John, (01)
This was a flawed architecture and flawed decision-making. You don't just go
from A to B overnight. You maintain these as distinct until the new B has the
kinks worked out, or you choose an A + B approach, not a B - A approach. It is
of course possible and prudent to retain your RDB structure and layer semantic
technology over it. (02)
Really, for most of the above paragraph, any technology can be the value of A
and B. Any new B will have this issue. (03)
I've seen many folks embrace "cloud computing", "service-oriented
architecture", "object-oriented programming", "enterprise architecture", etc.,
thinking that adopting these technologies in toto and at once solved all their
problems. Semantic technologies cannot save you from bad computer science and
engineering, nor from bad management. (04)
Thanks,
Leo (05)
-----Original Message-----
From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx
[mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of John F. Sowa
Sent: Saturday, February 25, 2012 9:50 AM
To: ontolog-forum@xxxxxxxxxxxxxxxx
Subject: Re: [ontolog-forum] What goes into a Lexicon? (06)
Dear Matthew, (07)
> So the real question is: Why have the proponents of semantic technologies
> not been able to demonstrate the business benefits that the adoption of
> those technologies would achieve? The same applies for ontologies. (08)
I'd like to cite an example of what happened when some enthusiastic
proponents convinced management that an RDF database would have
business benefits. (09)
I heard this story from a friend, whom I'll call DS (Doubting Skeptic).
DS was called in as a consultant to an IT department that implemented
some great new functionality using RDF and web-based tools. This was
running on the local Intranet, which was separated from the WWW by
a firewall. (010)
People, including management, liked these new functions. In addition
to the RDF included in web pages, they accessed data from the employee
database, which was running Oracle. However, they transferred the
employee data from the RDB to RDF overnight. During the daytime,
the web services ran with the previous day's employee data in
a read-only way. (011)
However, management wanted two extensions: (1) the ability to run
with the latest employee data, and (2) the ability to update the
employee database from the web services. So the RDF proponents
convinced management that RDF + SPARQL could do everything that
Oracle could do -- and in a more integrated way. (012)
The IT department spent some time and quite a bit of money to
modify all the software that accessed the RDB to switch over
to RDF + SPARQL. They demonstrated that every function that
had been performed by Oracle could be performed by the modified
software when operating on the same data. (013)
But before switching over, management wanted some outside
reviewers to verify the suitability of the new software.
Since DS knew Oracle and RDF, he was hired as one of the
reviewers. DS noted the following issues: (014)
1. RDBs use locking to ensure that each transaction that
updates any data would run to completion before any other
program could modify any field that the transaction would
use. But the RDF + SPARQL system did not have locking or
any other mechanism to ensure data integrity. (015)
2. RDBs support checkpoints and transaction logging so that
if a crash occurs it's possible to restart at the last
checkpoint and rerun all the transactions since that time.
But the RDF software did not do logging and checkpoints. (016)
3. The timing results for the SPARQL software were somewhat
slower than the RDB software, and the tests had only been
done by running one program at a time. Nobody had done any
realistic tests of what would happen under normal operating
conditions. (017)
These were the three main concerns, but DS went through a long
litany of other possible problems that might occur. DS suggested
that the IT group continue to run the RDB and the web software
concurrently. But they could add new software to allow the
web services and the RDB software to send incremental updates
to one another. (018)
The IT manager rejected that proposal because the transition
to an integrated RDF + SPARQL system had already been decided.
He informed DS that his services were no longer required. (019)
A while after the scheduled transition, DS called one of the people
working on the project (not the manager, but one of the so-called
"worker bees" who knew what was going on). DS asked how the
transition was going. Short answer: chaos. (020)
All three issues cited above plus others brought the system
to its knees. The performance of both the new web services and
the old RDB services were excruciatingly slow. The absence of
locking on updates had created "anomalies", but nobody knew
which updates caused the errors or what the correct data
should be. (021)
Furthermore, the end of the month was rapidly approaching, and
the employees were going to be very unhappy when they failed to
get their paychecks. Upper management wanted to return to the
old system to generate paychecks, but that was not possible --
they could map data from RDB to RDF, but they had no way to map
data from RDF to RDB. The RDB still contained the old data,
but there was no way to update it with any of the new data. (022)
The IT manager did not keep his job much longer. (023)
John (024)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (025)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (026)
|