FYI. Recent Comm Week article
- Date: Mon, 10 Aug 92 11:54:30 +0200
Regards,
Bernhard Stockman.
------- Forwarded Message
IN SEARCH OF OPEN SYSTEMS
by Donne Pinsky
In the real world, an open system is any system that works
with others.
Still waiting for Open systems Interconnection to fulfill its
promise, business users have turned to a mix of internetworking
hardware and software to glue their disparate computer systems into
effective networks.
The goal, users say, is to be able to continue to work in the
multivendor, multiprotocol environments of today, yet access any piece
of information or program from any other computer system.
At the heart of this practical model is the belief that
managers must work to interconnect proprietary equipment and software
using purpose-designed internetworking hardware and software, instead
of waiting for computer vendors to come up with true OSI products.
But things weren't meant to be this way.
Four years ago OSI was hailed as the future of networking. In
June 1988, close to 10,000 people attended the Enterprise Networking
Event, in Baltimore, Maryland, organized by the Corporation for Open
Systems International, a consortium working for the promotion of open
systems. The attraction: the world's first complete OSI network,
using OSI products from some 50 vendors.
Recalling the event, industry veterans say that it was a time
when major computer vendors were battling to establish a predominant
place for their proprietary networking protocols. OSI, they say, was
an attempt to agree on a neutral protocol.
But in the end, vendors agreed to coexist. And that has led
vendors to forge their own definition of open systems based on
multiprotocol routing--a definition that has little in common with the
OSI standards painstakingly pieced together by the International
Standards Organization.
The uncomfortable truth is that, four years after the
Enterprise Networking Event that announced OSI's arrival, OSI has made
little impact where it really matters--in the real user networks.
OSI standards define in some detail the protocols and
interfaces needed to support an open system. In theory, the
seven-layer OSI model allows interoperation between systems that
adhere to the model or, at least, incorporate the functionality of
each layer at each end of the communications link.
Although most users agree that adherence to the OSI model
could help them find a longer-term solution to their interconnection
problems, the majority recognize the reality that most networks are
going to employ many different protocols for years to come.
"Only a few years ago, everyone was talking about OSI," says
Joe Hielscher, strategic business development manager for multimedia
networking requirements at Ungermann-Bass Inc., in Santa Clara,
California. "But today, multiprotocol routing has taken over, and
from what the market is telling us, it is clearly here to stay."
Multiprotocol routing allows users to support different
proprietary protocols such as IBM's Systems Network Architecture and
Digital Equipment Corp.'s DECNet over the same network. Multiprotocol
routers are intelligent devices that direct the flow of data over the
corporate network. They also provide a single point of control and
management for the numerous protocols that large companies typically
use.
The chief way users hold these networks together is through
Transmission Control Protocol/Internet Protocol, which they use to
support their multiprotocol traffic (see table). Designed originally
for heterogeneous traffic traveling on the research community's
Internet network, TCP/IP has spread quickly into commercial networks.
"Back in 1988, OSI was touted as the Cadillac of open
protocols, whereas TCP/IP was the mid-sized economy car," says Bill
Biagi, director of information technology at the Corporation for Open
Systems International. "But OSI Products have been slow in coming,
whereas TCP/IP has evolved and added new capabilities."
FOLLOWING PROTOCOL
(Percentage of Traffic on Protocol, Based on a Survey
of 400 Multinational Companies)
Current In Two Years
------- ------------
OSI 2 4
Bisynchronous 4 3
SNA LU6.2 5 7
DECNet 10 8
Asynchronous 8 6
IPX 15 16
TCP/IP 15 23
SNA SDLC/HDCC 33 28
Source: Business Research Group, of Newton, Massachusetts
EXPLOSION IN DEMAND
Real-world open systems are as different as the companies that
build and use them. Each one uses a blend of different ingredients,
such as local and remote bridges, multiprotocol routers and servers,
protocol converters, application gateways and intelligent hubs, to
hold together the hodgepodge of different computers and networks the
company uses.
"What you see today in large corporations are huge, sprawling
networks with all kinds of devices hanging off of them," says Thomas
Wood, senior industry analyst at consultancy Business Research Group,
in Newton, Massachusetts.
The result has been an explosion in demand for these different
multiprotocol internetworking products. The worldwide market was
projected to reach $416 million in 1991, up 67 percent from 1990. It
should reach $900 million by 1994, according to the Gartner Group
Inc., in Stamford, Connecticut.
