In-Depth
OTMs - ORBs for the enterprise?
Middleware is just plumbing, claimed most software experts during
the 1990s. While some middleware variations were described as smart, they were plumbing nonetheless. However, evolving
technologies may be changing that truth. Chief among these is the object transaction monitor.
Object transaction monitors (OTMs) represent a rung on the evolutionary ladder of component development and deployment.
It is a ladder that is being climbed so quickly that experts cannot even pause to agree on nomenclature -- is it
Object Transaction Middleware, Object Transaction Manager, Object Transaction Monitor or Object Transaction Server?
While the standard terminology is still debated, the core concept is that of a server-side marriage of object-to-object
communications as represented by object request broker (ORB) technology, the reliability and scalability traditionally
found in transaction processing monitors (TPMs) and the management functions required by applications composed
of distributed objects. To further confuse, the technology typically, though not always, runs in a Web-enabled
environment.
The OTM is a natural, evolutionary step. Raw object request brokers are, as stated, just plumbing. They provide
little more than a means for object-oriented communication. In response to customer needs, vendors have added services
to their ORBs, and, sometimes in concert, the Object Management Group has added new standard services to its CORBA
standard specification.
The requirement for event management, so common in the transactional enterprise, has proved to be just as necessary
to firms migrating to object-oriented software technology. "If an ORB is in the enterprise, it is, by definition,
an object transaction monitor," said one company representative from among the OTM ranks. In fact, for many
IT shops the first crucial service they require is an object transaction monitor.
Vie and other views
Although the company does not offer an object transaction monitor, middleware maker Vie Systems, Lyndhurst, N.J.,
has a perspective on the newly evolving OTM market segment. Arising out of the financial service ferment, Vie makes
what Business Development VP Scott Saber
describes as a middleware system for application-to-application message transformation and intelligent message
routing. The parallels he draws to describe OTM are the natural parallels to the database transaction monitor category.
Object implementations have proved that they need a central monitor, said Saber, in the same way databases have
a central monitor for sending out events and monitoring actions associated with those events. For example, he said,
if objects are used in a series of events that comprise a transaction and one event in the series fails, you want
to be able to roll out the entire transaction. The analogy to the world of database transactions is clear and present.
You must maintain the order of events.
Saber states the essential problem that object transaction monitors seek to address: "Ordinarily, in the object
world, objects can fire at any time." And a misfiring in an established order of execution can be critical.
Thus, ORB vendors such as IBM; Iona Technologies Inc., Cambridge, Mass.; Expersoft Corp., San Diego; Inprise Corp.,
Scotts Valley, Calif. (which gained an ORB via its purchase of Visigenics); BEA Systems Inc., Sunnyvale, Calif.
(which gained its ORB via purchase of Digital Equipment Corp. technology); and others have come to offer transaction
monitoring of objects.
From outside the CORBA-compliant camp comes industry heavyweight Microsoft. The firm's COM+ (now combined with
the Microsoft Transaction Server) can also be described as an OTM. But COM+/ MTS, or another of the CORBA-based
OTMs, might just as easily be described as an application server or transaction processor -- it depends on the
marketing hat or direction of the industry winds.
For his part, industry analyst Ed Acly sees the outlook for OTMs (which he classifies as Object Transaction Servers
or "OTSes") clouded by the arrival of the application server. This hot new technology is a collection
of services that is sometimes just a stone's throw away -- at least in physical terms -- from a Web server. "Not
long ago there were people who thought they had the 'latest thing' in [object-oriented technology] and CORBA,"
said Acly of Framingham, Mass.-based International Data Corp. "Now they think the application server is the
hottest thing."
In any case, the newness of all this may be overstated. "We just have this fairly generic problem of transaction
management. It didn't start with ORBs. The services that have been added to [basic] ORBs are needed to make ORBs
real," he said.
