Reality today for many midrange-to-enterprise level IT shops is a “one of everything” environment that demands lots of individual attention, and, in increasing numbers of cases, a data center that is running low on power, cooling, and space. Server virtualization can help some by extracting operational management of individual servers to an overall software level, but after a decade of virtualization, only about 60 percent of servers are virtualized according to multiple surveys. And virtualization has its own complexities and adds processing overhead, which only increases the power and cooling load. It cannot help data centers that are running low on resources.
CIOs seeking a better answer to the problem of server and storage sprawl or a path to improving operational efficiency should take a hard look at an option they may not have considered — a mainframe. Midrange organizations in particular often see mainframes, if they think of them at all, as dinosaurs — huge, complex, expensive, using esoteric technologies and water cooling. But the dinosaurs evolved into birds, and today’s IBM System z or Unisys Forward! And ClearPath are not your father’s mainframes.
An entry level system is about the size and cost of a high-end server ($89,000 for Forward!, $75,000 for System z). But unlike that server, the mainframe comes fully integrated with its storage (flash and/or disk), and networking, and is fully virtualized.
Actually IBM’s VM operating system was the first virtualized solution on the market, going back to the 1980s when the term “virtual machine” was virtually unknown. The big advantage for users — while servers are designed to run one application, mainframes are designed to run multiple compute loads simultaneously, allocating resources dynamically as demands change to meet each application’s service level specifications.
- Fully automated
All that happens automatically, with no need for human intervention under normal conditions. And over that, these system run industry standard Linux, the same Linux that most shops run today with the addition of a few commands specific to the underlying platform. That means that ITOs can port their existing Linux software and run it with no customization. IBM mainframes simultaneously run zOS, which runs DB2 and other often shared utilities with a UI that makes operations transparent. Unisys Forward! also runs standard Windows, and Windows can also be run on a VM of IBM System z.
Management of the entire system is fully automated. Mainframes have been known to operate dependably and efficiently for years with little or no human attention. And because operational efficiency and high utilization have always been mainframe issues, going back to the 1950s, today’s systems are designed from ground up for fully automated load optimization and maximum productivity.
And mainframes are designed to be hugely scalable. This means that a midrange organization can bring in a low-end mainframe to replace two or three servers. Over time it can grow that box into a fully integrated, virtualized data center handling most of its in-house compute and database load while maintaining that fully integrated, highly automated environment. Or it can bring in a larger system and basically forklift much of its data center and realize the benefits of a fully converged, virtualized environment that was designed from the ground up as a single system rather than kitted together from an amorphous collection of hardware.
The advantages start with built-in instant fail-over, dynamic resource allocation, virtualization across compute, storage and networking out-of-the-box, and of course a single vendor and service contract covering the entire environment, which greatly simplifies contract management.
Today’s mainframes also are designed for the realities of today’s data centers. Except for the highest end systems, water cooling is a thing of the past. And while a raised floor is nice and can make cabling easier, it is not required.
Server consolidation in a weekend
So what does this mean in reality? Speaking at the recent IBM Enterprise 2013 conference, Kevin Barber, assistant director of the MTM Center at the University of Arizona’s School of Pharmacy, described how the school migrated its full environment, including the management of outcomes for a large patient population, to a System z box over a weekend. Basically the mainframe was installed on Friday, the software was all loaded and tested on Saturday, and the entire environment went live on Sunday with no hickups or problems. Today it is being run by a single Linux programmer on a part-time basis.
Obviously not all compute loads are suited for a mainframe environment. Large, high performance transactional environments are best kept on individual, unvirtualized servers with lots of flash storage. And Big Data analysis involving large quantities of social media or other public cloud data is best done as close to the data as possible since moving terabytes across the Internet is costly and time consuming. However, once the first pass through that data — for instance to extract any comments about an individual company or its products — is complete, it may be desirable to move that smaller data set to the mainframe to combine it with internal sales and marketing or other data, for instance to analyze the effectiveness of a marketing campaign or customer response to a new product.
On the other hand, use cases for a mainframe are many and easy to identify. They certainly start with data centers that are reaching capacity — a forklift migration to a mainframe that replaces multiple servers and cuts both demands on floorspace and on power and cooling is much less expensive and complex that building a new data center and relocating existing operations. It also is a good option to consider to support growing operations. A mainframe can also be an option for supporting expansion into new geographies — basically a company can clone its mainframe environment and drop it into the new territory as a data-center-in-box to support all business needs there while providing a full remote backup for the home data center.
IBM is, of course, the dominant player in the mainframe market worldwide, and todsy’s System z is a solid, advanced architectre. However, it is not the only choice. Unisys Forward! Is certainly competitive, and CIOs exploring a mainframe option would be wise to talk to both, to find what each offers for the organization’s individual needs and to gain negotiating leverage. Having your IBM coffee mug on your desk when the Unisys sales team comes to call and your Unisys mug out when IBM drops in, and whenever possible timing purchases for the end of quarters, when the sales teams are pushing to book as much business as possible, can only help you to get a better deal.