For the longest time, midrange customers and business partners have wished that IBM would better quantify the ease-of-use characteristics of the AS/400 that it has boasted of for the better part of a decade. AS/400 customers all know the benefits that the system provides in terms of ease of setup, programming and administration, but what no one has ever known is how these characteristics stack up against other midrange servers. Such quantification is important if IBM ever hopes to expand the AS/400 base beyond its current enthusiasts. That is undoubtedly why International Data Corp has just released a new study that examines the quickness with which AS/400 shops can design, install, test and roll out applications compared to PC, Unix and other midrange servers.

By Timothy Prickett Morgan

The IDC study, which examined actual midrange customers in the United States and Germany, is the first time anyone has quantified the margin that the AS/400 architecture enjoys in application deployment compared to alternative midrange platforms. IDC is very careful, as you might suspect, to not name names when discussing alternatives to the AS/400 by their operating system; the study discusses generally how AS/400s stack up against PC LAN servers and other midrange servers. If you read between the lines, however, PC LAN server means Windows NT – Novell NetWare is not really an application environment so much as a file and print serving environment – and Midrange means one or another version of Unix with a smattering of Digital VMS and HP MPE. IDC is, as usual, trying to have it both ways (or rather, as many ways as possible) and not taking a stand on its numbers and the exact operating environments it examined at customer sites. But that’s just IDC for you. This doesn’t negate the results of IDC’s surveys, it just makes the results less specific than they ought to be. IDC sought to answer exactly the kinds of questions that all midrange customers are asking themselves and their potential vendors every time they deploy new applications. IDC polled midrange shops about how long it took them to get new applications installed on new server platforms. The assumption was that customers were replacing legacy applications on new platforms. The consultants performing the survey specifically excluded sites that were upgrading e-mail to groupware as well as government institutions and research and other technical customers who use midrange servers in ways not typical of the commercial midrange base. IDC did not specify exactly how different the beginning and ending platforms were. (An AS/400 customer moving from V3R2 and RPG IV on a CISC box to V4R1 and ILE RPG on a RISC box was probably considered as moving legacy applications to a new platform, which gives the AS/400 a serious advantage compared to moving an AS/400 application to a Unix or NT server.) IDC also looked at how customer satisfaction correlated to ease and speed of deployment of applications (obviously, there is a very strong correlation, but asking this question separately is a viable way to check what customers really think about the platform’s ease of use). IDC examined four different phases of application deployment:

Testing phase

The design phase included all aspects of figuring out the architecture of the application. IDC asked about how long it took to architect the application, determine data types, link modules, migrate data from legacy applications, translate that data and modify existing code or completely rewrite it. The installation phase included loading the operating system on the new server, installing applications and their associated data on the server and configuring the end user network for the server. The testing phase included the normal testing companies perform before cutting over to new applications; customers sometimes tested with production data, sometimes not. The rollout phase included running old and new applications side by side using production data and eventually shutting down the old applications. The results of the IDC polls show that the AS/400 holds a significant lead over PC and Unix servers, at least according to all the customers who use the machines. Remember, this is not AS/400 customers rating how well they rolled out NT applications; IDC’s data concerns Unix customers rolling out new Unix applications, or NT users rolling out new NT apps. According to the study, the average AS/400 customer took 25 man months to get a new application up and running compared to over 33 months for PC LANs (that means NT mostly). The AS/400 offer better speed of application deployment than other midrange environments (mostly Unix), which required almost 32 man months on average to deploy new applications. The IDC midrange poll indicates that the AS/400 enjoys a speed advantage mostly because it offers tight integration of hardware, operating system, database and development tools. In many cases, the lack of choice of an AS/400 – there’s one hardware supplier, one OS, one data base and basically one language (RPG) that customers use – is an advantage. This has always been the AS/400’s main advantage, in fact. When there’s only one way to do things, everybody agrees and gets on with it. Being a mature environment is also part of it, too. Building commercial applications on an AS/400 is a well- understood task; people have been doing it for 20 years if you count the System/38. Unix has had half as long to mature as the IBM midrange, and NT half again as much time. Maturity counts, and the IDC results show it. IDC tracked the time it took to deploy applications both with and without the design phase because some customers thought the design phase had very little to do with application deployment. (The users polled who expressed this opinion were obviously stupid. This is the most important part of the process. If you architect an application correctly and choose the right platform and tools, you minimize the pain you encounter down the line as you build the applications and deploy them on the servers.) Including the design phase, AS/400 servers showed an appreciable advantage over PC and Unix servers. PC server applications took 32 percent more time on average to design, install, test and roll out than AS/400 environments; Unix and other midrange environments took an average of 26 percent more time than the AS/400. Moreover, the AS/400 really showed the value of an integrated environment when the design phase was excluded. PC environments required 58 percent more time to get up and running after they were designed and written than AS/400s; even Unix environments took close to 30 percent more time.

