|
June 26, 2000, Volume 3 - Number 10
Brought to You By Application Servers
|
Until recently, nearly all enterprise software products were built to a specific operating system running on specific hardware. Product evaluations began with a quick look at the specs — On what platform does the hardware run? Unix? We can’t use it; we’re an NT shop. For new companies building a system from scratch, the operating system lock worked in the other direction, with their preferred software forcing the choice of operating system and often the hardware, too. Software vendors who wanted a broader market had to develop multiple versions of their products to accommodate all the major operating systems. But even developing Legato Networker for HP-UX, for example, isn’t enough. Which version of HP-UX do you have — 10.01, 10.10, 10.20, 10.30, or 11.0? The list of product versions becomes unwieldy, and expensive, very quickly.
Not only is it expensive to create multiple versions of software to run on every possible operating system, but the process of building to an operating system is costly as well. Software that sits directly on top of the operating system has to include many system services such as load balancing and fail-over. These features are essential in order for the software to run properly and the system to operate efficiently, but they are complex to code and time-consuming to test. Even though other software running at the same site may require similar system services, developers create software built to operating systems from the ground up, much like craftsmen fashioning a custom musket from stock to barrel.
Within the past year, the increasing use of application servers has fundamentally altered development environments and market strategies. Developers can now build one version of a software product designed to run on one specific application server. That application server will run on various platforms, slashing the amount of code product builders have to write and making operating systems essentially interchangeable. Also, application servers provide many of the system services that used to be part of software development, in a sense automating the skilled labor of writing load-balancing and fail-over code. Interchangeable parts and mechanically reproducing skilled labor changed the musket market and ultimately the entire factory system a long time ago. Now those same concepts are dramatically shifting software development paradigms and altering markets throughout the computer industry.
Vendors who have been quick to seize the opportunities application servers offer have already reaped the benefits of their foresight. For example, Blue Martini Software, established in 1998, embraced application server-centric development very early. The company brought its first product to market in a record-breaking nine months. Its Customer Interaction System is an enterprise-scale, turnkey e-business application built to run on the BEA WebLogic application server. That server in turn runs on Windows NT, Sun Solaris, HP-UX, IBM AIX, Compaq Tru64 Unix, Siemens Reliant Unix, SGI Irix, IBM 0S/400, and Linux operating systems. By using application server technology, Blue Martini created a product that is compatible with all those environments, but its developers only had to rush one version to market, not a dozen. A WebLogic license is embedded in the Customer Interaction System license, and customers have turnkey functionality on multiple platforms.
Avoiding runaway version development costs was only one benefit of Blue Martini’s approach. Because an application server handles essential system services, there is simply less code for Blue Martini developers to write and maintain. Developers master e-commerce, and they let BEA WebLogic handle the critical, but common, functions of scalability, availability, clustering, load balancing, fail-over, security, and communications with legacy systems.
Not only does building to an application server save time and money, it can also give a fledgling company added credibility. High-profile enterprise-level customers such as Harley-Davidson, Levis, and Gymboree had greater confidence in Blue Martini’s product because of the established reputation of BEA WebLogic. They knew that a product that had already proven itself in battle would handle mission-critical systems, so their purchase decision was less risky.
Name recognition also worked for OneSoft Corp. and its OneCommerce e-business product, although its development environment is slightly different. OneCommerce is built on Microsoft’s COM+ technology, an integral part of the Windows 2000 server platform, which is a competitor to Java-based application servers. While it is technically an operating system and not an application server, it provides some application server functionality, and the lines between the operating system and application server features are blurred. OneSoft benefits from the Microsoft name, and because Windows NT/2000 runs on hardware from a wide range of vendors, OneCommerce can reach a broad customer base. From the developer’s view, writing to a Windows 2000 system is similar to writing to an application server, because they are working with an application service, and not an operating system, interface. The application server element of the operating system handles the system services, and product development becomes quicker and more cost-effective.
