IBM Pushes Mainframe Performance Envelope

The fastest machines in the data center just got a whole lot faster. IBM today announced the zEC12, a mainframe based on 5.5 GHz processors built using the latest 22-nanometer technology.

According to Doug Balog, general manager of System z, the zEC12 now has 101 configurable cores that are about 25 percent faster than processors used in the first-generation zEnterprise mainframes that IBM launched in 2010.

In addition, the new mainframes feature Flash Express solid-state memory technology that Balog says not only significantly increases performance, but also helps reduce the overall size of the mainframe footprint inside the data center. Overall, Balog says this latest version of the mainframe increases system capacity by 50 percent.

Other additions to the mainframe lineup include a configurable cryptographic co-processor, called Crypto Express4S, which offloads the processing of, for example, encryption or digital signatures, from the main processors.

Finally, IBM also announced that it is upgrading the enclosure of the zEnterprise Bladecenter Extension modules that allow Intel Xeon or IBM Power Series servers to be tightly coupled to the mainframe to accommodate future Xeon and Power Series processors that Balog says will require support for different types of blade server configurations.

Since revitalizing the IBM mainframe in 2010, IBM has been driving an effort to converge the management of workloads across mainframes and distributed computing systems. While that effort remains a long-term work in progress, interest in high-performance mainframes has increased steadily as the number of transactions and the amount of data that needs to be processed continues to escalate.

The next challenge, says Balog, is to make mainframes a more transparent part of a cloud computing ecosystem by extending the self-service management capabilities of IBM Tivoli systems management software out to the mainframe. Once that occurs, the mainframe should appear to be just another available resource in the cloud.

In the meantime, the mainframe remains the most scalable resource in the data center that, from a total cost of ownership perspective says Balog, is still more efficient than trying to manage an equivalent number of densely packed x86 servers that increase management costs with every node that gets added to the cluster.