By Timothy Prickett Morgan
If computers were like cars, where status comes with the amount of money you can afford to conspicuously consume on a finely engineered but nonetheless overpriced vehicle, the AS/400 would be well ahead of PC servers in the race. Maybe there was a time when the IBM label on a computer was an immediate recognition of corporate status — in the 1980s, companies would pay over list price to get moved ahead on the waiting list for IBM’s mainframes. Maybe companies, once having chosen IBM machines, had to pretty much stick with IBM machines because unlike cars, which all burn the same (or nearly the same) gas, computers all run their own special operating systems and customized application code. In a very real sense, once each computer is turned on and loaded with a company’s most vital assets – its information and business rules – it becomes its own special machine. That’s how most AS/400 customers, despite their occasional displeasure with IBM about frequent product introductions, high prices and absurdly rapid equipment value depreciation, feel about their AS/400s. In its more than ten years of existence, the AS/400 has arguably been the most popular general purpose computing environment in history; OS/400 started out as a proprietary environment, but has lots of Unix APIs and supports Java, TCP/IP and other internet standards about as well as IBM’s other platforms. Well over 200,000 unique customers have bought more than a half million machines (the number is probably close to 600,000 processors, including new machines and processor upgrades). Sales of AS/400 processors and disk subsystems have brought about $55bn in revenues to Big Blue since 1988 and probably in the neighborhood of $35bn in gross profits; AS/400 customers have probably bought upwards of $100bn in other IBM equipment and services, including PCs, maintenance, software and so forth. The AS/400 business has been and continues to be a big business for Big Blue, even by the standards of Microsoft and overvalued internet stocks. Intel and its PC server partners and Microsoft and its systems and application software partners have been working for years, and will probably have to work for yet another year, to bring Windows 2000 and its associated PC servers to market and give it the kind of respectability that the AS/400 has always enjoyed. That Microsoft seems intent on and capable of making its Windows 2000 platform the most popular proprietary operating environment for the support of strategic, enterprise- wide applications – not just print, file and web serving as it is relegated to in most shops that use NT – is obvious. The question is, can Microsoft really match the AS/400? The other obvious question is can the AS/400 survive against the full weight of the Microsoft machine?
Scalability advantage
Tom Bittman, the director of server market analysis at Gartner Group and a man who knows the AS/400 well – he used to work for IBM’s Rochester Labs – has spent a lot of time and energy trying with his staff to work out how the AS/400 stacks up against Windows NT and the forthcoming Windows 2000. There’s no simple yes or no, live or die responses to the questions posed above. The issues are complex, and so are the answers. After arguing about it, the Gartner Group consensus goes something like this. Right now, the AS/400 has more raw scalability than a PC server running NT. A big Northstar server can have twelve 262 MHz Northstar PowerPC processors that can handle about 45,000 TPC-C TPM and 27,000 Domino mail users; memory scales up to 40 GB and disk capacity will, with the February 9 Northstar announcements, go up to 4.2 TB. A four-way Pentium II Xeon server running Windows NT can process about 20,000 TPC-C TPM and handle between 6500 and 7000 Domino Mail users. The AS/400, says Gartner, has another scalability advantage over Windows NT servers in that its OS/400 operating system does a much better job of handling mixed application workloads. By 2001, the first scalability advantage will pretty much be closed and even if not, irrelevant for all but the largest server customers. But Gartner says that the sweet spot in the AS/400 market among small and medium businesses will be in reach of NT by the end of 1999. Gartner says the AS/400 will still enjoy advantages in supporting mixed workloads even by 2001. But it is likely that this is a much more sophisticated argument than most server sales reps can deal with. Reliability, of course, is one of the hallmarks of the AS/400. Not only that, but it has had failover capability, provided by third party software vendors Lakeview Technology, Vision Solutions or DataMirror, for years longer than Microsoft has had its Wolfpack failover clusters. With V4R4, OS/400 will be gutted to provide the basic plumbing that will eventually allow third parties to create Tandem-like fault tolerant clusters with load balancing as well as failover among nodes in a cluster. Microsoft wants to add the same capabilities to NT. For the moment, Gartner says that the AS/400s enjoy better inherent reliability (that’s hardware and software) as well as better failover capability. In a recent Gartner study of 190 companies, AS/400s were found to have an average of 99.994% reliability (5.2 hours of unplanned downtime a year) compared to 97.44% (224.5 hours of downtime) for Windows NT servers. (In the study, about a tenth of the AS/400 and NT servers were clustered, increasing the reliability numbers some.) These advantages, however, will at least as far as Gartner is concerned, only be sellable characteristics of AS/400s for another two years, since by 2001 the company expects that NT servers will have just about the same reliability. On pricing, NT servers will of course continue to have an advantage compared to AS/400 servers, despite IBM’s pronouncements that AS/400 servers are equivalently priced to NT servers – they are not and never have been. Gartner says that the high price of AS/400 hardware is offset in large part by the fact that to do multiple jobs, customers have to buy big NT server farms, putting different applications on distinct machines rather than loading them up on one big AS/400 server, which is easier to use and simpler to manage. By 2001, NT servers will still have attractive initial prices, just like today, and AS/400s will have some total cost of ownership advantages, although probably less than today. Perhaps more significantly, Gartner expects that NT will continue to attract more ISV investment dollars and could eventually have an applications portfolio that rivals that of the AS/400, which is the largest in the computer business. The shortage of skilled NT techies now and in 2001 and the ease of administering an AS/400 will continue to play into the hands of IBM. The upshot of all this, says Gartner, is that while Windows NT is not an appropriate platform for AS/400 customers to consider using today, over the next two to three years it will become an appropriate alternative for about half of the AS/400 base to use instead of an AS/400. No one is saying AS/400 customers will or should move to NT, but rather that it will be a practical option. In the meantime, IBM seems content to have Windows NT as the only real supported environment on its Integrated PC Server for AS/400s and Gartner expects that IBM will start pushing mixed NT- OS/400 environments, with the AS/400 as the central data base and application logic split between the AS/400 and NT. That this runs counter to its drive to get everything native on the AS/400 is ironic, but IBM wants to have it both ways. It wants to sell native AS/400 applications when it can (it makes more money that way) and sell a very tightly controlled NT solution when it can’t. Whether or not this is what AS/400 customers want, only time will tell.
