From NASA retiring its 45,000 sq ft data center to the U.S. Federal government saving more than $2.8 billion, many of the biggest data center owners in the world are making the long-term effort to scale their data centers more efficiently by consolidating the current hardware on an ongoing basis. Trying to figure out if there is more “juice” in a densely deployed virtual environment might be one of the most challenging but also expected progression of IT operations. Their increasing demand for more efficient computing power and storage space makes the growth of a scalable and elastic data center a necessity.
“Can Turbonomic tell me when to purchase new hardware, both in the near and far future?”
A data center consolidation project - The Demand for Efficiency
One of my Turbonomic customers - let’s call them Company A - is undergoing a data center consolidation project. With thousands of VM (virtual machines) deployed per hypervisor and an annual VM growth rate of over 20%, consolidation projects have been deemed a high-risk goal at Company A, since the performance impact after hardware consolidation has always been a major uncertain factor whenever Company A made any technical and business decisions for their mission-critical corporate IT operations. More rigorous SLA requests and trickier use cases fly around the help desk at Company A all the time, and with major resources invested in developing automation and private cloud, Company A has created a successful global IT system centered around self-service platforms and one-step workflows. However, when we first discussed our expected use cases with Company A, one of the first questions we were asked was, “Can Turbonomic tell me when to purchase new hardware, both in the near and far future?”
A consolidated data center that guarantees application performance
Being a new but small piece of software in their complex IT operation, our discussion with Company A went on for a while before Turbonomic captured a realistic set of goals for their next round of data center consolidation project. We took our time to explain and define the elusive desired state of the application workload and virtual infrastructure with Company A. Was it 70% utilization across ESXi physical memory? How about 65%? Or 85%? When Company A started to brainstorm numbers like a guessing game, we had to show them that one of their many vCenter servers observed a physical memory utilization change from an average 70% over the weekend to a peak of 95% the following Monday morning. Just like real-time control of performance, we can never rely on a simple number to be the target of consolidation. Instead, the target has to respect the changing demand of the workload, because a consolidated data center is worthless if it is not capable of running applications with guaranteed performance.
“The key to successfully executing a data center consolidation plan is a shift in the IT organization’s mode of operation.”
Consolidating your data center comes with a different mode of operations
After running some offline projections of saving hardware and merging clusters in the Turbonomic lab, we were able to determine the overall feasibility of proceeding with a data center consolidation plan for Company A. Proving out the feasibility of improved efficiency on a hardware level is just the beginning. The key to successfully executing a data center consolidation plan is a shift in the IT organization’s mode of operation. From the moment we enabled automation, we have seen hundreds of actions executed hourly that confirmed our initial estimation of consolidation feasibility. Our risks avoided dashboard/report captures the nature of these actions. This categorized view was quite eye opening to the IT organization’s management team.
Number of risks avoided by Turbonomic on 1 cluster in 48 hours
We then mapped out the types of applications and constraints that Turbonomic has to follow. With over 10 years’ worth of data collected from both the legacy and new hardware, we were able to define a clear set of requirements to save on hardware purchasing, track OpEx savings, and contribute to Company A’s chargeback calculation when deploying new VM.
This close collaboration with Company A will not only result in gained efficiency for their IT organization, but also establishes an entirely innovative way of IT operations, which we at Turbonomic call Future Mode of Operations.
Will the consolidation plan proceed as smoothly as we imagined? I cannot wait to find out as our exciting journey with Company A continues.