[comp.databases] Summasy of responses to database question

remco@tnoibbc.UUCP (Remco Bruyne) (02/12/90)

Hello, 

A short time ago I posted a request about information concerning database
systems capable of networking, storing pictures and with support to create
and maintain applications with a reasonable amount of work. I got some
responses some of which were very useful.

The question was:
>We are looking for a database system which suits the following needs:
>
>1) Should be able to run on several types of hosts.
>2) Should not get too slow with databases of about 100,000 and more records.
>3) Reasonable costs, may be dependent on the host system used.
>4) Should be reasonably easy to generate/maintain and change applications.
>5) Should be able to interface to other DBMSs (maybe with some interface
>   software).
>6) Capabilities for storing images/line pictures should be present now or
>   in future implementations of the product. Also think of interfacing to
>   existing drawing/digitizing programs.
>7) Capabilities for network communication, because it is necessary that
>   information can be retrieved from remote systems.
>    
>That's all :-).
>
>Probably this system doesn't exist, but when you know of something that
>comes really close, could you let me know? 

Here follows a summary for the interested (in the order in which
I received the messages):
----------------
Date: Tue, 30 Jan 90 15:02:10 -0800
From: Donovan Hsieh <donovan@csl.sri.com>
Message-Id: <9001302302.AA01620@julius.csl.sri.com>
Subject:  Wanted: DBMS for distributed information containing images

In remco@tnoibbc.UUCP of comp.databases, Remco Bruyne writes
"Wanted: DBMS for distributed information containing images":


I guess most existing commercial RDBMS should meet your requirements
more or less except (6).

This is also pretty much depending on how you want to
manipulate your images/line pictures data. If you simply want
to store them as it and bring them back to whatever tools for
processing. There are two ways you can handle it:

a. Store them into a special long fiield which may handle data up
to 4KB (vendor dependent). Of course, you may want to compress and
decompress them before and after the access to save storage space.

b. Another way is to implement it as a database application instead
of using the database system feature. In your database schema, a special
field which contains either a file pointer to those images files,
or a reference path name which may lead your applications to access
the file and pass to the tools.

If however you want a fine grained control over the internal data
structure of those images and line picture files. I would suggest you
to use an OODB which is available on several vendors. This gives you
the richest data modeling capability in the long run. Currently, those
OODBs are almost implemented in a client/server architecture on
the network. Some even provide backward SQL compatibility to certain
extent.

Donovan Hsieh
Computer Science Lab
Stanford Research Institute
Menlo Park, CA
------------------------------------------------------------
From: prc@erbe.se (Robert Claeson)
Message-Id: <9001310919.AA16119@maxim.erbe.se>
To: remco@tnoibbc

Check out version 6.3 of Ingres (the one with knowledge management
and object capabilities). It should fit your bill well.

Robert
------------------------------------------------------------
Date: Wed, 31 Jan 90 16:45:52 +0100
From: johan@cwi.nl
Message-Id: <9001311545.AA01859@kangaroo.cwi.nl>

[This response was in Dutch] I give a short summary in my own
words and hope I correctly reproduce the main information...]
To my opinion Ingres satisfies all requirements except 6.
A future version (not the next) will probably have support for this.


>1) Should be able to run on several types of hosts.
I know of:
Apollo, Dec (VMS en Unix), Gould, HP, NCR, Pyramid, Sequent, Sun
286 en 386's (Unix, OS/2, MS(DOS), the latter with restrictions).

>2) Should not get too slow with databases of about 100,000 and more records.
Ok. [I left this untranslated :-)]

>3) Reasonable costs, may be dependent on the host system used.
Depends on the number of users.

>4) Should be reasonably easy to generate/maintain and change applications.
SQL + Quel, which is comparable to SQL, but a little more powerful.

