The zEnterprise BladeCenter Extension (zBX) is central to consolidating and streamlining your data center infrastructure. It also is the key to IBM’s hybrid computing concept. In the z196 or z114 you can mix z/OS and Linux workloads, run Cognos and WebSphere, but it is not until you add in the zBX can you truly run hybrid workloads including all the above and Windows and UNIX/AIX. Then you can pack a multi-platform IT infrastructure in two boxes tightly connected by a private 10Gb network and manage it as a single system.
That is a powerful idea. Data center managers are attracted to hybrid computing for the efficiency it brings their heterogeneous infrastructure. Figuring out if it is right for their organization, however, gets tricky. IBM doesn’t readily make available information about use cases and deployment models. In addition, the costs surrounding the zBX remain something of a mystery to z data center managers, which discourages adoption.
For instance, a long-time mainframe enterprise data center manager told Dancingdinosaur: “I assume the cost of System p server capacity is approximately the same, regardless of whether server capacity is purchased in terms of P7 cores for a rack-mount server or for a blade server, but I’ve not been exposed to blade and rack pricing charts.”
IBM insists that it is pricing the blades as commodities, competitive with comparable blades from other vendors. OK, maybe you can figure a P7 blade will run you around $6000 depending on how it is configured.
But you still don’t know the price of the zBX itself. The zBX is not just any commodity blade cabinet. It includes sophisticated networking, including a built-in switch. The data center manager made his best guess: “I’ve not been provided with detailed costs from IBM, but I suspect that a simple two-blade starter system would cost about $350,000 for a zBX frame.” Good guess but high. A data center manager who ordered two zBX cabinets told Dancingdinosaur the cost for each was somewhat above $200,000. At $200k the insurance company, which also runs AIX and Windows workloads, might have considered a zBX.
IBM has never liked to talk about prices for a number of valid reasons. They operate in many different countries where currency plays a role, and the company often sells through partners who need some pricing latitude. Still, when it comes to zBX adoption, this reticence is hampering adoption.
When IBM does talk about mainframe pricing it prefers to couch the conversation in terms of the cost per workload or TCO. The zEnterprise can handle many more workloads than its distributed system rivals and even IBM’s own distributed Systems p and x. By amortizing the number of workloads over the cost of the zEnterprise, including the zBX, blades, software, management, middleware, and integration the cost per workload is not just competitive but often beats its competitors.
John Shedletsky, IBM VP/Competitive Technology, does exactly this in a presentation called zEnterprise Economics. You can find it here. This is an excellent presentation and comprehensive in the scope of its analysis. Also, it is very hard to quibble with his numbers or math. The only problem: data center managers cannot readily break down Shedletsky’s numbers and apply them to their own situation and workloads. At the end of the presentation, a data center manager still can’t tell what it will take to buy a zBX and two blades to run the company’s workloads.
As it turns out, a midsize mainframe shop that just wants to run one or two blades probably shouldn’t be looking at the zBX anyway. As one IBM staffer told Dancingdinosaur, “Ideally the organization wants to deploy sufficient numbers of blades to have a major impact.” He recommended companies plan to deploy a minimum of three or four blades although amortizing 8-12 blades across the cost of a zBX would be ideal.
Then there is the question of which workloads. IBM has said it does not expect organizations to virtualize and consolidate all their Windows applications on the zBX. Rather, they should choose those workloads that most benefit from the attributes of the zEnterprise. These, most likely, will be two-tier applications with either the data or some application logic already residing on the zEnterprise. That could be a Windows application that frequently accesses data from DB2 or needs to call critical mainframe application logic through CICS.
IBM offers its tuned-to-the-task methodology to help in selecting workloads. More helpful in sealing a deal would be some realistic use cases or model workload deployment scenarios. In subsequent pieces Dancingdinosaur will delve into appropriate use cases and deployment models for the zEnterprise and zBX.