Administrator > Performance

Performance/The Mainframe Energy Efficiency Value Proposition

Visualize this: a single, large, scaled-up system that shares communications resources, memory and I/O facilities driven by a single (yet redundant) power subsystem. Now contrast this with dozens or hundreds of distributed, scaled-out systems each with its own power supply, memory, CPU, I/O and network interface facilities. And remember this: these scale-out distributed systems need external hubs, switches and routers in order to communicate between each cooperative system within a distributed, networked environment (these also add additional acquisition cost and require additional power). Which approach do you think is inherently more efficient?

If you chose the large, scaled-up system environment, you’re right! By building a scaled-up environment that operates on the principle of sharing resources, mainframe design trumps distributed computing architecture when it comes to energy consumption.

From an energy consumption perspective, mainframes have one of the most (if not the most) compelling value propositions in the computing industry especially when compared to distributed system-computing approaches. This claim can be made based upon three criteria:

• Mainframe (scale-up) system design advantages
• Component efficiency
• Level of sophistication of mainframe virtualization

Each of these aspects deserves closer scrutiny.

Mainframe System Design Advantages

When it comes to energy efficiency, one can argue that a mainframe (scale-up) system’s design is superior to that of a distributed system’s design for several reasons including average utilization, internal networking and the sharing of components.

First, scale-up designs waste less energy by virtue of being better utilized. Average utilization rates for mainframes run in the 75- to 90-percent range; whereas utilization rates for distributed, scaled-out “tower” application and database servers frequently run at less than 20 percent. What this means is that mainframes process workloads more efficiently and use less power doing so than distributed, underutilized server environments.

The second major energy-saver with respect to scale-up systems is internal networking. By using internal networks to communicate between processors, the external, power-consuming devices used in distributed systems architecture can be eliminated. What this means is that IT buyers need not waste money on loads of network interface cards, hubs, switches and the like, nor do they have to power each of these devices separately. Depending on the size of a given distributed systems environment, using a mainframe or scaled-up environment can save hundreds of thousands of dollars annually in terms of energy usage, equipment acquisition and maintenance.

The third major energy saver is the design emphasis around sharing components such as memory cache and I/O pipes. Instead of multiple components devoted to single machines (as in scale-out architecture), components are heavily shared in a scale-up architecture. This sharing of resources results in less energy being devoted to powering underutilized memory, disk and I/O than to fuel distributed computing environments.

Next page: >>