[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference ulysse::rdb_vms_competition

Title:	DEC Rdb against the World

Moderator:	HERON::GODFRIND

Created:	Fri Jun 12 1987
Last Modified:	Thu Feb 23 1995
Last Successful Update:	Fri Jun 06 1997
Number of topics:	1348
Total number of notes:	5438

1206.0. "A competitor to watch: Interbase" by BAGLEY::FOSTER (Old RALLYator Sighted in Midwest) Wed Nov 18 1992 21:58

I attended an Interbase seminar yesterday put on by Borland.  The quality and 
glitz of the presentation were excellent, and the room was filled to over-
flowing.  Oldtimers will remember that Interbase began as Rdb/ELN, Jim Starkey's
answer to Rdb/VMS.  Jim's main idea was to avoid locking by using a time-stamp
mechanism to serialize transactions.  This is still the main selling point of
Interbase.  However, Interbase is now the mainstream database product of the
world's largest database vendor (in units sold).  Borland is moving to make
Interbase the UNIX-based server underneath both Paradox and dBASE, and this 
looks like a formidable competitor.  (Paradox and dBASE together own 75% of the
PC database market.)  Among UNIX databases, Interbase has moved from 0% to 4%
market share in one year.

They also have an impressive number of reference accounts.  Unlike us, they
list their successes and discuss them in detail at the beginning of their
presentation, before the techie stuff.

As for the techie stuff, they claim that they are the first of the "third wave"
of database technology.  Prominent features:

- No-lock concurrency, which gives better performance in mixed read/update 
applications (they call it "On-Line Complex Processing").  They claim to beat
the Ingres's and Sybase's by a large margin

- Database recovery in less that one second (they don't use Undo/Redo logs).

- Flexible blobs as part of the architecture (not tacked on), with blob filters.

- Transactional event alerters.

- User-defined functions in the database.

As for competition, they mentioned Informix, Oracle, Sybase, and Ingres 
repeatedly, but they only mentioned Rdb to say that their support for DSRI 
made it easy to simply unplug Rdb and plug in Interbase.  "Your applications
will continue to run unchanged, and in fact will run much faster."  

If you want more info, send mail.

T.R Title User Personal
Name Date Lines

1206.1 32370::BERENSON Database Architecture, Standards, and Strategy Wed Nov 18 1992 23:48 56

T.R	Title	User	Personal Name	Date	Lines
1206.1		32370::BERENSON	Database Architecture, Standards, and Strategy	`Wed Nov 18 1992 23:48`	56
	>- No-lock concurrency, which gives better performance in mixed read/update >applications (they call it "On-Line Complex Processing"). They claim to beat >the Ingres's and Sybase's by a large margin READ: They have terrible data consistency problems. If you insist on having data consistency, they'll give it to you by single threading all transactions through the database. This is called turning a vice into a virtue via marketing. >- Database recovery in less that one second (they don't use Undo/Redo logs). READ: They use multiversioning inside the database to do recovery. They do not get good record clustering because the database is littered with old versions. I doubt they do inter-table clustering at all. Nor hashing. Nor index-only retrievals. I'm not even sure they have a viable media recovery strategy. >- Flexible blobs as part of the architecture (not tacked on), with blob filters. This is either incredible BS or they've architected blobs differently than DSRI. It is quite clear to anyone who has every looked at blobs that they were a tacked on wart to the DSRI architecture. Starkey was the one who architected them in DSRI originally, so I tend to believe OSRI isn't any better. >- Transactional event alerters. Now this is neat. >- User-defined functions in the database. And so is this. >As for competition, they mentioned Informix, Oracle, Sybase, and Ingres >repeatedly, but they only mentioned Rdb to say that their support for DSRI >made it easy to simply unplug Rdb and plug in Interbase. "Your applications >will continue to run unchanged, and in fact will run much faster." There are two incredible pieces of BS here. First, their DSRI compatibility is circa Rdb V2.0. Second, the cases where they will run faster are somewhere near 0 for any non-trivial application. It takes a pathological case for them to really show any benefit over Rdb or any of the rest of the competition. The way they sell the product is to target it for the pathological cases! There is a massive battle shaping up in the PC space between Borland and Microsoft. Borland's strategy is (in my characterization) to win by forcing the archaic positional semantics of DBASE and PARADOX into relational database engines with their own Interbase is the forfront. Microsoft's strategy is to move towards relational databases using ODBC. Borland has existing marketshare going for them. Microsoft has the hotest of the DBASE clones (FOXPRO) and the brand new Microsoft Access going for them. Hal

>- No-lock concurrency, which gives better performance in mixed read/update 
>applications (they call it "On-Line Complex Processing").  They claim to beat
>the Ingres's and Sybase's by a large margin

READ: They have terrible data consistency problems.  If you insist on
having data consistency, they'll give it to you by single threading all
transactions through the database.  This is called turning a vice into a
virtue via marketing.

>- Database recovery in less that one second (they don't use Undo/Redo logs).

READ: They use multiversioning inside the database to do recovery.  They
do not get good record clustering because the database is littered with
old versions.  I doubt they do inter-table clustering at all.  Nor
hashing.  Nor index-only retrievals.  I'm not even sure they have a
viable media recovery strategy.

>- Flexible blobs as part of the architecture (not tacked on), with blob filters.

This is either incredible BS or they've architected blobs differently
than DSRI.  It is quite clear to anyone who has every looked at blobs
that they were a tacked on wart to the DSRI architecture.  Starkey was
the one who architected them in DSRI originally, so I tend to believe
OSRI isn't any better.

>- Transactional event alerters.

Now this is neat.

>- User-defined functions in the database.

And so is this.

>As for competition, they mentioned Informix, Oracle, Sybase, and Ingres 
>repeatedly, but they only mentioned Rdb to say that their support for DSRI 
>made it easy to simply unplug Rdb and plug in Interbase.  "Your applications
>will continue to run unchanged, and in fact will run much faster."  

There are two incredible pieces of BS here.  First, their DSRI
compatibility is circa Rdb V2.0.  Second, the cases where they will run
faster are somewhere near 0 for any non-trivial application.  It takes a
pathological case for them to really show any benefit over Rdb or any of
the rest of the competition.  The way they sell the product is to target
it for the pathological cases!

There is a massive battle shaping up in the PC space between Borland and
Microsoft.  Borland's strategy is (in my characterization) to win by
forcing the archaic positional semantics of DBASE and PARADOX into
relational database engines with their own Interbase is the forfront.
Microsoft's strategy is to move towards relational databases using ODBC.
Borland has existing marketshare going for them.  Microsoft has the
hotest of the DBASE clones (FOXPRO) and the brand new Microsoft Access
going for them.

Hal