The Four Hundred

The importance of taking legacy applications to the Web is not lost on iSeries shops. It may be that the decision about how to accomplish this task is causing problems ranging from perpetual postponement of the decision to making bad decisions and having to explain them. But that is just the dark side. There are ample e-business success stories, from customers using a variety of middleware and tool vendors. One thing is for certain: For iSeries shops, it's more than a WebSphere world.

After three years on the market, WebSphere is still trying to pick up momentum in iSeries shops. That's not to say it is nowhere to be found in the iSeries world, but it has frankly had its problems. WebSphere on iSeries has had its best success in the largest companies where hundreds of programmers are on hand and IT budgets reach Kiliminjaro heights. Among small and midsized businesses, WebSphere is spinning its wheels more than it is gaining traction. IBM hopes to change this story. Its marketing machine is a force, and its intentions about favoring Java and interoperability are clear. But in the real world, rather than the idealized world of marketing, many small and midsized companies are successfully moving their iSeries-centric applications to the Web with less-expensive and quicker-to-deploy tools that make it difficult for WebSphere to gain ground.

While IBM wades into the small and midsized business market, hoping to make its case with WebSphere, independent software vendors have been making some hay. Software vendors such as BCD, ASNA, LANSA, and mrc have been doing well in this arena, despite the marketing juggernaut that IBM has unleashed with its WebSphere strategy.

One issue that works in favor of many of the ISVs, when compared with WebSphere, is implementation time. There is urgency on the part of many companies to build Web sites with e-business in mind. Taking back-end applications to Web users is a priority, and many IT departments face mandates to get the job done yesterday. More often than not, the project managers, and the companies they work for, demand solutions that are quick, clean, and simple--as long as what they build today ultimately works within the framework of a long-term solution.

"The people we encounter as our customers and prospects have programming teams that want to implement, maintain, and add to their own solutions," says Roger Pence, director of education and training at ASNA. Pence says most iSeries shops are "wrestling with the bare-bones basics" of taking applications to the Web. "It doesn't matter whose tool you select; there's work to do. There's a lot to learn." The learning, Pence says, involves a lot of new technologies and often comes as quite a shock to the typical iSeries shop.

"Our point of view," Pence says of ASNA, "is let us show you how to leverage what you do know, the stuff you have on the back end, and then you can focus on the new stuff you need to learn." IBM in its WebSphere marketing "doesn't come clean with all that's required over the long haul." It causes companies to get in trouble in a hurry and need lots of help. "To build an enterprise solution [with WebSphere] is a lot more expensive and a lot more labor intensive than you are led to believe."

Pence is familiar with WebSphere disasters. "I'm dealing with two WebSphere projects right now that spiraled out of control because they expected to do too much too quickly."

Pence isn't the only person to note that iSeries shops are raising questions about the expense involved in WebSphere projects.

"I can tell you of an $800 million electrical company that burned through $100,000 and were totally dismayed because they did not get what they wanted with WebSphere," says Eric Figura, manager of sales and marketing at BCD. "They didn't even come close. The company came to BCD and said, 'We'll pay you for 40 hours of consulting to show us how far you can go with the application that we want.' After 38 hours of coding, using BCD's WebSmart, Figura says, the BCD programmers completed the company application.

"Things aren't happening for WebSphere in the iSeries marketplace," Figura says. "I hear it from the customers. They can't afford to be spending $20,000 to $200,000 on hardware upgrades just to run a particular application on WebSphere." Associated costs related to tier-based pricing on software and maintenance contracts can also factor into budgetary considerations.

Hardware and software upgrades are one aspect that contributes to the overall cost of implementing WebSphere. The cost of hiring skilled WebSphere consultants to write Java applications, to configure the server, and to maintain the development environment can dwarf the hardware costs. Installing, implementing, and using WebSphere requires considerable resources--financial and personnel. It's too much for many midrange companies. Even when the number of applications being brought to the Web is few, there is still the WebSphere learning curve to get past or the prospect of paying consultants to watch the shop, at least in the short term.

According to Figura, in the past year 165 companies have implemented BCD's WebSmart, which just a week ago was upgraded with a new release, WebSmart3. The success of WebSmart, many say, is directly related to its low cost and ease of use when compared with WebSphere. It definitely appeals to the iSeries small and midsized business market, where smaller AS/400s--Model 150s, 170s, and 720s, for instance--are running. "Those people have zero chance of running WebSphere," Figura says.

John Quarantello, who oversees marketing plans for IBM eServer solutions, doesn't deny this is a problem for WebSphere in this market.

"It is fair to say the initial success we've had with WebSphere has been with some of our larger customers," says Quarantello. "We have been trying to figure out a way that we can become a lot more pervasive with the smaller customers and let those customers be more successful with WebSphere. WebSphere is a robust Web application server; it loves CPU--the faster the CPU, the better. It also loves memory. You need to have a certain sized iSeries, Intel, or Unix machine to have adequate performance," he says.

"The biggest problems we run into are with companies trying to run WebSphere on a little Model 170," says Quarantello. "It will have horrible response times. The minimum CPW that people should be using for WebSphere is in the 300 CPW area. I'd recommend 500 CPW and a half gigabyte of memory as a minimum." He also suggests that users go through the iSeries' workload estimator to make sure the system is correctly suited for their anticipated transaction volume. Try to do something on a 200 CPW machine, Quarantello says, and you will not be satisfied with the results.

