Typically, the IT services buyer forecasts usage utilizing a number of factors at his/her disposal:
- Prior contracted capacity—that is, what was signed up for in the past.
- Forecasted business objectives, along with the inherited success criteria for objective completion, that will dictate IT demand—what will IT infrastructure needs look like over the upcoming quarters relative to in-budget and market-responsive projects.
- Risk tolerance by business line that dictate Service Levels which are acceptable in fulfillment of those aforementioned objectives.
A challenge exists in obtaining the right amount of IT capacity so there is minimum under-utilization (resulting in waste) and at the same time insuring that service levels can be achieved, especially given different business and service objectives.
It is important to consider and recognize the following differences in IT infrastructure demand versus supply as we see how a solution to efficiency and waste measurement of compute capacity based on the right set of metrics immediately adds value to the buyer of compute services/capacity:
- Contracted Capacity = Theoretical Maximum—What is the contracted maximum compute services the buyer has under commitment and SLA the supplier has agreed to provide.
Typically the “contracted capacity” is overspent. In traditional Data Center build-outs, given the lease cycles and infrastructure investment in forecasting real estate, connectivity proximity to systems, required resources for management and budget cycles, the best estimate possible is put forth in forecasting the required demand and then with an excess added to make sure there are no “career limiting” events due to a shortfall in capacity.
- Maximum Capable = MaximumCap ((Network, I/O Throughput, Storage, CPU) * (Concurrent Operational Load Of Software Pipeline Flow Prior To Transaction Failure/Bottleneck))—The “highest capable throughput of workload” that the buyer achieves given bottlenecks that exist within the infrastructure or due to architectural bottlenecks existing within the application layers interoperating within the same infrastructure. (Typically peaks will top out at this point.)
Often systems may be operating with peaks reoccurring at a regular frequency but without a negative event. Many IT organizations may not be truly aware of the architectural bottlenecks that prevent them from realizing the full contracted capacity without the needed introspective review.
All too often, the actual introspective is done around launch of a new service or remediation around a performance issue where measurement was necessary to determine the successful resolution, which leads to the next metric.
Leading IT organizations use analytics to determine the four levels of compute benchmarks described above in order to affect and realize options that are more cost efficient and reduce waste. The waste will be characterized into either Surplus Waste or Operational Waste.
To learn more about how to measure efficiency and waste using a tool-based approach for scenario-based analysis download RampRate’s white paper, IT Efficiency and Waste Quantification – Discussion of Methodology and Tools for Measurement.