In-Depth
Distributed object computing: The brave new world
The big competition for Corba then was the Open Software Foundation (OSF),
with its Distributed Computing Environment (DCE) standard. Meanwhile, Microsoft
was not really in the DOC game at all, other than making some vague statements
about how their new operating system still in development and, code-named Cairo,
would finally get some real object-oriented features, including the yet-unspecified
Component Object Model (COM).
A long way baby
No one can seriously deny that DOC has come a long way since then. This year,
the OMG has grown to over 700 members, and a fully interoperable Corba, joined
at the hip with Java, is making its debut on the Internet. The OSF has nearly
vanished, and Corba companies like BEA Systems Inc., Sunnyvale, Calif., Iona
Technologies Inc., Dublin, Ireland, and Visigenic Software Inc., San Mateo,
Calif., that are selling Corba technologies are becoming the new high-tech darlings
of Wall Street.
Meanwhile, Microsoft has released Distributed COM (DCOM), and is touting its
new object Transaction Manager, previously code-named Viper, which even Microsoft
now describes as an an Object Request Broker (ORB), just like OMG's Corba. Oddly
enough, DCE is now mainly an under-the-covers way to implement DCOM, and Cairo
is once again just a big city in Egypt (or, if you prefer, a small town at the
southern tip of Illinois).
Now the question is no longer whether DOC will become the backbone of global
distributed computing, but just how -- and how soon. Vendors and consultants
are frantically trying to put together their DOC strategies, while end users
are tire-kicking Corba and DCOM in droves and asking questions about the ROI
on large-scale DOC projects. (See Fig. 1.)
Fig. 1: ORB usage -- then and now
The world of distributed object computing (DOC) has expanded from
making the OMG's Corba and the OSF minimally interoperable to the
now interworking Corba and Microsoft's DCOM.
Source: Genesis Development Corp.
|
|
Yesterday's news
But that's just yesterday's news. The really late-breaking story is just starting
to gel as a whole new approach to DOC is shaping up around the not-so-new concept
known as 'business objects.' A more proper name is probably 'domain objects,'
but the term 'business objects' better catches the spirit of this trend and
has been in use throughout the industry for longer.
In short, a business object is a self-managing component that represents a
key business entity or process intended to be (re)used all over an enterprise,
or even across related enterprises. Conceptually, the notion of a business object
is the Rosetta stone of business computing -- it's the first time that the same
concept can be used to both describe a piece of the business and an executing
piece of the information system that supports that piece of the business.
This advance is a tremendous help to those folks who have been struggling
with business process modeling and (re)engineering. Previously, businesses that
engaged in such modeling exercises reached a no-man's land when they tried to
convert their reengineering results into new software. The paradigms for expressing
business models and software models were miles apart, and only a few very expensive
and complicated CASE tools even tried to bridge the gap -- with very limited
results.
Fig. 2: Mapping ORBs and business models
In this approach to logically mapping ORBs and business models,
the output of business modeling is a set of high-level business
components configured to run directly as distributed objects on
an enterprise ORB.
Source: Genesis Development Corp.
|
|
In the new approach, the output of business modeling is a set of high-level
business components, which can then be configured to run directly as distributed
objects on an enterprise ORB. After that, any business event in the real world,
like the placing of a new order, simply becomes the construction of a new business
object (an 'order') that then assumes an identity and life cycle of its own,
supported by the enterprise ORB. (See Fig. 2.)
This represents a huge step up for DOC, which previously was positioned as
specialized plumbing for exotic forms of client/server computing.
This earlier positioning relegated DOC vendors to selling their wares mostly
to distributed computing "techno-dweebs" working on a few very leading-edge
projects. With business objects, however, DOC suddenly becomes a major technology
enabler for business, with a broad target market leading right into the office
of every major CIO and CEO.
Even better, the upcoming blitzkrieg of DOC-enabled business objects already
has its path cleared in advance by today's number one technological buzz-bomb,
the Internet. Basically, what the Internet and its variations -- Intranet, Extranet
and the like -- have done is to knock down almost all the remaining artificial
and realistic barriers that once kept everyone stranded on their own proprietary
'islands' of computing.
Now, every I/S manager is on notice that the business may demand that almost
any system, even those dark and dirty legacy dungeons, may one day need to be
seen from almost anywhere else in the world to meet the insatiable demand for
electronic commerce. Just how insatiable is this demand for global accessibility?
