Occasionally throughout the summer I’ve been chatting and emailing with Saugatuck Technology analyst Charlie Burns about mainframes, IBM and Linux. Many people have argued over the past year that the mainframe is dying out (again), but Burns and some very telling market trends go against that grain with a 180 degree turn: the mainframe is surging, and it’s all thanks to Linux.

I’ll have an article up a bit later this week (or early next) detailing just exactly what is going on in this space, but for now I thought I’d include one of the recent emails Charlie sent me that covers some of the basic cost advantages of the mainframe.

Architecturally-based advantages in the hardware, the operating systems, and in the virtualization functionality enable mainframes to manage multiple diverse workloads based on business objectives and deliver exceptional cost reductions. If we compare the costs of using mainframes to those of conventional servers as noted earlier, we find the following:

• Technical support and maintenance costs. By consolidating and centralizing the capabilities of dozens of servers into a single platform, use of a mainframe drastically reduces the redundancies and differences that are de rigueur in server farm environments. If we accept conventional industry wisdom that states a minimum of 70 percent of IT costs are labor - and that the majority of labor costs are training and support - it’s easy to see how mainframes can quickly free up IT budgets for more strategic investment such as new application development.
• Software licensing and maintenance costs. Since most operating, middleware, and application software is licensed to each server it is used on, a mainframe offers substantial software savings. In a mainframe, the computing capacity applied to software can scale dramatically. Literally, hundreds of virtualized server images can operate in a single mainframe under a single license, thus, avoiding additional license and maintenance fees. In addition, the IBM System z has the capability of running specialized processors for Linux and for some application workloads. These processors are priced substantially lower than the base processors. Thus, the System z delivers both hardware and software saving on a broad scale when compared to individual x86 server platforms.
• User and IT training costs. Training costs tend to be driven by the number and complexities of multiple applications and operating systems. By enabling the use of all leading operating systems and applications within single platform, mainframes drastically reduce the need for training.
• Utility and environmental costs. Mainframes require substantially smaller amounts of power, UPS capacity, cooling, and floor space when compared to the environmental requirements of an x86 server farm with equivalent processing capacity. The mainframe’s advantage is even more substantial when one considers the reduced amount of storage and inter-connection equipment compared to an x86 server farm.
• Security costs. Mainframes enable centralization of software and application interfaces. Centralization of software enables vastly improved security management by reducing the number and types of access points. Additionally, because of its heritage, security is architected into the mainframe and is uniquely robust. For example the IBM System z family of mainframes provides security against information flow between virtual machines. The System z was first certified in mid-2003 as Evaluation Assurance Level 5 (EAL 5) by meeting the Common Criteria standard ISO 15408. Comparatively, virtualization on x86 server platforms require security to be added and layered as part of the operating system, applications, databases, and so on – further increasing both the complexity and cost of security, while adding more points of vulnerability due to incompatibilities between security systems and other software.

An Elementary Roadmap
Saugatuck recommends that every company with more that 20 x86 servers should perform a thorough evaluation of existing workloads and servers with the following steps in mind:

1. x86 servers yielding the largest savings should be migrated to the mainframe first (e.g., those with unique infrastructure support requirements)
2. x86 servers with the lowest utilization should be migrated early
3. Assets with an upcoming compelling event (e.g., need for capacity upgrade, lease expiration, etc.) should be migrated before incurring the expense
4. x86 servers/workloads should be aggregated by user department to leverage strong buy-in
5. Oldest technology x86 servers should be migrated early
6. Focus on real estate by freeing up contiguous raised floor space or eliminating sites as early as possible

Today SearchEnterpriseLinux.com ran an article on AMD, Novell and power management, and how the overall Linux community has warmed to the idea of “Green Linux.” Our sister site SearchDataCenter.com has, in my opinion, led the way on power and cooling from a hardware perspective, but today we’re starting to see some of the responsibility on this important issue shift to the software guys too.

Our article today (Novell, AMD synch up on power management) dealt first with AMD and its initiative to shift PowerNow! to Linux systems. Much of the work, said AMD’s director of commercial applications Margaret Lewis, is centered on incorporating AMD’s PowerNow! power management into SLES 10 SP1 and SLES 9 SP4. PowerNow! is a dashboard-based application that conserves power in AMD-based computers. The CPU’s clock speed and VCore is automatically decreased when the computer is under low load or idle to save power and reduce heat, Lewis told me this week.

The AMD news was, obviously, hardware-related, but later in the article we get to the Linux Foundation, and what that group is doing to raise power management awareness with Linux developers. According to the foundation, making Linux green is becoming more important in all aspects of Linux adoption, from desktop to server. At the recent Linux Foundation Collaboration Summit at Google’s Mountain View campus developers said will work to identify key projects and coordinate resources among its members to improve power management functionality in the Linux platform. There’s another meeting scheduled for this week in Ottawa, so we’ll just have to see how fine tuned things become.

Getting Linux, power management, and the data center to coalesce at a single point will be quite the coup if someone can pull it off. Right now, I think each of the factions (hardware, software, etc) are off doing their own thing. Regardless, it’s great to see Linux developers and advocates getting out in front of the issue.

UPDATE, June 26@1:30 pm.: For some more on the next generation data center and a place where many a Linux admin might be working very soon, check out Bridget Botelho’s scoop on Google’s brand new Iowa-based data center.