Midrange Insights

Analysis of News Events
  • Smaller Small Medium Big Bigger
  • Default Helvetica Segoe Georgia Times

OS/400 and Windows NT: Friends and Enemies

A lot of loyal AS/400 customers, by which I suppose I mean most AS/400 customers, get livid every time they hear anyone in the press or the analyst community mention Windows NT, or now its new name, Windows 2000, formerly known as NT 5.0. I have two things to say about this. First, I would rather not have to talk about Windows NT at all. While Microsoft has certainly gotten better at delivering stable server software since it entered the market in 1993, it often promises more than it can deliver. Say what you will about the conservative folks in IBM’s Rochester Labs, but with the exception of OS/400 V3, they have rarely overextended. Microsoft makes a business out of it. Which brings me to my second point. You don’t like hearing about Windows NT? Too bad. IBM, as well as AS/400 customers and industry analysts, will talk about little else in 1999 and 2000, with the exception of Java, millennium bug nightmares, business intelligence, and e-business.

If I know one thing about Microsoft, it is this: Microsoft generally gets things right on the third try. (Corel, Novell, and IBM’s OS/2 developers will agree on this one.) Windows 2000 is Microsoft’s third try at true enterprise computing, with NT Server 3.0 being little more than a beta, Windows NT 3.51 being somewhat useful, and Windows NT
4.0 being good enough for operations where downtime isn’t a critical issue or where doubling up on servers for failover capability is amenable to the budget. Windows 2000 will be the most important product that Microsoft has announced, ever, and that is why so many people will be talking about it. It’s not so much that Windows 2000 will radically change the way enterprise computing gets done, but rather that Microsoft is finally figuring out how to make an operating system that can compete, feature for feature, against OS/400, OS/390, and the major UNIX variants.

For those of you who don’t know about Windows 2000, here’s the scoop. Microsoft changed the name of NT 5.0 to Windows 2000 and announced that it would deliver four separate versions in stages. Windows 2000 Professional, formerly known as NT Workstation 5.0, will be the high-end desktop operating system. Windows 2000 Server is the new name for plain-vanilla NT Server 5.0; it will offer support for one or two processor servers and is aimed at small and medium businesses. Windows 2000 Advanced

Server is the new moniker for the Enterprise Edition of NT Server 5.0; this version will support up to four processors in a single machine as well as system clustering and load balancing across those clusters.

Windows 2000 Professional is in its third beta now and is stable and mostly finished, according to developers I know who are writing code for it. As I go to press in mid-December, Windows 2000 Server is not stable, nor is Advanced Server.

All three of these editions of the software are expected sometime in mid-1999, but expect delays if the server implementations don’t stabilize.

Finally, Microsoft says it will deliver, perhaps as early as the end of 1999, Windows 2000 Datacenter Server, which will support 16-way machines, up to 64 GB of main memory, and clustering and load balancing like the Advanced Server edition. I would advise companies to stay clear of using any of the server versions in production much before early 2000, and late 2000 looks like an even better idea.

If you haven’t gotten the idea yet, this is serious stuff. Bill Gates is dead serious when it comes to your money, because he wants to make it his. And he is beefing up Windows NT to try to entice you to give it a try, to join the cult of ActiveX, COM, and DCOM, which has the largest programming community in the world and which is expected to represent 45 percent of the $62 billion server market by 2003, according to Gartner Group. NT accounted for about 25 percent of the $52 billion server market for 1998; the AS/400 will probably have more than $5 billion in sales, and if IBM is industrious, it may be able to hold that sales level (Gartner Group doesn’t think so). While about 70 percent of those NT server sales will be low-end servers, the same holds true for the AS/400 base, which is increasingly dominated by the Invader line of AS/400e servers and other small machines.