To find out, take a moment to read the neighboring excerpt. (See "A
Virtual Fable.")
But, as we all know, today's Internet technology is not really up to that
kind of demand. Today, most companies are still struggling just to develop and
maintain relatively simple Internet sites, while maintaining appropriate security
and integrity for their underlying systems. They are doing this largely because
the most popular Internet protocols, such as -- HTTP and CGI, etc. -- were never
really designed for the kind of applications now being demanded by business
as commonplace on the Internet, much less full-blown global electronic commerce.
However, what is a meltdown for conventional Internet techniques is simply
critical mass for DOC. Real DOC protocols like Corba and DCOM, combined with
business objects, have the potential to handle not only the peak loads of today's
Internet, but also the truly high-volume, high-performance demands of tomorrow's
Internet successor -- the global electronic commerce machine we have begun to
call the 'Meganet.'
Mining the Meganet
How will the brave new world of the DOC Meganet work? Businesses would make
their basic commercial interfaces -- checking inventory, estimating costs, accepting
orders and payment and so on -- visible to potential clients through business
object interfaces described in some kind of standard business object interface
definition language (IDL), hosted by a standard ORB. The ORB would also host
the implementations of those business objects, but these implementations would,
of course, remain completely opaque to any clients.
Each hosting ORB would be connected through a standard inter-ORB protocol
like OMG's Corba Internet Interoperability Protocol (Corba-IIOP) to all other
ORBs through the Meganet. This would allow any ORB-enabled client to access
any ORB-based business object anywhere in the world, at any time, based, of
course, on suitable authorizations. (See Fig. 3.) The various
ORBs on the Meganet -- Corba, DCOM and/or their successors -- would handle the
necessary global request brokering, security, transaction management, concurrency
control and exception handling.
Fig. 3: ORBs and business objects coexist
Through the DOC Meganet, each hosting ORB would be connected
through an inter-ORB protocol, like OMG's Corba Internet Interoperability
Protocol (Corba-IIOP), to all other ORBs.
Source: Genesis Development Corp.
|
|
The business implications of this are staggering. The simple-minded Internet
servers of today only talk directly to end users with visual browsers, plunking
along page by page and they have to be gerry-rigged to support even simple-minded
business transactions. On the Meganet, however, servers will be serving up real
distributed objects, which can engage in arbitrarily complex transactions with
other objects. This puts the kind of business-to-business electronic commerce
described in "A Virtual Fable" much closer.
If you think the Meganet itself is still far off, guess again. Within a few
months, every Netscape browser and server in the world will support Corba-IIOP,
not to mention the platforms of almost every other major vendor, including Sun
Microsystems Inc., Mountain View, Calif., Novell Inc., Provo, Utah, and IBM.
Meanwhile, every Microsoft software platform will support DCOM. But, lest you
worry that a market fight between Corba and DCOM might leave the Meganet in
tatters, be forewarned that there is already a working, commercial standard
for Corba/COM inter-networking, supported by both that will handily support
the coming Meganet.
In any case, the certain winner of any upcoming 'battle of the business object
Meganet' will be the businesses of tomorrow, and with them, DOC. While there
are certainly many details yet to be ironed out, there is absoluutely no question
that there will be such a Meganet, and that it will run on some kind of DOC
engine -- Corba, DCOM or some hybrid successor. If so, this means that, in short
order, the commercial battle lines will be moving elsewhere. Where? Some experts
believe they are in fact already moving four-square towards business objects
themselves.
Making business objects viable
To learn how business objects will cavort across tomorrow's Meganet, there
is probably no better example than the OMG. The OMG may not be the only place
that DOC and Meganet history is being made, but, with its open process, the
OMG standards group is definitely one of the best places to see it unfold.
Last year, the OMG opened its doors wide to corporate end users, who are flocking
to form new working groups and task forces to define global DOC standards in
such business domains as finance, telecommunications, insurance, manufacturing,
transport and medical systems. This is a huge step forward for the OMG, which
previously limited its efforts to standardizing generic facilities like ORBs,
transaction managers, security services and event managers.
In addition, the OMG is currently involved in developing at least two new
standards that are likely to have a major impact on tomorrow's Meganet -- a
Business Object Facility (BOF), and a unified Object-Oriented Analysis and Design
(OOAD) interchange specification. Together, these new standards, both due out
by the end of this year, will form the backbone of both design-time and runtime
support for frameworks of distributed business objects, including those based
on the domain standards defined earlier. (See Fig. 4.)
