With so much POWER horsepower, why do we turn to x86-64 for non-core business workloads? Maybe we need to change our thinking of what core business workloads are and then put them on a core business server.
Being almost four years into a Power 720 Express, it's just about upgrade time in my shop. Or at least time to get the ball rolling into looking at some new hardware.
With the new PureFlex for IBM i solution I covered in the last Wheelhouse as a potential opportunity because we have a number of System x servers in our shop, my thoughts are turning to one of my favorite topics: consolidation.
We have a couple of older x3550s with VMware and a total of maybe 12 or 13 virtual machines that were put in place over time by departmental Windows advocates who liked to ramp up a virtual server when a simple IBM i-based service would've done just fine. Those certainly play a part in our decisions going forward.
While we could just upgrade those machines and continue down the path of running separate hardware, the ability to run PureFlex is certainly appealing. Cost will always be a factor, and I'll let you know how that investigation goes.
With regards to consolidation, I like to squeeze as much horsepower as I can out of existing investments. Coming from the IBM i world, we're all aware that we can run code written many, many years ago on modern hardware without a recompile. The investment protection is built in. But what about new investments? How much are we as a community exploiting IBM i technology in order to get as much bang for the buck every three to five years when we upgrade? Do we see IBM i as the consolidation platform that it really is and has been for many years? I'd hope so.
But the fact of the matter is that while you see high-end machines on display in customer testimonials, major announcements, and press releases, most companies that run IBM i are the small-to-medium-sized businesses (SMBs) with an entry-level machine and a couple of POWER6 or POWER7 cores. They're a trucking company with 400 users or a transmission parts manufacturer with 40 users. They have some Windows servers and maybe some Linux servers, but IBM i runs the core business. That core gets the investment protection.
One of my goals as an IT manager is to ensure as much of the infrastructure gets the kind of IBM i investment protection as possible. That means I need to run as much load as I can on IBM i and POWER technology rather than on x86-64. While I'll probably always have Windows running in my shop, and the jury is still out if my amount of Windows virtual machines will justify a small PureFlex investment for my shop, the question I'm faced with is, what Windows servers can I migrate to IBM i to achieve the same ease of management, security, scalability, and simplicity that my core business has?
First, what is the core business? ERP? Sales and order management? Warehousing? Or do components that get bolted on after the fact become part of that core? I think they do. Email is hardly thought of as part of the core business. Neither is instant messaging. Or mobility. But in fact they really are. If we shut down our email, chat, or mobility services for an hour, we can count how many people are looking for our heads on big, pointed sticks.
Unless we think of these services as core business processes, then they'll continue to run on second-thought x86-64 servers.
Some are a no-brainer, such as IBM Notes Traveler (formerly Lotus Notes Traveler). I could have kept a Windows virtual machine with a Domino server and IBM Notes Traveler around and working just fine. I could have kept the separate backup process so that my Traveler databases and server would be recoverable in case of a disaster. I could have kept going with Patch Tuesday twice a month, ensuring Windows servers were up to date so that they'd be protected from vulnerability exploit attempts that were sure to follow. And on and on and on.
No. Putting that one service on IBM i means that I never have to think about it. Set it and forget it. That's the integration value of IBM i. For every solution moved to IBM I, the backup is integrated. We give that solution the value of single-level storage and 128-bit addressable space. We give it the bulletproof security. We give it the management simplicity. The solution gets that by default just by consolidating the workload on our core business server.
For years, we ran on a single core. Until now. In September, I finally loaded that single POWER7 core with enough workload to justify a second IBM i license in order to turn on another core. The benefits of loading up my single IBM i partition is well worth the cost of another core. IBM Connections put me over the limit.
As a Lotus Quickr customer with an existing support contract, we were entitled to the full IBM Connections set of products. So instead of loading up a good few virtual machines on x86-64, I figured I had just enough horsepower to put it on IBM i and Power Systems. Just enough...and users finally felt a new workload when it ran. The point is that I could've tuned up a number of virtual machines to do the job, but loading it on IBM i was simply the best move to make. Once we added the second core, the thing just smoked. In fact, everything just smoked. More importantly, Connections, as a continuation of what we started with Quickr, will eventually be a core business workload. It needs to go on IBM i with the rest of them.
What if some solutions out there aren't supported on IBM i but would be great to move to IBM i? I'm a big believer in Samba and what it can offer my company. Not just for file sharing, but as an eventual replacement for Microsoft Active Directory in order to push out group policies to domain members as a primary domain controller, plus serving up Kerberos and LDAP.
Group policy. Domain controller. LDAP. Kerberos. File shares. Eliminate one of those and see how well the business runs. They need to be thought of as core business processes, not just as add-ons or services that augment the core.
While I could run Samba on x86-64 as a Linux virtual machine, I can also deploy Linux on POWER in its own partition with the Hardware Management Console (HMC) or on a guest partition managed by IBM i. Even though I don't have an HMC at the moment, the guest partition can still be quite effective because of the IBM i integration reasons stated earlier. Linux as a guest partition is actually a single IBM i object called a network server storage space. Essentially, you create an object that carves out a section of your disk storage and you build a Linux partition inside it. Linux, while independent of IBM i, takes advantage of virtualized IBM i objects (think tape drive, DVD drive). It can be backed up and recovered using standard IBM i save/restore commands. While Linux just sees a single volume (or multiple volumes, depending on how many network server storage spaces you allocate to it), it will take advantage of the collection of disk drives you have, resulting in optimum performance.
Since most SMB customers have cores to spare on their entry-level Power servers, why not take advantage of those cores instead of spinning up a new VM under x86-64? They're just sitting there. To me, taking advantage of those cores means fully exploiting the investment we made in IBM Power Systems.