Ther are 4 levels of interactive use:

name     enduser       development-          support for 
       ease-of-use     time (order)    development    maintenance
                                           of applications

QBF         +           uur               -             +
SQL         -           uur               -             -
4GL        ++           week              +             +
3GL       +++           maand            ++             +

SQL has a 'tty-interface', the others a form-interface.
A screen represents a form containing fixed texts, fields with attributes.
The form-definition can be made easily for QBF, 4GL and 3GL.
[stuff deleted]
4Gl can be converted to 3GL (C, Fortran, Pascal , Cobol, Ada) with
embedded SQL-calls and form-control statements.

>5) Should be able to interface to other DBMSs (maybe with some interface
>   software).
There is a gateway system for information exchange to and from most
well-known systems (DB2, SQL/DB, ... )

>6) Capabilities for storing images/line pictures should be present now or
not yet

>7) Capabilities for network communication, because it is necessary that
>   information can be retrieved from remote systems.
TCP/IP, based + other options.

Other characteristics:
   good multi-user properties (locking; data-integrity
      by handling multi-statement transactions atomically.
   slow in single-record operations (large overhead)
      compared to other systems fast mutli-record operations.
   requires a lot of resources (+/- 50 MB binaries end support files);
      data in Ingres-tables with Random_access and some indices easily
      needs 10 times more space than a corresponding sequential Unix-file;
      A server with some clients need about 10 MB of RAM;
      abundant cpu and io necessary

                                                    Johan Wolleswinkel
CWI, Postbus 4079, 1009 AB  Amsterdam,  Kruislaan 413, The Netherlands
Phone: +31 20 5924122    Fax: +31 20 5924199   Telex: 12571 (mactr nl)
Internet:   johan@cwi.nl            or          ...!hp4nl!cwi.nl!johan

Johan, thank you for your extensive explanation which was very
useful.
------------------------------------------------------------
From: dlewis@cper1.Dayton.NCR.COM
Message-Id: <9001312154.AA01138@uunet.uu.net>

Progress Software Corporation has a 4GL called Progress.  You can access
foreign databases on Unix and NCR Mainframes running our proprietary OS 
called
VRX.  You can access Total databases (from cincom) and Oracle databases.

In May, they will be releasing the Progress to Progress database access so a
DB can access multiple Progress databases.

This would be one possibility.

>Dave
------------------------------------------------------------
From: uwm!dgh%unify@relay.EU.net (David Harrington)
Message-Id: <9001311723.AA23624@unify.UUCP>
Organization: Unify Corporation, Sacramento, CA, USA

Unify Corp., 3870 Rosin Court, Sacramento, CA 95834, is a developer of
application development technology and tools.  We have a toolset that consists
of a visual, WYSIWYG, GUI application screen generator combined with a powerful,
event-driven 4GL, which is called Accell/SQL.  Accell/SQL runs on Unify's OLTP
RDBMS, Unify 2000, as well as Oracle V5.x & V6.x; Informix/SQL, Sybase and SCO
Integra.

As regards your requirements:

1. Unify 2000 runs on > 20 hardware platforms.
2. Unify 2000 is rated at 150 TPS on the TP1 benchmark.
3. Unify 2000 is priced based on # users and size of the h/w platform.
4. Accell/SQL gives you the application development environment that should
   easily meet your development needs, now and in the future.  
5. Accell/SQL interfaces to (and runs on, as a native implementation) all the
   databases listed above.
6. The Open Look release of Accell/SQL will provide full graphical
   capabilities, including the ability to manipulate bit-mapped images and
   interfaces to drawing packages.  Unify 2000 supports this capability with its
   BLOB data type.
7. In addition, remote database access is available to Accell/SQL applications
   via Unify's Accell/NET product, via SQL*NET for Oracle and Sybase's
   client/server architecture.

Looks like this is what you need.

-- 
David Harrington		                       internet: dgh@unify.UUCP
Unify Corporation		                 ...!{csusac,pyramid}!unify!dgh
3870 Rosin Court                                          voice: (916) 920-9092
Sacramento, CA 95834                                        fax: (916) 921-5340
------------------------------------------------------------
Date: Wed, 31 Jan 90 13:00:15 PST
From: sybase!cyoung@relay.EU.net (chris young)

Sybase runs on many different flavors of UNIX, VAX/VMS, STRATUS, and others.
And there are probably a few platforms which are native to Europe of which
I am not aware.

