Legacy Application Modernization Strategies Hinge on SOA
Published: October 9, 2006
by Timothy Prickett Morgan
Aberdeen Group has just completed a detailed study of the legacy application modernization plans of small, medium, and large enterprises, and not surprisingly, companies are looking to the services oriented architecture (SOA) approach to creating distributed applications to help them both modernize their legacy applications and to make their composite applications more flexible and therefore giving their businesses more agility. Some companies, however, are looking to simply get rid of legacy applications on mainframes, i5/OS and OS/400, and Unix servers to get rid of the legacy problem.
To reach the conclusions in a just-released report called The Legacy Application Modernization Benchmark Report, Aberdeen conducted in-depth interviews with 120 companies of all shapes and sizes and locations. Aberdeen purposefully sought out companies with mainframe, i5/OS and OS/400, and Unix servers as their core back-office systems because it wanted to know how different kinds of IT shops felt about legacy application modernization and SOA and approached the issue of making legacy applications play better with each other, with new applications, and with the Web-based Internet technologies that are basically the standard for the front end of applications at this point in the history of the data processing. Aberdeen's data set was skewed toward bigger companies located in North America, but included companies of smaller size and from other areas. About 52 percent of the survey's respondents were from North America, with 35 percent from Europe, the Middle East and Africa, and 8 percent from Asia/Pacific. Approximately 43 percent of the respondents were from companies that had in excess of $1 billion in sales, which 42 percent came from medium sized businesses with between $50 million and $1 billion in sales, Only 15 percent of Aberdeen's respondents were from small businesses with under $50 million in sales. This stratification in the survey does not fit the way companies are distributed by geography or by company in the real world, of course. And this is often the case in studies performed by IT analysts. You will have to ponder for yourself how this might affect the conclusions you draw from those that Aberdeen draws from its own data.
The study and the accompanying report that Aberdeen put together was sponsored by a number of interested parties, including Ateras, which sells legacy application and modernization tools that convert mainframe applications to Windows .NET applications; IBM, which is trying to rule the SOA roost with its WebSphere armada of products and is more times than not the target of legacy modernization efforts by competitors; Clerity Solutions, which just bought a set of mainframe rehosting tools from Sun Microsystems; Micro Focus, which has COBOL application rehosting environments for mainframes and i5/OS-OS/400 servers; and Nexaweb, which has tools that build new Ajax and Java interfaces for legacy applications.
So what did Aberdeen learn from these companies? Lots, says Peter Kastner, vice president and research director for enterprise integration at Aberdeen. "From previous work, we have determined that users are embracing SOA with all of their might, and they believe that SOA is the panacea for all of their application integration nightmares," explains Kastner. Aberdeen estimates that application maintenance--making necessary tweaks to code to support end user or government reporting needs--and application integration--making applications share data in a secure and auditable fashion--comprise about 40 percent of the aggregate IT budgets in the world. Kastner says that while there are some differences among i5/OS, mainframe, and Unix platforms in terms of how they approach legacy application modernization and how they are deploying SOA technologies to address that issue, the biggest determinant on where a company is along the road to SOA is size, not platform. "This is not about the platform so much as it is about the size of the company," says Kastner. "Large organizations have IT processes that are the result of company organization charts that look like printed circuits, and they have 30 years of legacy applications."
Speaking very generally, Kastner says that about half of the companies polled are in the process of implementing Web services, which he has dubbed "SOA Lite" in the presentations he gives. They are implementing Web services--using XML, SOAP, Ajax, and other technologies--on a project by project basis, testing it out and seeing what the return on investment will be. "But by just mashing up corporate applications, they are not taking advantage of some of the excellent tools out there, and by working on a per-project basis, companies will end up with an architecture by accident. This is like the streets of Boston, which evolved from cow paths." Anyone who has tried to drive in Boston and seen the expense of the Big Dig project to rebuild transportation in the city knows this is not necessarily a good way to get an architecture.
Kastner says that Aberdeen is expecting modest spending on SOA technologies in 2006, and that spending on myriad SOA tools could increase significantly in 2007, as legacy application modernization efforts pick up steam and tool vendors compete to help companies do the modernization. But here is the kind of goal that everyone is after. One of the big New York banks that has been working through its legacy application modernization efforts in conjunction to a move to an SOA, had about 1,200 applications running on mainframe, Unix, Windows, and other platforms. These applications have millions of lines of code, and no one wants to mess with them. But this bank's SOA projects are done, it will have approximately 80 services that the IT department has created, which can be mixed and matched in any way managers and end users want to create new applications. No one will ever see a green, black, beige, or any other kind of legacy screen.
Legacy Application Modernization, By Platform
Everybody has their own reasons for wanting to modernize their legacy applications with an SOA approach, but they all have one thing in common: the SOA approach, generally speaking, is something that just about the entire IT industry--across all platform and application vendors--agrees is the way that applications have to be built.
The 32-page Aberdeen report is jam packed with data, and you can get a freebie copy compliments of IT Jungle by clicking this link provided by Aberdeen specifically for our subscribers. In the coming months, we will be asking you to participate in Aberdeen's studies as part of an agreement we have worked out with the consultancy. Under that agreement, we ask you to participate, and Aberdeen gets better reports and gives IT Jungle some extra content as a thank you to our subscribers. (There is no money involved, just in case you are wondering.)
