Application servers have the dubious merit of being at the centre of the latest wave of industry hype, but there are now so many offerings on the market that it has become difficult for users to know what to choose.
While many are not based on industry standards, still others are not even what they claim to be, so how should customers go about finding the right product to fit their application needs?
Mike Gilpin, vice president of consultancy Giga, provided a definition of what an application server actually is at the consultancy?s Middleware Choices conference in San Francisco this week.
?Application servers are the next evolution of enterprise middleware and comprise a mixture of Web, Corba and distributed transaction processing technology. They enable distributed applications to be built, deployed and managed more easily than was possible before and their emergence has synchronised with the emergence of Java,? he explains.
Some of the confusion in the industry stems from the fact that the market for these products is growing so quickly. While client/server took 10 years to develop, it will only take about five or six years for the application server market to mature, Gilpin believes.
?While about 15 per cent of application development projects are based on distributed components now, this will increase to 25 per cent next year and rise to 50 per cent by 2000. However, there?s still a big legacy of applications and by 2003, the distributed applications portfolio will still only make up one-third of all applications out there,? he forecasts.
But, unlike most mature industries, there will not just two or three vendors in a steady state, he continues. There are more likely to be six or seven that can make a good profit out of this market because it is quite diverse in scope.p> Unsurprisingly, Internet based business is the number one application, followed by companies involved in merger and acquisition activity, particularly in the banking sector. In such a situation, an application server can act as a hub tying a number of disparate systems together.
Organisations are also finding that they need to significantly rearchitect their systems to cope with the increasing demands of globalisation. This means they can no longer take their applications down for four hours during the night while they undertake batch processing, but need to provide 24x7 availability. Introducing an application server into the middle tier can provide more flexibility.
Next on the list is the desire to improve customer service and customer relationship marketing. An application server enables users to enhance and add new systems and processes such as data warehousing more easily to create a single unified view of the business and the customer to help call centre staff do their job more efficiently.
Finally, organisations are also using these products to integrate their supply chain and reduce disintermediation because many systems are no longer just being used internally, but being exposed externally to customers and suppliers.
As a result, says Gilpin: ?The macroeconomics of the middleware market means that this technology is addressing problems companies have as regards tying disparate systems together and creating a fabric to hook everything together. The market is evolving to higher value solutions and will in future support standards that will make middleware more ubiquitous than transaction processing monitors.?
But for the moment, the most important thing to bear in mind when purchasing an application server is the need for reliability, availability and scalability, because the products are part of the server infrastructure.
Apart from that, Gilpin says, there is no individual item that people care about to the exclusion of all else.
However, in generic terms, they do want their systems to be easy to implement and manage, particularly because, up till now, it has been too difficult to deploy and administer distributed applications.
They also want their application servers to be able to connect to new packages and to transaction and database oriented legacy applications, so that they knit into the overall business process. Offerings should be standards based, supporting Java, Corba and the like, to enable portability, and be easy to develop to and reuse.
Unfortunately, Gilpin claims, server development tools are relatively primitive at the moment. Their automation levels are quite low and overall the offerings are not as good as they ought to be, which makes it important for a given application server to be popular and have the broad support of ISVs so that packages are readily available.
This means that, in future, there will a limited number of application servers that people want to have, he adds.
Also, as a rule of thumb when choosing their strategic middleware platform, users should pursue an architecture driven strategy, which should be as broadbased as possible in standards terms. If multiple application servers are deployed, it is best to choose a family that is based on similar standards or they may not work together, Gilpin warns.
Any decision should also reflect which packages users are most likely to run on their application server. They should not go for catch-all offerings because different servers are optimised for different applications.
?If you want high availability and reliability, use one of the Unix based products. We have limited expectations of Microsoft Transaction Server (MTS) until Windows NT 5.0 is out of the stable in 2000-2001. Although Microsoft Com and MTS are fine themselves, they are currently limited by NT 4.0. I?d also be cautious about using [Sun] Enterprise Java Beans until this is proven by the end of 1999. There?ll be a lot of change before then,? he advises.
So, where do Web application servers fit into this equation?
Philip Costa, a Giga analyst, explains: ?A different kind of discipline and architecture is required to build Web applications and this is where middleware and application servers fit into the Web."
"Early Web sites were developed by marketing departments for publishing and broadcasting, but as we move to the next phase, we?re seeing significant changes,? he says.
?Web applications are now focused on Extranets and ecommerce, and the next phase will be customers interracting with their accounts, which demands a different paradigm. But while tools and servers are converging on a common architecture, not all solutions are equal and Web application servers will require integration with enterprise middleware,? he continues.
As a result, a Web application server needs to support diverse clients based on such languages as ActiveX, Java and XML in future, to provide a richer way to access application data.
They also need to be able to hook into older installed systems as well as new ones, however, to access corporate data. ?But the challenge of the last 12 months and the next 18 months will be to make Web application servers scalable, reliable and manageable,? Costa says.
In addition, users need to look at their existing skill sets to ensure they can make the technology work in their organisation, and should also limit the number of different types of application servers they introduce because, in most cases, they have only limited resources.
Customers should also look for a platform that is supported by a range of development tools, Costa advises, because they need to get their applications to market quickly. ?There?s a narrow timescale for delivery, so a project should take no more than three months or it?s a lost opportunity,? he says.
In essence, however, there are two main classes of application server - transaction processing monitors that have developed into application servers and HTML based general servers that incorporate robust features. These still have a long way to go, however, Costa warns.p> This sector, according to Costa, is dominated by vertical application servers for ecommerce and store front packages and is not yet out there as a general platform. However, the more mainstream vendors are starting to merge their publishing and transaction platforms to create more general purpose offerings, which should start to appear over the next six months.
Although the major client/server vendors had been caught off guard by the demand for application servers, Costa reckons they will gain lost momentum in early 1999 because the world still needs to support transactions and business logic. This means they will move more aggressively over the next six months, in particular buying up new technology and beefing up existing products.
Including a 15-inch Intel Core-powered device weighing less than a bag of sugar
Tuomo Suntola's ALD technology extended Moore's Law, but was only adopted by chip-makers in 2007
Trump proposes a $1.3bn fine and a round of firings to un-bork ZTE
Findings could mean new optical frequencies to transmit more data along optical cables