When it comes to application workloads, IBM wants to take the debate over the hardware platform off the table.
For more years than anyone cares to remember, the main issue in the data center has been whether to run an application on a mainframe, RISC server or Intel-class processor architecture. As more applications migrated to lower-cost Intel servers, this issue became especially nettlesome for IBM as it saw more application workloads migrating off the mainframe to servers other than IBM’s in the name of application modernization.
While IBM voiciferously argued that the total cost of enterprise computing was less on a mainframe, customers increasingly opted for Intel-class servers because the cost of acquiring a $2,000 to $10,000 Intel server was more attractive than paying hundreds of thousands of dollars for a mainframe.
Today, however, IBM is moving to eliminate this issue all together. The company today not only rolled out the zEnterprise, its fastest and most power-efficient mainframe to date; it also unveiled a new data center architecture that allows systems to be deployed on a private network under a common management framework that spans IBM mainframes, p-Series servers running AIX or Linux, x-Series servers running Windows or Linux, IBM mainframes running Linux or a new generation of dedicated Smart Analytics Optimizer appliances for running analytic and business intelligence applications.
But the zEnterprise mainframe is only the first instance of what IBM refers to as a new Flex architecture, says Ambuj Goyal, General Manager for development and manufacturing in IBM’s Systems and Technology Group. IBM, he says, has already deployed this new architecture for certain customers using an iSeries server as the hub rather than a mainframe, and the company intends to let customers make any server they choose the hub of the Flex architecture.
In addition, Goyal said the Flex architecture will be extended out to not just physical servers, but any combination of virtual servers. IBM is also working toward creating multi-core processors, where several cores in the system will be dedicated to running specialized tasks, such as graphics or encryption.
According to Karl Freund, vice president of strategy and marketing for the System z in IBM Systems Group, this approach means customers can determine what application workloads to run across a 10 Gigabit Ethernet network that is managed by a unified resource manager framework.
This form of data center convergence is driving recent reorganizations within IBM that have brought the software and hardware groups of the company under the same team of executives. The basic idea, says Freund, is that IBM people across the company will now focus more on how to best serve a customer’s application workload, versus competing with one another for the same business.
And because the architecture runs across a private network, performance is significantly enhanced as the need for the number of switches in the data center is sharply reduced, Freund said. Unlike Oracle and Hewlett-Packard, which have essentially mothballed their Sparc and Intel investments, said Freund, IBM continues to invest in multiple processor architectures as part of an effort to optimize application workload performance.
According to Rod Atkins, IBM senior vice president for the Systems and Technology Group, the zEnterprise represents a new dimension in enterprise computing as heterogeneous systems in the data center start to converge around new hybrid architectures.
The only question is whether a Flex architecture is going to be seen as the right answer to how the next generation of the data center should be managed, or something IBM should have done long before customers started migrating away from mainframes and UNIX servers years ago.
