Windows Consolidation with the System i: Is It Happening?
September 11, 2006 Mary Lou Roberts
Earlier this month, Forrester Research issued a brief report, “Windows on IBM System i–Short-List It?” The author, Brad Day, Forrester’s vice president of computing infrastructure, notes the challenges faced by small- and medium-sized shops that often operate two different server environments–one for mission-critical applications and another for standard infrastructure software. While shops may have dedicated servers to support mission-critical applications, “these same customers have also deployed X86-based SMP and blade server designs to run infrastructure software like Microsoft Exchange, Web applications, or file-and-print facilities on Microsoft Windows Server.”
Day points out, however, that these same shops often lack the staff and skills to manage these two environments effectively. “Forrester’s SMB clients point to a number of technology pain points created by this dual infrastructure. . . . In the Windows environment, managing and monitoring multiple Windows servers and administering user IDs and passwords across multiple, distributed computing infrastructures is a complicated and resource-draining task.”
This certainly matches the experience of the OS/400 and i5/OS shops I have talked to. In fact, I’ve often been amused that IT managers at these companies will often not even mention any other servers until asked to be more specific, as if they are embarrassed to admit they have Windows boxes. But when the question of whether or not they have any other platforms installed, they respond, “Oh sure. We have a bunch of Windows servers that we use as print servers and for email and the Web.” When then asked about the size of the IT shop, it’s not unusual to hear, “We have a total of seven people: three on the AS/400, two network people who work on the Intel servers, and two help-desk folks who deal with passwords and email support,” or some similar breakdown. And that’s the side of the staff that’s growing; the business applications staff has been stable (or in some cases, even decreasing) for years.
Forrester asserts that many SMB organizations are not shifting toward a hybrid model with one system that supports Windows-based office productivity and infrastructure software in addition to mission-critical applications, and that the System 15 servers offer a way to achieve this consolidation using i5/OS. “In addition, an IT operations manager can create additional deeper systems integration through the Integrated xSeries Server and adapter technologies for IBM BladeCenter Windows integration and operations.”
The report finds that, “The integration of BladeCenter with System i is one approach in minimizing the systems and operations management challenges in the Windows environment.” This integration, the report claims, delivers three capabilities:
In conclusion, Forrester recommends that, “Existing System i5 customers should put this integrated i5/BladeCenter product offering on their shortlist. . . .” The Forrester report does not mention the fact that the Integrated xSeries Server–which presumably should be renamed the Integrated System x Server now–even exists or is a better option for small shops who do not need a 14-server blade chassis with their iSeries or i5. That’s because IBM is, whether it admits it or not, focused on bigger OS/400 and i5 shops and selling iron to them. This way, it gets more money per sales pitch.
The specifics of the argument for consolidating Windows servers on the iSeries and i5 may be enhanced by recent BladeCenter announcements, but the theme is not new. Moreover, people always talk about running Windows on the iSeries or on the i5, but what it really does is run next to it. Windows is not a native workload on Power-based servers. It runs on internal X86 co-processors or on external X64 rack or blade servers. Windows runs with the i5, not on it.
For some time now, IBM has been making the consolidation pitch a backbone of the iSeries and i5 marketing strategy, and a July 2005 white paper produced by Andrews Consulting Group still holds a prominent spot on the IBM System i Web site.
The author, Lee Kroon, makes the argument that, Windows PCs and servers have increased productivity by enabling workers to make better and faster decisions, and they have “become as integral to business operations as the telephones on our desks.” However, Kroon goes on to point out that, unlike telephones, “Windows-based networks require regular attention from IT staffs to keep running. Moreover, as the number of Windows servers grows, the time spent to manage them grows at even faster rates. As a result, Windows system administrators are spending steadily growing amounts of time to keep their servers and networks working properly.”
The solution, the author concludes, lies with the consolidation of these servers onto the iSeries and i5 using the IxS coprocessor or the IxA, and iSCSI adapters–a strategy that will “reduce the amount of time that a company will spend administering its Windows applications; improve the availability, security, and performance of the Windows systems; and give Windows system administrators more time to work on issues that will deliver added value.”
Forrester and Andrews Consulting are certainly not alone in observing and touting the benefits that should exist from tightly coupling Windows servers with OS/400 and i5/OS platforms. George Hamilton, research director for Yankee Group, points out that that trend to server consolidation has been going on for some time. “Virtualization enables you to do that. You don’t have to run as many physical instances of servers; you can run them logically and better optimize your resources. It makes sense for companies to do that.”
Of course, if Windows was native on Power servers, it could run inside the same kind of logical partitions that i5/OS, Linux, and AIX run in, and this would result in much more efficient use of Power hardware. It would also obviate the need for IxS co-processors or iSCSI links and external servers–blade or otherwise. In a System i way of looking at the world, maybe a logical partition should be a blade and that is the end of that. No mixing and matching hardware. Keep it simple. But that can only happen if Microsoft and IBM work to get Windows Server 2003, or the future Longhorn Server, native on Power iron. Which they could do fairly easily. They just choose not to.