These days, the application server, as Acly rightly points out, is usually a Java phenomenon. "There seems
to be a move to use Java on the top and CORBA [for communication] on the bottom," he said. This is because
CORBA is relatively language neutral and can better support multiple languages and, in turn, legacy systems. OTM
vendors may be at a disadvantage, Acly indicates, to the degree that they require CORBA IDL programming. Many wish
to do new programming with Java and have CORBA plumbing underneath, he said.
Are OTMs -- or, as Acly says, OTSes -- better defined than application servers? "With application servers
right now it is like an inkblot [Rorschach] test," said Acly. In other words, people see what they want to
see in an application server. "OTSes are a little bit better defined," he added. "The OMG has specified
what these services are. But an application server is less clear.
"What do you want it to be? SAP calls some of its products 'application servers.' Tool makers describe their
products as application servers. Middleware 'guys' call their products application servers," said Acly.
One may surmise that, if they are only slightly more defined than the application server, OTMs at least bear consideration.
Nomadic nomenclature
Whatever the nomenclature, it is becoming clear that OTMs reflect the changing character of middleware. Actually,
a more precise statement is that component development standards (such as COM, CORBA and Enterprise JavaBeans)
are driving changes in component deployment environments (represented by what are called OTMs for now). Development
models are essentially standardized contracts between developers and the respective models that control some aspect
of object behavior, such as how to invoke an object. Deployment tools then execute that contract at runtime.
If standards are utilized, the controlled behavior is identical in the runtime environment, regardless of the deployment
tool or environment. Carrying this concept to its natural conclusion, deployment tools can provide a development-independent
runtime environment. Today, most IT organizations can only imagine selecting a development suite for its value-add,
such as visual assists, application wizards and the like. Selecting a deployment environment on its own independent
merits, such as proven scalability or ease of administration, is another dream of IT. While the dream is still
unfulfilled, experts say the momentum is shifting.
Consider for a moment that browser-enabled clients can be created using different authoring tools that interface
to server-side logic. Java applications can now be written to run on multiple Java Virtual Machines (JVMs). And
the probability is high that CORBA-compliant objects written in different languages, using different development
tools, can communicate with each other. Standards are the common enabler -- standards and the market's increasing
demand for compliance to these standards.
An emerging concept heard from middleware gurus is that components execute within the context of a container. This
is already a reality on the client, where applets are running in the context of a browser-enabled Web page container.
Enterprise JavaBeans (EJB) and COM extend the concept to a server-side rendering with the "middleware"
as the container. This notion is described as a deployment platform by Mark Carges, BEA Systems' vice president
of advanced product development. The deployment platform carries out the development contract and provides the
infrastructure for delivering the "standardized" services.
Again, the industry is struggling with the terminology for this concept. OTM is one synonym, application container
is another. Application server is perhaps the most popular term today, although it has become the "catchall"
phrase for anything that moves on the server. OTM and application container have become the more common terms among
middleware-oriented vendors, while application server has the nod from vendors selling integrated development and
deployment facilities.
The industry also appears to have come full circle with its naming conventions. Consider that CICS provides a "container"
for most of the world's consequential transactional applications. And, observers note, CICS servers were once called
"application servers" -- a name that continues in the IBM lexicon.
"Clear and consistent terminology has never been one of the industry's strong suits," said John McGuire,
senior product manager for enterprise products at Iona Technologies. "On the vendor side, we seem to change
words at the drop of a hat. But the constant is the customer problems that we're trying to solve. Those haven't
changed, and addressing key deployment issues is what this market is all about."
Different, but converging
As the market converges, vendors with vastly different perspectives (and strengths) are taking aim at the component-oriented
bull's-eye. Vendors with such products can be divided into a variety of segments, including:
3Middleware-centric vendors whose general strength is developing middleware technologies. Some, such as BEA Systems
and Hitachi America Ltd., Brisbane, Calif., come from the TPM arena. Others, like Iona, Expersoft and Visigenic
(prior to its acquisition by Inprise) sprang into existence riding the CORBA ORB wave. Still others evolved from
the Web server environment, where the value-add centered more on deployment than development. Finally, there are
specialist companies that concentrate on a focused middleware capability, such as messaging.