The growth of OSI-based products has crawled by comparison.
Real-world open systems reflect the often conflicting
priorities of real-world businesses. For example, the recent
recession led many companies to implement belt-tightening measures
that forced network managers to think first and foremost about
protecting their company's existing investments in proprietary
systems.
Yet they also had to make sure that the company network could
grow and change in line with new business requirements. This meant,
among other things, progressively migrating applications from the
mainframe out to distributed systems, on new and existing sites. The
problem: Different information systems were not initially designed to
interoperate, and a way--any way--had to be found to do so.
Real-world open systems--the cobbling together of many
protocols--was the answer.
GROWING DISENCHANTMENT
The growth of multiprotocol routing also reflects growing user
disenchantment with "open" products from computer and communications
product vendors. Users claim that very few of them actually live up
to the "open" tag of providing for all levels of connectivity for all
types of systems that a typical business uses.
Many vendors acknowledge this.
"Today, 'open' is a meaningless marketing buzzword used by
absolutely every vendor out there...we [vendors] have mangled that
word beyond recognition," says Michael Zisman, president and chief
executive of Soft*Switch Inc., a Wayne, Pennsylvania-based supplier of
electronic mail products.
In addition, users say that open systems solutions available
today are more expensive than what some refer to as the "proprietary
glue-ware" that allows different systems to interoperate.
"For example, one of our sites in Europe using DEC's VAX
systems wanted to use an electronic mail system residing on a IBM
mainframe at the company's U.S. headquarters," says Gary Pettifer,
technical services manager at Air Products and Chemicals Inc.'s
European subsidiary, in Walton on Thames, England. "The open systems
solution would have meant installing a DEC gateway to allow the VAX
users to communicate with the SNA network on the other side of the
ocean, but that would have cost us about 16,000 pounds [$29,600]. So
we ended up equipping each one of the PCs at the site with a separate
communications card to talk to the mainframe, which cost us less than
one-fourth the price."
Users say that they have no assurance that pursuing an OSI
strategy will guarantee success, so they have a hard time justifying
the extra expense of OSI products to their top management.
Most users admit that their long-term preference to implement
open systems flies in the face of their short-term business
requirements.
However, OSI supporters say that businesses that do not adopt
a long-term commitment to OSI have a lot at risk.
"It's really a question of long-term strategy and the
perceived importance of the corporate network as an integral part of
that strategy," says Chris Sluman, managing consultant a Sema Group
Consulting, in London. "Every business has a five-year business plan.
But very few have a five-year information systems plan to accompany
it."
Others doubt that OSI will ever provide users with a viable
solution.
"Standards have become somewhat of a sham," Business Research
Group's Wood says. "What it boils down to is that TCP/IP is here, and
OSI is not."
Nearly two-thirds of 400 large users recently interviewed by
BRG revealed their preference for compatibility with their existing
systems over standards-based solutions.
"For the long term, we try to adhere to OSI as much as
possible," says Hans Peter Angliker, telecoms manager of Swissair,
Switzerland's national airline. "But to solve today's problems, it is
important that our network technology support multiple protocols."
TOUGH GOING
Even those companies that have made the long-term OSI
commitment find the going tough.
Take the case of Hoechst Group, the Frankfurt, Germany-based
chemical company. In 1985, Hoechst was on of the first multinationals
to adopt an OSI strategy, with the aim of integrating its voice and
data communications systems. In 1986, Hoechst also was among the
first companies in Europe to implement an X.400-based E-mail system.
"Our move was intended to put pressure on vendors by showing
that we were serious about open systems. But the move to OSI has been
disappointingly slow," says Harald Nottebohm, Hoechst's recently
retired head of voice and data communications.
Nottebohm says that Hoechst's top management support for OSI
was key to the company's commitment to OSI. Despite this, however, he
says that a continuous debate over what the company's priorities
should be went on between the planners and the operations staff within
his own department.
"People responsible for the day-to-day operations of the
network would say, 'It is all well and good to pursue open systems,
but what do we do about the problems we have to solve today?'," he
says.
Yet despite OSI's lack of real impact, most observers agree
that the OSI movement had some very positive side effects.
"Vendors needed something to motivate them to become more
open, and the emphasis on open systems made them far more open than
they would have become on their own," says Ungermann-Bass' Hielscher.
"And perhaps in the future, OSI may still become a dominant language.
But...things certainly didn't happen the way OSI zealots thought or
hoped they would."
------- End of Forwarded Message
