months.