I have a particularly vivid memory of sitting at a table eating lunch with John Thompson, who was AS/400 general manager when the AS/400 division was announcing its first server models in 1993, just as Windows NT first got to market. I asked him if he thought NT was a threat to the AS/400. He laughed, shrugging off the whole idea. I didn’t laugh, and neither did a few other people at the table. Not long after that, Thompson was in charge of IBM’s software business, and then he saw very clearly that Microsoft was gunning for him. Last year, for the first time in history, Microsoft sold more software than IBM. I don’t think Thompson is laughing any more, nor is anyone else at IBM when it comes to NT. IBM’s attitude about Windows NT, particularly in the AS/400 base, seems to be “Embrace and extend”—the first three parts of the Microsoft battle cry: “Embrace, extend, exterminate!”

Like you, I wish IBM had been chanting all three back in 1993, when NT was still a joke, but as the mainframe business was imploding, Big Blue was too busy trying to reconfigure itself to think long-term. IBM is a better company than Microsoft; it has always been a less greedy citizen of the world and one that had some sense of its social responsibilities. As vicious as IBM has been through its nine decades of existence, it has been more or less tamed by the courts and has fostered a large part of the computer community (either directly or indirectly, as intrepid techies have developed alternatives to The IBM Way) in which we make our respective livings. Microsoft has a long way to go to match IBM in this respect.

A Second Opinion

Gartner Group’s server analyst, Tom Bittman, doesn’t mince words and is one of the few analysts in the business who will speak frankly about the issues that concern a lot of AS/400 shops. In a recent presentation he gave at the AS/400 Web Summit in New York, sponsored by The 400 Group, Bittman laid out the competitive landscape in the server market between 1998 and 2003, paying particular attention to comparing and contrasting OS/400 and Windows NT.

“Just a few years ago, IBM was selling the RS/6000 as its strategic growth system. That has changed,” said Bittman. He said that RS/6000s are now platforms for server

consolidation and specialized functions like Web serving, and that IBM has refocused on growing its OS/390 and OS/400 bases. “The AS/400 has the best ability to mix workloads today aside from OS/390. Windows NT can run multiple applications, but they may not run well,” said Bittman. Perhaps more significantly, Bittman said that in many market niches, NT already has better breadth of applications than Sun’s Solaris and IBM’s AIX UNIX variants, and that by 2001, NT will catch up to IBM’s OS/400 and Hewlett- Packard’s HP-UX implementation of UNIX. “OS/400 still has the largest and broadest application suite, but it is being challenged by UNIX, and in the future, by NT.”

Bittman said that while Windows NT was not a viable alternative to the AS/400 in 1998, it certainly will be on par over the next three years. “The sweet spot of the AS/400 market will be in NT’s reach in 1999, and they [Microsoft and PC vendors] are going to make a lot of money.” Bittman believes that an NT server will be a viable alternative for about half the AS/400 base by 2001 but expects that the AS/400 will remain competitive until at least 2003, based on IBM’s and Microsoft’s plans for the midrange. He fully expects that most AS/400s will be running mixed OS/400-NT shops by 2001, using the AS/400 for primary databases and critical applications and NT for some core applications and infrastructure like Web, print, and file serving. Bittman said that he expects AS/400 customers to install up to 40,000 Integrated PC Servers (IPCSs) running Windows NT by

Whether or not the NT-IPCS cards are a first step in a gradual migration from OS/400 to NT or the foundation of a long-term coexistence of OS/400 and NT depends on how good Microsoft is at delivering on its promises for Windows 2000 and its associated BackOffice suite of system software. If Windows 2000 is good, reliable, and understandable, a large migration could occur, and certainly there will be lots of coexistence even if it doesn’t. (No one is saying that this should happen, but that it may happen in a kind of Brownian motion that often captures markets.) If Windows 2000 stinks on ice, AS/400 shops will sit tight.

After fighting it out with other Gartner server analysts, Bittman said this is the consensus among them all concerning OS/400 and NT:

• Scalability: The AS/400 enjoys considerable scalability benefits today, but Wintel servers will have almost the same high-end scalability as AS/400s by 2001. For mixed workloads, AS/400s will likely hold their lead at least into 2001.

• Reliability: The AS/400 currently has better inherent reliability, both for hardware and software, and with the use of third-party software, better failover capabilities as well. But NT will almost catch up by 2001. “NT has some stability problems,” said Bittman. “But that will be fixed over time.”

