Uncovered capacity saves an online services provider $125 million and positions it for future growth.
A fast-growing global online services provider found itself in a data capacity crunch much sooner than anticipated. The 88MW of critical data on 100,000 servers around the world was forecast to double over the next three years. The tech team believed there had to be a better way to add capacity, especially given that internal analytics calculated average CPU utilization at less than 15 percent. The company called McKinsey to help evaluate its infrastructure and data options, and find a way to make future data growth more cost effective.
The McKinsey team of enterprise services specialists first developed comprehensive baselines of the client’s storage costs and assets. The team detailed server counts, facility counts and sizes, server locations, utilization levels of servers and facilities, as well as power usage effectiveness (PUE) measurements of existing facilities. A simultaneous effort documented asset financials, including server capital expenditure (CAPEX), facility CAPEX, server labor support, facility labor support, and energy costs.
After completing the baselines, the team tackled the client's central issue: identifying and testing more cost-effective data storage options. It did so using two usage models enhanced by McKinsey industry analytics:
- a demand forecast model to test out several data growth scenarios
- a financial model that tested the impact of using next-generation modular facility technology
The models shed new light on the client's actual situation and showed that the company had some breathing room—and more capacity at its disposal—if it made several changes that would both optimize its data storage utilization and drive down operating costs.
The client was able to capture $125 million in savings—and reduce its original budget by more than 10 percent—through:
- using modular facilities to preserve capital through incremental builds
- replacing expensive tier III capacity with lower-tier capacity for noncritical and cloud-enabled applications and decreasing per unit costs by leveraging cheaper materials (e.g., steel vs. concrete) and cheaper cooling
- improving server utilization through cloud computing