The percentage of new software products built to run on application servers is increasing very rapidly, especially among Internet applications. The time-to-market value for vendors and the compatibility and performance benefits for customers make it a positive trend. Also, leveraging application server technologies lets application developers focus more on development of business logic, as they should, instead of on systems-specific issues. Because they can concentrate on a standards-based skill set, application developers can spend more time improving the quality of their applications. It’s quite certain that new software products will be built to application servers, and nearly all enterprise customers will eventually adopt a multitier architecture. Time will tell how that leap in technology will change the way those companies earn and retain their customers.
The old system development model let vendors lock in customers by developing proprietary, incompatible hardware, operating systems, and software products. The enormous investment of an enterprise computer system made major system changes cost-prohibitive. Now, adding a new layer of application servers to a legacy system essentially releases that lock. Although there are still significant compatibility issues, the potential exists for companies to implement the software they need on whatever system they want. Players in these changing markets have the opportunity to make positive changes toward standardization and interoperability. Unfortunately, now that one lock is released, most vendors of application servers, operating systems, and hardware are simply trying to change the locks.
Consider the Java-based application server market. Java has the potential to be a write once, run anywhere standard, although that hasn’t quite come true yet. Incompatibilities arise among older and newer versions of the Java Development Kit (JDK), and Java isn’t fully standardized. But rather than working to make Java a stronger standard, most application server vendors attempt to chain customers to their products by adding specialized and incompatible features. Any applications that use these nonstandard features will most likely not work on other application servers. Differences also exist in the ways application server vendors implement certain application services. All products are based on the relational model, which is currently a stronger standard than Java, but to port database applications from one relational DBMS to another is very difficult. Companies intentionally make incompatible that which is standard and compatible in theory in an attempt to hold on to market share.
Microsoft is also attempting to lock in customers, although it uses a COM+ server rather than a Java-based application server. Its strategy is convincing customers that the operating system and the application server are the same things — and in Windows 2000, they essentially are. The system provides both functions, but the lines between them are blurred. If vendors of the other Java-based application servers want to compete with Microsoft, they need to become more standardized as a group, rather than more unique.
As I stated earlier, Internet applications built on application servers will increasingly be the ones that provide the best value. Customers seek value, and they have to select products that will work on their systems. As product offerings continue to expand, people will make purchase decisions based on which products work on the most common, standard types of application servers. Currently, implementing Blue Martini’s product means we have to implement BEA WebLogic. A comparable product that runs on a few different application servers will be much more attractive to customers. Also, companies do not use only one type of pre-packaged software. They may have one package for e-commerce, another for system monitoring, and another for managing backups. As new application-based products arrive on the market, companies will select those versions that are the most broadly compatible, and will work with the application servers they have already implemented. Software that plays well with others is going to emerge as the market leader.
The operating system market will also be altered. Because a startup company’s preferred software packages will run on application servers, and thus be compatible with any platform, compatibility issues will no longer force the selection of an operating system. Assuming that costs and performance are roughly comparable among all platforms, then the CIO and his or her team will most likely choose to implement the system they already know. But what motivated those team members to learn that particular system’s development tools in the first place? Turnover is high among IT professionals, and programmers are constantly developing and maintaining marketable skills. If professionals choose to learn the workings of a system that is extremely specialized, that means their expertise has a limited market. Most would rather learn system tools that are more common and standardized so they can use their skills on a wide range of systems. Because IT staff will generally be more familiar with the most standard operating systems, they will tend to implement those systems more often. Customers will create a standard-seeking demand spiral.
The 21st century computer industry is certainly more complex than the musket market of the 1800s, but basic economic truths are proven repeatedly, whether it’s the industrial revolution or the information revolution. Manufacturers and customers both benefit when there is competition within the framework of strong industry standards, and products win when they are most broadly useful, not most blatantly quirky. Now it’s a race to set those standards.
Bin Li, Ph.D., (bin.li@marchfirst.com) is team lead for MarchFirst’s (formerly Whittman-Hart & USWeb/CKS) New York programming services group. He received a Ph.D. in computer science from New York University, and was formerly project development manager for Citicorp.
RESOURCESBlue Martini Software Inc.:www.bluemartini.comOneSoft Corp.: www.onesoft.com |
 |
Most Popular This Week
IE Weekly Newsletter
Subscribe to the newsletter
| |
|
|