Anyway, as the table below shows, all of the different kinds of shops that Aberdeen talked to--IBM mainframe, IBM midrange, and Unix--are focused on quickly getting changes done on IT systems--meaning the applications--and lowering their application integration costs. Mainframe shops are more concerned with integration costs because they tend to have the biggest data centers, the widest collection of IT systems, and the most complex applications. Unix shops are more interested in time to market with changes, basically because Unix was a platform that companies adopted to get rid of legacy platforms and applications. (Ironically, there are very few Unix shops with one Unix, one database, and no Windows or Linux.) Because of the integrated nature of the i5/OS and OS/400 platform, these IBM midrange shops, according to Aberdeen's surveys, are less interested in integration issues. The dominant non-IBM platform at AS/400, iSeries, and System i5 shops has been Windows for the past decade, and that has not changed. IBM has done an excellent job integrating Windows with its i5/OS and OS/400 platform--stopping short of a native implementation of Windows on its Power processors. Because iSeries and System i5 shops have been consolidating workloads onto these machines by their very nature, consolidation is less of an issue.
About a quarter of the i5/OS and OS/400 shops that Aberdeen talked to as part of this study said they had good results with their legacy application modernization efforts, and only 9 percent said they had poor results. The good news is that 46 percent of the iSeries and System i5 shops polled said they planned to extend or surround their legacy applications in order to modernize them, and they cited the economics of the platform and their loyalty to the i5/OS and OS/400 platform as the reason for why they would do this. Another 18 percent of those shops polled said they would modernize those applications on the i5/OS and OS/400 platform, which means going a bit further and implementing an Enterprise Service Bus, an SOA repository and services registration, and other best practices in SOA. That is a 64 percent retention rate. The bad news is that 11 percent said that they would modernize their applications off the i5/OS and OS/400 platform because of the better economics of alternative platforms, and another 25 percent said they would simply replace their legacy applications entirely and move to another platform to do their modernization.
Over on the mainframe, about 31 percent of survey respondents said they would extend or surround their legacy applications on the mainframe itself to modernize them, with another 19 percent saying they would do the extending or surrounding of these legacy applications using tools that ran on other platforms--presumably Windows, Unix, or Linux, which have vastly better economics. Another 35 percent said they would replace the legacy applications, and would do so either by replacing them with native mainframe applications or by using applications that run on other platforms. (For some mysterious reason, Aberdeen did not separate the two, and I have a gut feeling that there are very few mainframe shops throwing out their legacy applications for new code, while intending to keep the mainframe. Quite frankly, many modern applications simply are not supported on the mainframe--so it isn't even an option.) Another 15 percent of mainframe shops said they would modernize their legacy mainframe applications on the box either by hand or by using tools that automate the process. That leaves about 46 percent of the legacy applications staying on the mainframe, at least 19 percent half-way off the mainframe, and perhaps as much as 35 percent moving off the mainframe.
Unix platforms, by their very nature, are more alike than not, and as you might expect, the legacy application modernization issue is not hitting Unix platforms as hard as they are hitting the IBM mainframe or IBM midrange boxes. Unix platforms are also younger, and so are--in a very general sense--the programmers and administrators who make them work.
So it is not surprising to find that about 58 percent of the Unix shops surveyed by Aberdeen said they intended to keep their Unix boxes and Webify their legacy apps by extending and surrounding them right on the same Unix platforms. Another 16 percent said they would go even further, and modernize the applications using the full set of SOA tools on their Unix platforms. Only 16 percent of Unix shops said they would surround and extend their legacy applications on another platform (while keeping those legacy apps running on Unix, of course). Presumably, Unix shops look to Linux and Windows to run the tools that extend their Unix legacy applications. Only 11 percent of the Unix shops polled by Aberdeen said they were planning to replace their legacy applications and move off Unix at the same time.
When you add all of the legacy application modernization strategies up across the platforms, you get a pie chart that looks like this:
As you can see, extending and surrounding legacy applications is clearly the dominant choice that IT shops are making. And there are a lot of applications that need this work done. Rather than do this work, companies with mainframes and midrange platforms that have programmers and administrators who are themselves starting to head toward retirement are looking hard at their options. It is no wonder that such mainframe and midrange shops are looking to replace rather than extend their applications, or that 26 percent of all of the companies polled are taking the replacement path to legacy application modernization.
At first glance, it seems a little disturbing that more companies are not implementing the full-on SOA approach to legacy app modernization, which Aberdeen calls Technology Modernization in the chart above. That means picking a Web application middleware stack from IBM, BEA Systems, Oracle, or a handful of other middleware and application development companies and really architecting a company-wide SOA. This is undeniably the right way to attack this problem, and no one would argue it--at least not in a perfect world without time constraints, budget pressures, political platform proponents, and programmer preferences and prejudices.
Harte Hanks Buys Aberdeen
In a separate development, Aberdeen has been acquired by Harte Hanks, a $1.2 billion market research and survey company that is traded on the New York Stock Exchange. Harte Hanks is probably best known in the IT business for its acquisition of Computer Intelligence, a market research firm based in La Jolla, California. Computer Intelligence built one of largest and most sophisticated database systems for tracking IT buying patterns and installed bases of gear at hundreds of thousands of companies in North America and Europe. Harte Hanks says it will keep Aberdeen as a separate subsidiary and will retain its brand as well.
The combination of the Harte Hanks databases, which include some 680,000 companies and 2 million IT decision makers, and Aberdeen's surveys and analysis is intended to give Gartner (which ate Dataquest many years ago), IDC, and Forrester Research (which ate GIGA) a run for the money. The CI databases in particular will allow Aberdeen to have a much larger pool of survey respondents in its studies, which should make its results more statistically significant.
Harte Hanks will have 7,000 employees and 2,500 IT vendor clients after the Aberdeen buy.
Everybody Loves SOA, Aberdeen Survey Says
Major IT Firms Join to Standardize SOA