3Development-centric tool vendors, with strong development acumen, fill in the deployment side with a combination
of acquisition and/or homegrown middleware. Inprise first came to the market with Borland development tools, added
ORB functionality, and completed the ITS application server with homegrown facilities -- such as load balancing
and application management -- and technology acquired from Visigenic. Some Web Server vendors also fit into this
category, though most also sell respectable middleware.
3Megavendors such as IBM, Microsoft, Oracle Corp., Redwood Shores, Calif., and Sybase Inc., Emeryville, Calif.,
which have the wherewithal to build strong products in multiple categories, comprise the final segment. IBM's newly
christened WebSphere -- probably the broadest-ranging vendor effort -- is described as an amalgamation of ComponentBroker,
TXSeries distributed TPMs, VisualAge development tools and San Francisco component frameworks. Oracle offers the
Oracle Application Server and struggling Sybase, if it can deliver on its promise, has the makings of a strong
cross-model server based on its Jaguar CTS. Microsoft has potentially changed the rules by moving its OTM functionality
into NT.
With a fluid scorecard that could rival Abbot and Costello's 'Who's on First?' skit, the market is awash with partnerships
and acquisitions. The alliances include Iona with IBM's TPM groups, Forté Software Inc., Oakland, Calif.,
with the Inprise ORB unit, and Hitachi with the Inprise ORB unit. The recent list of acquisitions include Kiva's
purchase by Netscape, which in turn has agreed to be acquired by America Online Inc., Dulles, Va.; BEA buying WebLogic;
Sun Microsystems' acquiring NetDynamics; and PostModern Computing's sale to Visigenic, which in turn was bought
by Inprise.
At this point in the convergence, vendors are remaining closely linked with their heritage. Yefim Natis, vice president
and research director at the Gartner Group, a Stamford, Conn., consulting firm, shares his perspective of the market.
"A maturing IT community -- one that cut its teeth on client/server and is now moving into object-based development
and deployment -- has come to expect a certain level of capability and robustness from its Web-enabled tool suites
and middleware," he said. "OTMs are one step up in the application stack from TP monitors and ORBs. There's
a lot of essential functionality there, but much of that functionality, like transactional integrity, has become
a given in today's market. The problem has been that a lot of the middleware products have just been too technical
and hard to use."
While each analyst group takes its own cut at product and application taxonomy, Natis sees two classes of OTM development
efforts emerging: systematic and opportunistic.
Systematic development efforts are strategic in nature, have a long-term effect on the corporation, are extremely
well funded ($300,000 and up) and typically have a full-time staff, according to Natis. The purchase decision in
these instances is slanted toward vendors with a middleware heritage and a proven track record. "Organizations
involved in systematic development are buying a long-term partnership with their vendors," he said. "Of
course, the products have to deliver the goods, but there's also a very strong focus on the long-term relationship
and service aspects." Here, Natis sees companies like BEA, IBM, Iona Technologies, Kiva (now Netscape/ AOL)
and WebLogic (now BEA).
Opportunistic development efforts are much shorter projects that aim to take advantage of an immediate business
opportunity. With these relatively low-cost projects ($30,000 and up), IT organizations are under intense time-to-market
pressure, and looking for short learning curves and high productivity. Here, Natis said, users lean toward product
suites with strong development tools and strong debugging/testing facilities. This vendor list includes Bluestone
Software Inc., Mount Laurel, N.J., Forté Software, Sun's NetDynamics unit and SilverStream Software Inc.,
Burlington, Mass.
Even though the market is still differentiated, Natis sees a coming convergence, to the point of commoditization.
