Q - Joe, cloud computing has been characterized in many ways, but is basically a new service consumption and delivery model inspired by consumer Internet services. Key components include on-demand self service, ubiquitous network access, location independent resource pooling and rapid elasticity. In recent weeks you have been advocating the mainframe for management of the cloud. Why?
A - Simple ─ a mainframe is a cloud-in-a-box.
Consider this: mainframe users want resources on-demand. Mainframes are designed to be able to provide services on-demand ─ based upon pre-established prioritization. If you want additional resources on a mainframe, you can ask IT to provide them and voila ─ if you have priority, you get them. Over time, I expect IBM to front-end mainframes with software that will give users more control when requesting resources ─ but the fact remains that mainframes can be easily provisioned to meet user requests for resources today.
Ubiquitous network access is another of your criteria. Mainframes can be connected using TCP/IP to the Internet. The Internet is ubiquitous. Hence, you get that with a mainframe.
Location independent resource pooling (in other words, the resources show up in a pool ─ and no one cares where those resources are ─ all they care about is getting access to those resources. Mainframes offer the market’s most advanced resource management, virtualization, and provisioning. Hence, they meet this location independent resource pooling descriptor.
Finally, elasticity (expandability, flexibility, etc.). Name a more scalable, elastic, flexible environment when it comes to making resources available. Weigh mobile partitioning, advanced virtualization management, and virtualized resource management as you seek to find a superior architecture. You can’t find any general workload processor more flexible and easier to manage (considering how many resources that it controls) than a mainframe.
Q- In your experience, have customers who have adopted service management from system z realized measurable ROI? Can you share any examples?
A -Most customers that I talk to don’t sit around and say “this is how much it cost me to run back-up/restore and other functions that I manage using service management software. What they say is “hey, look over there. Those are my five mainframe managers who manage the equivalent of hundreds of x86 servers. The math is pretty obvious ─ you need a dozen or a couple of dozen managers and administrators to manage x86 environments ─ and far, far fewer managers/adminis-trators to manage a mainframe. (And remember ─ in some geographies, the salaries, benefits, and sick leave for x86 managers can cost a company $100,000 per year. Take ten or twenty IT managers out of the equation and pretty soon you’re talking about real money…)
In Brazil, when I asked a mainframe user who manages his mainframe ─ he said “me and that other guy over there”. He was managing his company’s SAP environment with 2 people. That’s mission critical computing ─ the company’s livelihood ─ with two people.
In Arkansas, another mainframe manager said “I’ve replaced dozens of x86 servers ─ and will replace dozens more ─ with our mainframe. Here are the five offices of the people I use to manage this environment. People don’t believe me when I tell them this ─ but its true!”
Incidentally, mainframe managers have been using dashboards with monitor, visualization, and control facilities for years. They don’t even think of it as service management software. But it is.
Q - Not every customer has a mainframe. At what point(s) should a business consider adopting the platform for service management?
A -Not every customer has a mainframe ─ and not every customer needs a mainframe. It you’ve got a ton of applications and you want to run on the most efficient general workload processor ─ you should be considering using z/OS, zVM, and/or Linux on a mainframe. If you don’t have enough work for a mainframe ─ but want to run high-RAS (reliability, availability, security) applications in a mission critical environment ─ use IBM POWER systems (my opinion: SPARC and Itanium are dead-end architectures). If you want to run Windows on Linux on x86 Xeon processors ─ you’re going to deal with a bunch of distinct servers that will need to be administrated and managed. And you’re gonna need a ton of people to manage those discrete servers. To reduce the need to have to use as many administrators and managers (and to save associated salaries/benefits), as well as to reduce costs related to human error that cause downtime and potential loss of business, you need to adopt service management software.
Bottom line: use a mainframe if you have a lot of work to do ─ and you don’t want to spend a ton of money managing that environment. Buy a Unix server if you don’t have enough work to do to keep a mainframe busy. Buy an x86 server if you need Windows ─ or if you want to run Linux on x86 platforms. In each case though, use service management software to reduce you management costs.
IBM is a client of Clabby Analytics and has provided compensation to Clabby Analytics for participation in this interview.