Fig. 4: The OMG's structure
These new standards will form the backbone of both design-time
and runtime support for frameworks of distributed business objects.
Source: Genesis Development Corp.
|
|
Let's first look at the BOF. In short, the goal of the BOF is to make it a lot
easier to host over Corba the runtime versions of the kind of domain object models
that most business users want. It is no secret that Corba is a very generic standard
designed to support virtually any kind of DOC problem. Thus, if you want to host
a straightforward business object like SalesOrder directly over Corba, there could
be dozens of complex decisions to make about transactions, security, persistence
and so on. Such a scenario also requires dealing with many interfaces.
However, a BOF can automate much of the housekeeping chores that would be
hand-coded in Corba. Corba would be automatic. In other words, analysts working
in some particular line of business could define a high-level business object
model, and then use the BOF to host the runtime business objects associated
with that model with minimum effort. The BOF itself would use Corba services
to manage those objects in a very consistent way.
Since the BOF will be hosted over Corba, and since Corba will soon be all
over the Internet, the facility BOF could rapidly become the standard business
process for publishing global electronic commerce interfaces. Once these interfaces
become standardized, the makings of the Meganet will truly arrive, and the 'Virtual
Fable' will be a fable no more.
Using the BOF will be even easier when OMG completes the standardization of
interchange formats among various OOAD tool vendors. These tools typically provide
a means for capturing domain object models, displaying the models graphically
and maintaining various forms that support documentation. Some tools can also
generate and/or reverse engineer C++, Smalltalk and Java code. Until recently,
each tool stored their models in a highly proprietary format that made exchanging
or sharing models among different tools very difficult.
OMG is now involved in a major effort to create a notational Rosetta stone
to be supported by every OOAD tool. (See "Converging on OOAD Agreement" by Viktor
Ohnjec, Application Development Trends, February, 1997.).
This is good news in any case, but it is particularly convenient for the BOF.
Once all OOAD tools can read and write in a common format, all the information
can also be shared with the BOF, which can use it at runtime to manage the various
business objects that are created. Such a standard could become another point
of convergence for OMG and Microsoft. That's because Microsoft, which previously
had no apparent OOAD tools strategy, recently announced a strategic alliance
pact with Rational Software Corp., Santa Clara, Calif., a maker of several object-oriented
tools. In addition, Microsoft has agreed to support Rational's proposed OMG
standard for OOAD tools.
Distributed-object computing opens up a host of new activities in addition
to the technology areas mentioned (DCOM, Corba, BOF, OOAD, etc.). Specifically,
vendors are beginning to move toward integrated solutions. (Rational is working
with Microsoft and has acquired several companies in recent months, most notably
Pure Atria, Sunnyvale, Calif., and Software Quality Automation, Burlington,
Mass.) Meanwhile, Platinum Technology Inc., Oakbrook Terrace, Ill., is integrating
myriad acquired solutions. Sterling Software Inc., Dallas, has recently acquired
Texas Instruments Inc., Plano, Texas, whose high-end tools will be merged with
the offerings from Sterling's Atlanta-based development tool unit. Although
many of the buzzwords change -- like business objects and component-based development,
etc. -- the expectations of DOC remain.
With the move to components and business objects, however, some suppliers
are also introducing development tools that can leverage existing legacy systems
and approaches in distributed business applications. Cambridge, Mass.-based
Riverton Software Corp.'s HOW product, for instance, lets developers custom
build -- or adopt and extend -- business objects and technology components to
produce applications that are partitioned along presentation, business logic
and data access boundaries. Such tools were unheard of even a couple of years
ago. The likelihood today is that more and more of these tools will emerge as
BOF and OOAD maturity increases.
As the details behind the BOF and OOAD specifications become solidified later
this year, a path will have been be cleared for the various industry-specific
groups to deliver specifications for their respective domain object models.
Those models, specified in the OMG standard OOAD notation, can then be expected
to run over an OMG standard BOF with appropriate ORB support. When BOFs become
widely available, and as the domain models become mature, the incredible commercial
possibilities of DOC distrobuted object computing and the Meganet will start
to become obvious to everyone.
The world is changing, and DOC is tracking that change.