News

The Mainstream -- April 2008 -- Issue 29

How much does it really cost to grow or consolidate workloads on the mainframe versus the equivalent distributed servers?

With the introduction of the IBM System z9 mainframe in 2005, analysis pegged the potential savings at up to 50 percent over other platforms, when calculating all relevant factors, such as labor, software, hardware, maintenance and energy costs.

Three years later, the introduction of the IBM System z10 Enterprise Class mainframe – with enormous price-per-performance benefits – may definitively put the cost question to rest, and help companies redefine their IT business strategy. By migrating key mainframe and distributed workloads to the new generation of Big Iron, users can significantly reduce the total cost of enterprise computing, while gaining the performance, business resilience and security offered by the mainframe.

Bigger Iron
With its release in March, 2008, the System z10 EC becomes the most significant mainframe redesign in a decade, while continuing with the modular design and four- book structure for flexibility and availability. Capacity and performance improvements over the System z9 EC include:

• 50 percent more powerful engines, with 70 percent more system capacity
• Three times more memory (up to 1.5 TB)
• Chip frequency 2.6 times higher than System z9
• Processing power equivalent to nearly 1,500 single core processor x86 servers
• Ability to run at 100 percent utilization rate, for long periods of time, based on demand
• Use of up to 85 percent less floor space and power than equivalent distributed servers and 15 percent less power than a System z9 mainframe

The significant performance leaps can help companies deal with key operational challenges, including the need to manage an exploding amount of data coming through the network – and the complexity caused by a proliferation of individual servers.

Consolidating to System z10 EC
Consolidating many applications onto a smaller number of larger servers has become a popular strategy for reducing server sprawl, and many users choose to consolidate on the mainframe. The performance and capacity of System z10 can provide profound savings over a distributed approach and drive additional savings with the introduction of mainframe specialty engines. The System z10 Integrated Information Processor (zIIP) and the Application Assist Processor (zAAP), redirect workloads, such as certain DB2 and XML cycles and Java cycles, off the mainframe central processor. This strategy saves MIPS and related costs.

Likewise, using a System z10 to consolidate x86 applications to the Integrated Facility for Linux (IFL) engine, can further reduce the price per unit of work, and radically alter cost comparisons. It drives consolidation ratios of up to 30:1 for mainframe versus distributed processors.

For example, users can host up to 760 Linux x86 single core servers on one System z10 mainframe running 26 cores (IFLs), versus about 248 cores when virtualizing on new multi-core x86 servers. Not all servers should be consolidated to the mainframe and individual results may vary. The System z10 scenario also provides room to add 38 more cores, without the need for additional floor space.

The System z10 running 26 IFLs generates about $24,600 in electricity charges per year, compared with $133,000 for the 760 blade-type servers.

Similarly, the System z10 scenario requires adding one to three people to an existing mainframe management staff, compared with about 38 IT professionals needed to manage the 760 unvirtualized x86 servers.

Factoring dividends of upgrading to the System z10 EC
In addition, System z10 performance and capacity increases can deliver other IT dividends to current mainframe users, who can save significantly on software licenses, especially when consolidating z/OS and other types of applications on the platform.

Because distributed software is often priced on a “per processor core” basis, the software and services and support savings can grow dramatically as users replace many cores with fewer IFLs.

Other dividends come from the ability to process about 50 percent more work on each System z10 processor and each IFL specialty engine using the same resources. So when moving from System z9 to System z10, performance gains might allow consolidation of 10 additional processor cores for every 20 x86 distributed servers consolidated on Linux, and eliminate the same number of software service and support charges per year.

System z10 can also lower total cost of ownership with pricing reductions on the monthly license charge software costs and related maintenance charges.

System z9 users can glean up to 5 percent savings on overall software monthly charges and up to 10 percent on related maintenance charges when they upgrade to the same capacity on a System z10. System z990 users can gain a combined discount of about 25 percent in those two areas when they upgrade to the same capacity level on the System z10. Together, over three years, the two pricing discounts may cover up to 70 percent or more of the System z10 server purchase.

Continuing to grow on the System z10 will continue to help reduce cost per unit of work even further with costs usually based on incremental MIPS capacity and incremental pricing for I/O, memory, specialty engines and other resources. Customers who already own zIIP, zAAP and IFL engines can typically roll them up to the System z10 without additional cost.

Ultimately, consolidating on a much more powerful System z10 Enterprise Class can help you strategically reduce your IT budget costs, while providing the security, scaling and integrity you rely on from the mainframe.

For more information:
System z10 and the new enterprise data center on IBM TV
White paper: A fresh look at the mainframe, David Rhoderick