Anyone who manages the technology demands of a large business faces plenty of exciting moments on the job, but I think it’s safe to say that calculating the energy costs of your company’s IT systems isn’t among those moments.
I was reminded of just how hard it is to factor energy efficiency into purchase and configuration decisions while reading some recent claims in the media around the cloud, and I remembered some simple but often overlooked ways mainframes solve tough energy efficiency dilemmas.
The Power of One
A device that can handle more data with fewer resources sounds like the definition of efficiency to me. No matter how much power it may have, a cluster of servers is still comprised of multiple devices, and every device involved in a clustered system multiplies issues of space, heat production, and power requirements. With up to 141 configurable processor units and 10TB of memory in a single machine, current mainframes offer comparable power to a large cluster of x86-based servers while saving floor space and energy output. That’s important for organizations that are looking to reduce their carbon or physical footprint or meet energy efficiency thresholds or capacity limits.
Limits of Capacity
One of the most energy-efficient aspects of mainframes is rooted in the system’s design. From their inception, mainframes have had some of the highest resource utilization rates of any hardware, often exceeding 95%. Many other systems are designed to run at 70% capacity or less in order to allow for system-originated maintenance, clean up, and checkpoints. If a hefty percentage of a system’s capacity is always busy processing self-generated tasks, then those throughput figures don’t really contribute to efficiency, do they?
When Less Is More
Think about a car engine. Not every cylinder is firing every time you press on the accelerator. If this were the case, the concept of fuel efficiency would be non-existent (and gas would likely be even more expensive.) Some engines even use a concept called variable displacement, which can dynamically shut off a cylinder or two to optimize energy production. Now, what type of computing device is most similar to a variable displacement engine? That would be the mainframe. The processing demands on any computer shift moment by moment, and mainframes are designed to easily shut down some processors when load is not present.
Computing the Cost
Too often, business environments demand short-term successes, which result in short-term decision-making. A classic example is considering the cost of acquisition rather than the cost of ownership in hardware and software. While one system may cost significantly less to buy and configure, there are significant costs that can pile up over six months – including electrical usage and heating/cooling. Figures from manufacturing promise significant savings over the lifetime of ownership. I’ve even heard of stories where due to power capacity limitations, like inside of the Washington D.C. beltway, the only computer resources that could be added were mainframes.
Using Hardware Well
In comparing the efficiency of computing systems, a vital question is often overlooked: How effectively does software utilize the hardware? We’ve all experienced problems with applications that run poorly on non-native systems. Whether or not a piece of software can perform as intended, as well as use all of the available processing power, can have a huge impact on efficiency. In the case of mainframes, the hardware/software match is often a best-case scenario. Applications and operating systems that were designed prior to recent leaps in memory, I/O and processing are able to take advantage of these advances without some of the inefficiencies that non-native hardware/software pairings can introduce. That has a direct effect on electrical usage and efficiency.
People Power
We’ve been focusing on the efficiency of processors and cooling systems, but what about the human factor? How system administrators use their time is an important part of the energy efficiency equation. Once again, mainframes make a difference. Multiple smaller systems take more time to manage than fewer large ones. This may seem at first like a small point, but, like other issues explored here, the long tail effect can be significant. Consider that multiple smaller systems can each have multiple differences in configuration and more. Multiple small issues have a nasty habit of turning into bigger ones.
It goes without saying that energy efficiency is essential to a company’s success. But I’ve witnessed too many situations where a drive for greater efficiency occurs without considering the longer view or subtle details. Those who do take a full look at their options, however, may be well served by the impact of Big Iron.
Best Microsoft MCTS Certification, Microsoft MCITP Training at certkingdom.com