>2) Should not get too slow with databases of about 100,000 and more records.
Sybase consistently provides high performance on all of these platforms.
Benchmarks for Sun machines are available.  Databases can run into the
Gigabyte range.


>3) Reasonable costs, may be dependent on the host system used.
Sybase is not inexpensive, however it provides one of the lowest price/per-
formance ratios available.  Our pricing schedule should be available
from the Sybase office in the Netherlands.  (see below)


>4) Should be reasonably easy to generate/maintain and change applications.
Sybase enforces referential integrity and other business rules at the
database level.  Therefore a change in the business rules requires
changing the rules in one place, in the database.  Ensuring that each
application gets the change and is properly coded does not constrain the
integrity of the system.  Among many other benefits, this architecture
provides for lower maintenance costs.

With the Application Productivity Tools (APT), application development
and prototyping efforts are reduced.  Forms are easily generated in 4GL
code which can be modified/extended to suit the developer's needs.


>5) Should be able to interface to other DBMSs (maybe with some interface
>   software).
Sybase's Client/Server Interfaces provides for customized front-end (Open
Client) and back-end (Open Server) applications.  This means that you can
develop front-ends that are able to talk to differing back-ends and back-ends
that can talk to differing front-ends and back-ends that can talk to 
differing back-ends.  This is accomplished through a self-describing
data protocol called the Tabular Data Stream (TDS).  The protocol is not
a network protocol (ie it does not replace TCP/IP, for example) rather
it is a data messaging protocol that is transmitted between the front-end
and the back-end via the network.  What all this means is that it is
possible to develop an Open Server application that acts as a go-between
for the Sybase SQL Server and another database engine.  In fact, Open
Server applications have been built that communicate between Sybase and
DB/2, IMS, and RMS.

Open Server is a product containing the necessary routines to build a
custom server.  This means that the developers themselves can build the
server and do not need to wait for the vendor to produce a gateway 
product to the desired database.  Most vendor gateways try to provide
functionality that may not be needed for a particular application.  And
most gateways don't allow for custom code that provide specialized
functions which an application may need.  This provides for greater
flexibility when developing interfaces between databases.


>6) Capabilities for storing images/line pictures should be present now or
>   in future implementations of the product. Also think of interfacing to
>   existing drawing/digitizing programs.
Sybase has a binary datatype that allows up to 2GB of storage per row.
I am aware of applications in the US which interface with Sybase, however
I do not know the company names.  If this is an issue, let me know.


>7) Capabilities for network communication, because it is necessary that
>   information can be retrieved from remote systems.
Precisely because Sybase has invested in the Client/Server architecture,
communication between database servers is matter-of-fact.  Not all
components are available yet (such as transparent data access between servers)
but many of the essential parts are.  Two-phase commit is one example.


I am assuming by your phone number that you are loccated in the Netherlands.
We do have an office in the Netherlands.  (Sorry, but I don't know the city).
The phone number is 31 569205.  (I hope that is correct.  If not, let me know
and I'll track down the correct number and a person's name to contact.)

Thanks,

Chris Young
Consultant
Sybase - Denver, US
phone: (303) 721-3308
cyoung@sybase.com
------------------------------------------------------------
Date: Fri, 2 Feb 90 16:18:29 PST
From: sybase!tiger@relay.EU.net (tiger whittemore)

I believe the Sybase RDBMS should handle all of your criteria very
well.  I may be biased as an employee, but I believe Sybase has a
very good reputation, and features to support the functionality
you've mentioned.

You can contact Sybase at 415 596-3500.
------------------------------------------------------------

-- 
---------------------------------------------------------------------------
 Remco Bruijne   USENET: remco@tnoibbc.ibbc.tno.nl    PHONE: +31 15 606437
---------------------------------------------------------------------------