"We are covering the same ground as the mainframes," he said. "At one time, there were something
like eight different TPMs on the mainframe. Now, for all intents and purposes, the market's consolidated down to
two -- CICS and IMS. We're also moving toward a two-option deployment environment. We believe one will be Enterprise
Beans -- a synthesis of EJB and CORBA -- with a programming model closely resembling EJB [but multilingual] and
with the internal architecture closely resembling CORBA. We believe the other will be COM+."
The real value proposition
Tokyo-Mitsubishi International (TMI) Plc is the London-based international capital markets subsidiary of The Bank
of Tokyo-Mitsubishi Limited. As a core operation within the bank's expanding global investment banking network,
TMI's Web site provides a wide range of international capital markets services to governments, central banks, supranationals,
corporations, financial institutions and portfolio managers.
We constantly read how financial institutions such as TMI use new technology to fuel external competitive initiatives.
But this institution is also using New Age technology to streamline its internal processing, not merely reface
existing applications. As a partner in its technology-based endeavors, TMI chose London- and Coventry-based Parallax
Solutions Limited -- a consultancy and development firm that has attracted customers such as BMW AG, Chase Manhattan,
Ford Motor Co. Limited, Lehman Brothers, and Rolls-Royce & Bentley Motor Cars.
Mark Wimpory, Parallax technical architect, virtually lives with TMI. Says Wimpory, "We're here to help TMI
design an architecture, and implement standards and applications that will take them into the next century. As
part of that effort, we're moving to open standards with object-oriented development and the likes of Java and
CORBA."
For TMI and Wimpory, a single-vendor solution from Inprise proved most attractive on an application currently under
development -- a Straight-Through-Processing application that streamlines the coordination between the front, middle
and back-office functions. Commented Wimpory, "We all know products based on these technology standards are
supposed to work together. After all, that's one of their real value propositions, isn't it? But reality can be
quite different from the proposition, and we didn't want to land in a situation where different vendors were saying
'Not my problem, it must be them.' We've all been there, and it's not an easy situation to resolve."
The fact that Java has been a moving target has only exacerbated this situation. But things might be getting better.
"Since [Version] 1.1, Java hasn't taken any more 180-degree turns. Maybe some 20-degree shifts, but no major
about faces," noted Wimpory. "And if you look at EJB, that's also stabilizing. While I'm not yet ready
to mix-and-match different tool suites, this is where the industry is heading, and it will mean a tremendous advantage
for IT organizations and companies like Parallax."
With regard to application management, Wimpory echoes the sentiments of virtually every architect interviewed for
this article. "Application management is a must-have for these CORBA-based systems. It's not an option,"
he said. "Can you imagine building an application, handing it over to the administrative group and then telling
them there's no way to manage the application? Some development teams might want to do that, but I don't."
Focus on management
As Iona Technologies' McGuire said, OTMs, or whatever one calls them, are about deployment -- and that tenet is
significant, as it was with client/server, the Web and objects. The first wave of technology focused on development,
with virtually no regard to administration and management until organizations began deploying systems. Then came
a hard realization of what was missing.
Perhaps we are finally learning the lessons of history, for OTMs focus on deployment. Companies like Forté
can attest that developing an application is only a first step. While competitors still swipe at Forté's
proprietary object model, the company was one of the first to figure out that deployment and management of distributed
objects is critical. The problem has not changed, notes McGuire. "Objects and components can be a deployment
landmine," he said. "How they're partitioned and distributed can seriously affect system performance.
You need a management facility that can identify bottlenecks."
Vendors appear to understand the issue, and thus users are seeing an emergence of facilities for managing an application
made up of distributed objects. This is not system management on the scale of Unicenter TNG from Computer Associates
International Inc., Islandia, N.Y., EcoSystems from Compuware Corp., Farmington Hills, Mich., or IBM's Tivoli unit.
These facilities are better dubbed application management, which is one step up on the management stack from system
management tools. Except for object-oriented 4GL vendors that invested in proprietary application management facilities,
object-oriented application management is in its adolescence.