• Initial Price: There’s no doubt that NT servers are considerably less expensive, at least for raw iron, than AS/400e servers, and a lot less than AS/400e systems. Gartner says that once customers start trying to load up complex workloads on their servers, NT servers take more iron, and that drives up prices. Gartner says that the total cost of ownership for NT and OS/400 servers should be about the same if the systems are managed correctly.

• ISV Support: The AS/400 currently has better independent software vendor (ISV) support than NT right now, especially for big application suites. But Gartner expects ISVs to shift their priorities to NT by 2001 and that AS/400 support will actually wane a bit while NT surpasses its current level of support. Why? ISVs are cranking lots more money into NT development because that is where they think they can make money. “But being there will not guarantee success,” said Bittman. “Many AS/400 vendors jumping into the NT market will fail and be bought out.”

• Vendor Support: Bittman said that the AS/400 has the best geographic coverage of any server in the market and that support for NT servers is still very immature and that worldwide coverage is not very good. NT will continue to improve as vendors like IBM, HP, Compaq, and Dell push their products hard worldwide and continue to increase their services revenue from NT support. However, the quality of support for NT servers will not match the quality of support for the AS/400 even by 2001.

Storage Consolidation Saves Money, Too

Last month, “Midrange Insights” focused on server consolidation (see “Midrange Insights: The Pendulum Swings Back from Distributed to Consolidated Servers,” MC, January 1999). Customers who can’t or won’t consolidate their disparate servers can nonetheless save some money by doing the next best thing: consolidate their server storage.

The key to storage consolidation is so-called “enterprise storage servers,” the name that many vendors and consultants have for giant storage arrays that can handle the storage needs of multiple and incompatible servers simultaneously. IBM’s Versatile Storage Server, or “Tarpon” disk array, which supports AS/400, UNIX, and NT servers (and by next year, mainframes, too) is one example of an enterprise storage server; EMC’s popular Symmetrix line of storage arrays, which supports the same environments, is another good example. There are two main reasons why enterprise storage servers are important options for data center managers. First, because these machines can support the major operating systems and servers, the task of buying disk storage and redeploying it as conditions change is greatly simplified. Second, because data is stored on a common box with a single management interface, it is (in theory) easier and cheaper to manage.

At this point, there isn’t a lot of empirical evidence to back up storage vendors’ claims that customers can save money by consolidating their disk data onto enterprise storage servers, but many customers have bought Symmetrix or Tarpon arrays just in case they do change server platforms. A study by International Data Corp (IDC), based on interviews with customers who had very large enterprise storage servers (ranging in size from 2 to 10 terabytes in size) connected to multiple UNIX, NT, and sometimes mainframe servers, is just about the only data that is available to show the kind of savings customers can expect from storage consolidation.

The IDC study, called Demonstrating the Value of Enterprise Storage, indicated that in a typical decentralized data center, the typical storage manager is responsible for about 100 GB of capacity and the management costs for that capacity account for about 55 percent of the total yearly storage budget. That is a lot of money to spend on running around. When all the storage arrays for the distributed servers (which might be in the data center or elsewhere on the corporate campus in various departments) are co-located in the data center—but are still tied to their servers, operating systems, and disk management tools outside the glass-house walls—then a storage manager can handle about twice as much data, over 200 GB, while management costs fall to 35 percent of the total storage budget. When companies consolidate the data on their disparate but co-located storage arrays onto enterprise storage servers, then a single storage manager can handle 750 GB of storage, and related management costs drop to 15 percent of the storage budget.

There’s a catch, of course. Enterprise storage servers typically cost twice as much as plain old disk arrays. IDC’s research suggests that, when all of the math is done, storage consolidation will save companies only between 15 and 20 percent of their overall hardware budgets because of increased prices and maintenance costs for enterprise storage servers. Still, the simplicity and the savings make storage consolidation a compelling choice, and it is a trend that will likely catch on for those very reasons, even among smaller AS/400 shops.