Still, there are pitfalls even to virtualized environments, Hamilton says. “While it’s good in concept to consolidate your servers, there are a lot of connections that can get broken when you do it, especially with Windows servers that have a lot of direct-attached storage, and you have to be mindful of that. When you’ve operated a data center that has always provisioned servers in the same way and you have a process on how to deploy the server and get storage dedicated to it and get everything connected, and now you want people to do things completely differently, there are always challenges. It may sound good in theory, but there are always problems.”
For example, Hamilton notes that the dynamics between the two different groups of people supporting the two platforms will change as they will now have to work more closely together. There are two different disciplines, two different ways to provision things, different processes, and different technologies. In the long run, it seems to make economic sense to consolidate as much as you can onto servers that can share as many resources as you can. And you can manage that with fewer people.”
Wayne Kernochen, president of Infostructure Associates, which is part of the Valley View Ventures consultancy conglomerate, thinks that “it will be interesting to see to what extent there will be a demand for including Linux in this story. I would think that there will be an added benefit because the System i can handle business-critical applications, Windows infrastructure, and Linux infrastructure. That will be another plus for the System i.”
But if there is not yet any great groundswell among System i users to move in this direction, what might be the impetus in the future? Kernochen believes that it will depend on how much central oversight a company has on these different environments. “My sense is that there are a few sites where there is strong central oversight, and people are saying overall, ‘we need to save money.’ That’s where server consolidation tends to come in anyway. In those cases, there will certainly be a move to combine Windows and System i on the same box. But Windows infrastructure tends to be less often controlled from a central point in the organization, and where that lack of central control exists, there may not be much movement.”
Sageza Group analyst Clay Ryder believes that System i and Windows integration has been underplayed for a long time by IBM and the market overall. “One of the neatest things about the System i is that through the systems adapters and the integrated servers, you can pull a Windows environment under management within the System i context, and that’s a wonderful approach for helping manage Windows-based solutions and share the resources of the System i, and a lot of folks have missed out on that. It’s a piece of integration I wish IBM would beat its chest a lot more about.”
If they realized the capability and the benefit, Ryder believes, people would move in the direction of managing many disparate environments on the System i. So why aren’t more people doing it? Ryder responds that for many, they don’t even really know that this is a possibility. “They aren’t aware of the adapters that are out there to bring xSeries or other Intel-based systems into the i environment. They don’t know that the Integrated xSeries Server is an option.”
Echoing the age-old complaint, Ryder doesn’t see IBM really pursuing this as an option for its customer base. “I’ve told them many times that I think they are undervaluing the asset that they have here.”
In addition, Ryder notes that some people who have been on the platform for a very long time see the iSeries and i5 as boxes that sit in the corner on their own. They love it because it doesn’t break, but they don’t envision it connected with the rest of the enterprise through consolidation. “That’s great for a 1992 approach to the world, but data center management is the bulk of the cost of IT these days, and anything that reduces that cost is a good thing.”
According to Ryder, most shops are not even looking at integrating the Windows environments with their iSeries or i5 servers. “That’s a very simple task that they are not even doing. They don’t have to make the decision to move the applications inside the iSeries or i5. They are not even putting in the connector cards so that the iSeries or i5 can see and control the user password databases and things like that. It doesn’t require a rip and replace, or even waiting for an upgrade. They could do some of these very incremental sorts of things with just a cable and a card.”
Will the System i continue to sit in the corner separate and apart from the Windows–and indeed, Linux and Unix–boxes? No, says Ryder. “There’s been a huge explosion of storage area networks, Internet connectivity, front-ends to glue onto the applications–all kinds of reasons to bring these systems out of the corner. But as long as the mindset is that the System i is somehow special or different, it will be difficult. It’s just silly, though, to maintain any kind of silos unnecessarily.”
IBM itself is part of the problem, Ryder says, because it seriously underplays the capability of the platform. I shared with Ryder the fact that I had gone to the System i page on the IBM website and typed “server consolidation” into the search box, expecting to find IBM’s boilerplate wording on how great the System i for consolidation of a variety of environments. Instead, the first entry on the list (and most of the subsequent ones on the first page) directed me to a pSeries solution. I also pointed out that the words “integration” or “consolidation” do not appear on the IBM’s System i Web pages. In short, even when I was actively looking for it, I could not easily uncover any material telling me about how to use the System i to consolidate and/or integrate Windows (or any other) applications. (I’m sure it’s hidden there somewhere.)
“I haven’t been able to uncover the reason that IBM has not approached this with the System i with more gusto,” says Ryder. “It’s one of the great mysteries of the industry. Even if companies do not want to consolidate on the System i, they should at least be looking at integrating or providing some more systems management across platforms. Why does IBM make it so hard to find? Is the market so unreceptive to it that the pSeries guys get to do all the cool things even though it’s the same exact gear?”