According to a recent study by Meta Group, the AS/400 gets very high marks when it comes to supporting SAP’s R/3 enterprise resource planning suite, beating out more popular Unix and Windows NT environments. META’s study of the real cost of R/3 installations was not as large as it could have been – the study only surveyed 50 users of R/3 around the world, who chose one of the three platforms to support R/3’s core financial applications – but it nonetheless bodes well for the AS/400 division and its strategy of increasing the ERP market penetration of the AS/400 platform. As things currently stand, the AS/400’s share of the ERP market was 18% last year, down from 20% in 1996. More significantly, the AS/400 portion of the ERP market grew only at 32%, compared to the overall 41% growth rate of the ERP market as a whole in 1997. These are trends that IBM wants to reverse, and if it can get the word out about the AS/400’s distinct advantages, it may be able to get the AS/400 slice of the ERP market to at least keep pace with the rest of the industry. Meta talked to customers using the R/3 core financial applications using each of the three types of platforms. The Unix users had an average of 1,500 employees with 94 active R/3 users; they tended to use a two tier R/3 architecture, meaning that the database and application portions of the R/3 suite (which can be split and usually are in SAP’s benchmark tests) reside on a single server. Windows NT shops running R/3 (which META says were harder to find even though they make up more than half of R/3’s platforms shipped in the past 18 months) also used a two tier R/3 architecture, favoring Oracle’s DBMS rather than Microsoft SQL Server, that supported an average of 56 R/3 users at companies with an average of 340 employees. Obviously, NT is not a high-end solution, although customers supporting more than a hundred R/3 users on NT said they planned to go to a three tier R/3 setup so they could support more users and better manage the performance of their R/3 installations. AS/400 shops running R/3 fall somewhere between the Unix and NT installations. The AS/400-based companies surveyed had an average of 3000 employees but only 75 R/3 users (there were obviously some very big firms in the AS/400 data who were just getting started on giving access to R/3 applications). Like Unix shops, AS/400 shops tended to stick with the simpler two tier R/3 setup, which was made possible only through IBM’s giant Apache and Northstar SMP servers that are considerably more scalable than NT servers and are almost as scalable as the biggest HP and Sun Unix servers. For the purposes of comparison, the cost analysis data META gathered from all R/3 users surveyed was normalized for 100 users on each platform. Total costs for the R/3 installations were spread over five years.
Little cost difference
According to the META study, there isn’t a lot of cost difference between AS/400s, Unix servers or NT servers. NT servers came in with the lowest platform costs at $4,887 per R/3 user per year ($2.4m for 100 users over five years). But that was only slightly lower than for the AS/400, which had slightly higher hardware and services costs and slightly lower software costs and came in at an overall cost of $4,995 per R/3 user per year ($2.5m). Unix servers were an even more expensive platform even though Unix server costs were lower than they were for AS/400s because software and services expenses were higher. The Unix platform cost $5,157 per R/3 user per year ($2.6m). These are not big differences, of course, which is the good news for the AS/400 business. While META did not provide the specifics of what prices the R/3 users it surveyed paid for equipment, the results, based on our own knowledge of AS/400, Unix and NT server prices in the current market, suggest that AS/400 shops must be getting some of the best discounts IBM has ever given on hardware and software to essentially meet low-end Unix and NT servers on price; it is also likely that customers are buying much more beefy NT servers than they might have been led to believe are necessary by various benchmark tests. The META study also does not address the cost of scaling up to 200, 500, 1000 or more users. We suspect that the price metrics on the AS/400 versus Unix and NT change quite a bit once customers move up into larger R/3 installations. Even with specialized R/3 AS/400 application servers such as the ones that IBM has been selling for more than a year now, costs per user have to rise because memory, disk and other peripherals are more costly on bigger AS/400s than they are on smaller ones – and these more than offset the lower price per MIPS IBM charges for raw AS/400e processing power. Ditto for Unix. A high-end HP V2250 or Sun Starfire or Compaq/Tandem NT ServerNet cluster are all much more expensive even if they are more scalable. Oddly enough, it isn’t price that is the real selling issue for the AS/400 as far as R/3 is concerned, but rather time. According to statistics compiled by META, it took far less time to get R/3 up and running on an AS/400 than on either Unix or NT platforms. The typical AS/400 customer last year needed 9 or 10 months to get R/3’s core financials up and running, and this year, thanks to IBM’s Ready- to-Run-R/3 pre-loading and pre-configuring program for AS/400e servers, it is taking only about 5 months to get the R/3 financials running. (This five months is actually lower than most AS/400 customers planned to take). This compares very favorably with the 13 to 14 months it takes to get a Unix and NT R/3 suite working. Of course, IBM’s competitors are working to get their own speedy R/3 programs in place, and will likely close the time gap soon. But odds are, the AS/400 will continue to have some advantages due to the tight integration of its data base management system, application tools and OS/400 operating system.