Time is money

The design phase of the application deployment was the biggest consumer of time. AS/400 shops averaged 12.5 man months, while PC LAN shops averaged 13.2 man months and Unix and other midrange shops averaged 15.4 man months. This is where an integrated environment really shows up as a positive. Both OS/400 and NT offer an integrated environment (Windows NT and BackOffice are every bit as integrated as OS/400, DB2/400 and RPG; they just aren’t as rugged and robust.) PC environments enjoy an edge in the time it takes to install applications. AS/400s have an edge in application testing. Unix environments come out on top when looking at the actual roll out. But when you do the math and add them all up, the AS/400 takes less time from start to finish. All of this, of course, doesn’t just add up to time. Time, as Einstein proved, is money. (No, he didn’t really prove that.) Figuring on a salary of $40,000 per developer and $20,000 a year overhead on top of that for health benefits, office space, a workstation and coffee, each man hour comes to about $5,000. That means an AS/400 costs $40,500 less to get applications up and running than a PC server. AS/400s tend to be twice as expensive as NT servers of equivalent power, which is a serious detriment. But the speed of deployment of applications can buy at least some of that cost differential back. Exactly how much depends on the particulars of the server setups. The point is, IBM can make a case that even though an AS/400 system costs much more than NT or Unix servers in the same power class, there are quantifiable benefits of using the AS/400 that in time lead to cost savings compared to these cheaper environments. Of course, as IDC correctly points out, the real advantage that the AS/400 holds is not lower application deployment costs, but rather the fact that AS/400 customers can deploy more applications in the same time with the same money. Midrange shops want to get their application backlogs down more than they want to save money. Customer satisfaction is the other aspect of the IDC study that also weighs in favor of the AS/400. While few midrange customers surveyed by IDC in the United States or Germany admitted to being dissatisfied with their machines, AS/400 customers tended, on average, to be more satisfied than customers using other platforms. AS/400 customers are probably more satisfied with their platforms because most AS/400 shops have direct experience with NT and Unix (or both) and even though IDC did not ask them to compare and contrast these machines, when they were rating their AS/400s they were doing it anyway. The AS/400 scores high – or in this case, scores low because on IDC’s satisfaction scale lower numbers are better ones – not just because AS/400 customers are happy with the Advanced Systems and Servers (as well as their D, E and F machines), but because they have been made unhappy by their Unix and NT servers sometime in the past. For the speed of deployment study, customers rated their satisfaction with their platforms based on a scale from 1 to 5, with 1 being completely satisfied and 5 being very dissatisfied. The AS/400 got a 1.5 or 1.4 average rating on all four phases of deployment, and a slightly lower rating on all phases of deployment together (1.9). PC LAN and other midrange servers got a 1.8 or 1.9 rating (mostly satisfied) on the four phases, and lower overall ratings (2.1 for PC servers and 2.4 for Unix and other servers). The discrepancies between individual deployment phase ratings and the entire implementation is strange, but not unexpected. Users can be unhappy about other aspects of the platforms they choose that were not taken into account in the IDC study which nonetheless impinge on midrange MIS managers minds when they rated their platforms. PC connectivity is a real pain with AS/400s, and this undoubtedly lowered its rankings. Similarly, reliability and scalability are issues with NT and some Unix servers, as is the manner of integration of the operating system, data base and application development tools. While the IDC study goes a long way toward making a convincing argument for the AS/400, there are a few things to gripes about. For one thing, IDC was not terribly specific about what applications it was comparing on all these machines. A more rigorous analysis would examine the installation of particular application suites – say SAP R/3 or PeopleSoft or, soon Baan – on each platform. The installation data is available from the vendors who supply these suites. And for those customers who build rather than buy their applications, IDC should have been a little more specific about what applications they were building and what tools they were using. And finally, IDC did not really come clean on what the non-AS/400 platforms were. If they were NT and Unix, IDC should say that and, more importantly, show what differences there were as customers chose different operating systems, data bases and development tools. These choices, as the AS/400 illustrates, make all the difference when it comes to application deployment. á