Interview: IBM’s Center for Microsoft Technologies

In April, IBM Corp. opened the Executive Customer Briefing Center within its IBM Center for Microsoft Technologies, based in Kirkland, Wash.

KIRKLAND, Wash. -- In April, IBM Corp. opened the Executive Customer Briefing Center within its IBM Center for Microsoft Technologies, based here.

ENT Senior Reporter Thomas Sullivan sat down with Roy Clauson, director of the center, and Jerry Walsh, manager of enterprise solutions at the center, to discuss IBM’s involvement with Windows 2000 development, and how Big Blue plans to differentiate itself from competitors in preparation for the coming operating system.

Thomas Sullivan: What is the aim of the IBM Center for Microsoft Technologies, and the new Executive Customer Briefing Center?

Roy Clauson: When we started the operation, the intent was to get a closer working relationship with Microsoft. The first thing we did was the NT port to the PowerPC. The mission was expanded 4 years ago to cover all the Personal Systems Group (PSG) products, and all the Intel products. We’re totally focused on making Microsoft products run really well on our Intel-based products. And that includes, by the way, the Intel processor in the AS/400, so we’re going to put NT Server under the covers of an AS/400.

Jerry Walsh: With Beta 3 of Windows 2000, the barn doors have been blown off and Microsoft is close enough to the final version that the time has come for IBM to begin educating its customers about what to expect with the new operating systems, especially in relation to IBM hardware. The Executive Customer Briefing Center is where we do that.

TS: What role does IBM play in improving the reliability and scalability of Windows NT and 2000?

RC: A couple years ago we started getting a lot of interaction with customers looking at how to get higher reliability, availability and the five nines [99.999 percent uptime] in the Wintel platform. So that resulted in this initiative we call On Forever, in which we are trying to bring to the Wintel platform the notion that this is a mission critical server. We can diagnose problems online without taking the system down, we can add or replace defective parts online, we added PCI Hot Plug last year, and we’re looking at Hot Plug memory now and more online diagnostics. So we’re working to bring the NT/Intel platform into the same availability model people are accustomed to in their glass house. But, it’s going to be a progression over the next five years or so.

TS: The five nines program focuses on Windows NT Server, but doesn’t guarantee that the applications running on that server will also be up that long. How does IBM plan to improve the overall system performance and what will it take to achieve that next level, where the whole system reaches five nines?

RC: You have to look at the system in total. That’s going to become application-by-application specific. We are trying to make sure the hardware and the operating system are there and as this journey progresses then we have to look at what is needed in middleware and the applications as well. As we get the OS and hardware more robust, there will be functionality that gets added to the OS that applications will have to be aware of and leverage. Let me give you an example: In the PCI Hot Plug, we’ve got the ability today to add in redundant network cards. The drivers for those network cards have to be changed -- and that’s a software change -- such that they know when one goes down the load that was on it is going to fail over transparently to the applications above it. We had to make changes in the code to take advantage of this underlying support in the hardware. That kind of thing will propagate its way up as functionality is added into the OS. This is a long-term focus and sort of a mindset.

Jerry Walsh: What we’re finding with customers is movement away from best price, low bid decisions on things like $39 NIC adapters. They realize that in a lot of cases it’s much more worthwhile to invest in components that have been subjected to a higher level of testing. That’s what [Windows 2000] Datacenter [Server] is trying to do, and we see that over and over again with customers coming in and saying "this is the configuration I’ve locked down, this is the configuration I’ve tested and this is the only configuration I am going to employ. Wouldn’t it be nice if someone could articulate this and state it to technicians?"

To improve reliability, we need to do that with Windows NT, as we have done with the mainframe: Put components together and then just hammer them, so we know they are going to work together.

TS: How does the testing done here influence the development of Windows 2000?

RC: We find problems that get fed back to Microsoft. We’ve done some work with Microsoft in the [Advanced Configuration and Power Interface] area, which actually is a carryover from what we’ve done with Hot Plug on NT 4.0, and has sort of shaped their thinking in how they want to do it in Windows 2000. We went through all of the existing legacy systems that IBM ships, as well as all those that are in the process of development and will be out shortly before or shortly after Windows 2000. That is a fairly significant number of systems that have been tested, and various problems have been uncovered. We get asked if we are happy with where we are for Beta 3 and if it is okay to move forward. So we are involved with feedback in that arena as well.

TS: What is IBM Doing to differentiate Netfinity servers from its competitors, especially on an enterprise level?

RC: Our On Forever initiative is very focused at the Netfinity server. Also, light path diagnostics is a feature that I haven’t seen on competitor’s machines. When there are components that fail, little lights on the boards highlight the failing components so you don’t have to go in and physically check a number of cards to get to the problematic one. And we are trying to get to high availability and provide a more proactive approach to managed capacity growth, the predictive functions that alert you when it looks like your disk is starting to fill up, or other problems are going to arise and so forth.

In some cases we want to work with our competitors. For example, the On Forever initiative is driving a standard for PCI Hot Plug, so we don’t have a unique solution. There are areas where we like to differentiate ourselves, there are areas where we like to get the industry with us, depending on what the project is.

TS: What is IBM doing in terms of linking Windows NT servers to IBM mainframes via interconnect technologies?

JW: There is a whole component of the X-architecture blueprint that deals with the interconnects and the first one that is shipping today is the ESCON channel mainframe connector. But there are plans and approaches from a hardware perspective that have to be designed first that are all up to the switched-fabric technologies, and that is where we pick it up to make sure the interconnects link correctly with the Microsoft software.