But don't count WebSphere out. Developments with ISVs like J.D. Edwards, Intentia, and MAPICS--all of which have WebSphere as a base part of their infrastructure--will be important to WebSphere's future. And as more iSeries shops upgrade their hardware and operating systems, WebSphere will have the processing power and the features that will put it in a better light. "One of the things we are considering for 2004 and 2005," Quarantello says, "is to build advanced function around WebSphere Application Server. At some point in time, if you want to do advanced OS/400 functions, WebSphere will be a prerequisite in part of that."

Dan Galvan, recently hired as vice president of WebSphere marketing and partner programs, thinks the focus should be on integration. "A lot of companies have gone through the first couple of stages of e-business," Galvan says. "They've published on the Web. They have Web sites. Many have moved on to the stage of having a commerce presence on the Web. As soon as you do that, you realize it matters how you interact with the back-end systems. It matters how your applications work with each other--if they work with each other. It matters how your inventory systems tie in with your ordering system and how it all ties to your Web front-end."

Galvan says the story is shifting from feeds and speeds on the front end to how you do integration. This, he says, is a market shift in WebSphere's direction. As an example, he points to companies doing business as downstream suppliers to Wal-Mart. "Those companies have found that when a company like Wal-Mart creates a standard for interaction, it requires a level of integration that wasn't imagined before."

For the time being, the perspective on WebSphere is that it makes the most sense for companies that rely on multiple platforms, a heterogeneous server environment. It doesn't fit well for companies that don't have all those issues.

Joe Stangarone, president of mrc, is also very familiar with the iSeries market for Web application software. Productivity Series is mrc's tool for modernizing legacy applications for the Web.

Stangarone's observations include the WebSphere fans and foes. "I see AS/400 guys who hate WebSphere and guys who love it. They are across the board," he says. One thing that he and others at mrc have noticed in recent months, though, may indicate a trend toward Java servlets.

"I'm not trying to tell you that the majority of the market is ready for servlets," he says. "But in the last three months, we've seen a lot more interest in that and a little less interest in the CGI/RPG things that we've been doing for a long time. I think the market is getting sold on servlets because of the flexibility they provide. I've talked to people who won't use WebSphere for Web serving, yet they say they are committed to servlet technology. They say, 'I'm going to use the AS/400 for my database but find another platform to do those servlets.' Most say they will use Tomcat."

If Java servlet technology catches the attention of people who are convinced of Java's cross-platform benefits, tools such as mrc's Productivity Series and others will be positioned to help those who are looking for shortcuts around the Java learning curve, which has traditionally been a roadblock for OS/400 programmers. Servlets also offer increased performance over the old applet technology that many associate with poor Java performance.

A servlet works like an RPG program, in that 10 or 1,000 users can be connected to one servlet. There's only one copy of that servlet in memory. There might be 1,000 copies of the data for 1,000 users, but architecturally that is smart for a system that is going to support lots of users.

"Java is not a fast language," Stangarone says. "When you try to develop things in Java, it's like going back to development cycles of the 1970s or 1980s, where things took weeks and months to write. If you do things correctly, over time you can see economies in the fifth and sixth applications by borrowing pieces from the first four, but it takes a long time to write and debug stuff in Java."

That's one of the reasons WebSphere has been slow to develop on the iSeries, Stangarone says. But he believes that WebSphere is perceived to be slow because many have tried the free versions on underpowered machines. "The quick conclusion is, 'Why would I want this stuff? I can see why it's free.' I think that perception has as much to do with [the reluctance to use Java] as the Java learning curve."

Mike Gilpin, a vice president and research fellow at Giga Information Group, says that Java is not as much of a priority as other aspects of the iSeries. "In theory, it should be attractive to have Java everywhere in a shop that is using multiple platforms, including the iSeries. But the importance of that still seems to be taking a back seat to other factors, such as ease of implementation and administration. These are not the strong points of WebSphere."

Although the Java learning curve has hindered widespread adoption of the language within the OS/400 community (although ISVs and IBM itself uses Java a great deal), Java is also viewed somewhat suspiciously as an avenue to platform-independence, and therefore a forerunner to the ultimate demise of the OS/400 platform. Of course, platform-independence is the reason IBM is pushing WebSphere and Java. This gets to be a touchy issue, as many OS/400 zealots, who see "exit strategy" written all over this, constantly talk about WebSphere and Java. In that regard, it becomes a highly charged political battle.

ASNA's AVR Web application product is painted with that same brush because it converts RPG to Windows. "We are not about an exit strategy," says ASNA's Pence. "We have tons of customers who say the AS/400 is very important to them and will remain so for years and years, but they also say Microsoft SQL Server and Windows platform are important, too."

Compare what it costs to have a fully configured AS/400 Web server running WebSphere versus a Windows 2000 server running IIS, Pence says. "The difference is tens of thousands of dollars, and potentially hundreds of thousands of dollars, on the consultant end of things."

BCD's Figura sums up the matter of e-business middleware and tools on the iSeries succinctly. "The small and medium-sized businesses that depend on iSeries will leave the OS/400 platform if they don't have a solution that works for them. We are able to save a lot of companies from going off the platform." This, I think we will all agree, is a good thing.