For all you youngsters out there, maybe you wouldn't know a logical unit
if it hit you over the head. But for many enterprises, LU2
is a key protocol they use to integrate with mainframe or
LU2 supports communications between terminals and
screen-oriented mainframe applications, often running under IBM's CICS. (For
more on SNA and LU types, see IBM's
redbook from 1995) Many (most?) older CICS apps used an
architecture that did not foresee the need to interconnect
applications broadly - they exposed only a single interface,
the terminal-based user interface. In some cases the
UI logic, the 3270 screen formatting, is so tightly wound up in
business logic, or even workflow or screen flow (something like
the "page flow" concept in web apps today), that it is,
practically speaking, impossible to uncouple the UI from the
These days we have MVC patterns and multi-tiered architectures,
but older designs didn't have the benefit of hindsight that we
do. They also didn't have the same engineering constraints -
Twenty years ago, there was no "world wide web". The economics
of computer processing power were governed by Grosch's
Law, named after Herb Grosch, who held that compute power
increases in proportion to the square root of compute cost.
This meant, bigger computers were more economical. At the same
time, the cost of compute power was the factor to optimize for -
as opposed to the cost or speed of application development, for
example. People were cheap compared to the equipment. These
engineering and economic forces encouraged the designers of
older systems to invest in monolithic, powerful centralized
nodes of computing - a mainframe surrounded by dumb terminals.
Call it the Big System design model. LU2 is how
terminals talked to those Big Systems.
Grosch's Law eventually succumbed to Moore's
Law, which predicted that the density of transistors on a
chip would double every 18 months. The relevant corollary is
that compute power would get cheaper and cheaper, as density
increased. This was not quite the antithesis of Grosch's Law,
but it inverted the value relationship between people and
equipment, so much so that Grosch's law did not "become
incorrect" as some have said, but rather it became irrelevant.
A key implication of the compounding effect of Moore's Law has
been that, in any information system, the cost to optimize for
is no longer the equipment, but the people. A mid-sized
computer no longer costs half a million 1976 dollars. Instead
you can go to Dell.com and get it shipped next day, for under
$1000. But IT people haven't gotten much cheaper!
This inversion changed everything, including the idea that
monolithic systems were optimal. Thanks to the persistent
phenomenon that Gordon Moore described, we now design under a
distributed, composable model (eg, our current model of SOA and
"web services), and we (mostly) optimize not for the cost of the
equipment, but for the cost of the people that run, program, and
use the equipment.
So that explains why Big Systems were so popular, why they were
hard to connect to, and why we have sort of been gradually
migrating away from the Big System model, in macro terms. But
Big Systems are still here, still in place, working. The changes in compute economics haven't eliminated the need
to interoperate with legacy systems. And the rules are set - if you
want to talk, ya gotta go through a terminal protocol. So
today, LU2 is often used for application-to-application
communication, where the calling application impersonates a
terminal. Screen scraping.
In the next release of Microsoft's Host Integration
(HIS 2006), Microsoft plans to include a new feature called
"Session Integrator". Session Integrator is similar to HIS's
current Transaction Integrator feature, but it will work
with LU0 and LU2 connections (whereas TI works with LU6.2 and
TCP/IP). This was first disclosed by the HIS team at the IT
Forum 2005 event in Barcelona, in November 2005.
Session Integrator will be included in HIS 2006 Enterprise
Edition, which will ship as part of BizTalk 2006. If you buy
BizTalk 2006 (any edition) you get a copy of HIS 2006
included. HIS 2006 will go into public beta Real Soon
Now. The public general release of this product is planned
for later this year.
Even though web services are all the rage, and even though IBM
is implicitly discouraging SNA deployments by publishing
end-of-service dates for SNA-based network gear and software,
and by improving the TCP/IP connectivity of Mainframe systems...
despite all that...there are still many companies out there who
need SNA-based communications solutions in the near term.
That's what this new feature in HIS 2006 is designed to support.
Yes, Microsoft will still invest most strongly in webservices
efforts, but we also need to satisfy the practical requirements
of customers. Nobody can afford to start all over again. Every
company or organization has some existing "investments" or
systems in place, and not all of them support web services. We
need to do what we can to make interop with those existing
systems, possible and easy.
Host Integration Server
is a big part of Microsoft's answer to that