SECTION 1 -- INGRES
<> 01.001 What is INGRES? %Roy Hann
INGRES is a relational database management system.
There are two distinct strains of INGRES; there is so-called
"University" Ingres, which is in the public domain, and there is a
commercial version of INGRES developed and marketed by Computer
Associates. Most of this FAQ is devoted to the commercial version
which is called OpenIngres.
University Ingres is also known as Berkeley Ingres, Ingres89, and
"distributed" Ingres. (The latter term is unfortunate because it is
hard to know whether someone is talking about public domain
"University" Ingres, or the distributed database management
capabilities of OpenIngres. The term "distributed" should probably be
avoided.) University Ingres is the original version of Ingres
developed at UC Berekely during the 1970s to demonstrate the concept of
a relational database. It has a plausible claim to being the first
implementation of a relational database management system since it
pre-dates IBM's System R. This program eventually evolved into the
first versions of the commercial product marketed by Relational
Technology Inc. (eventually bought by The ASK Group and then by
Computer Associates). It has some of the features of the currently
available commercial version, but is architecturally different; the DML
is QUEL, and the performance and reliability are only fair to
adequate. NB: NONE of this FAQ is related to University Ingres unless
EXPLICITLY noted.
OpenIngres is a commercial product and costs money to buy. It is a
robust, state-of-the-art RDBMS with a reputation for including
first-class development tools. The database server component runs on
all the major Unix derivatives, as well as VMS, Microsoft Windows NT,
and even on Novell Netware as an NLM. The Ingres client components run
on all of these too, and also on AXP VMS, Microsoft DOS and Microsoft
Windows.
As well as a database engine, the OpenIngres product suite includes a
range of development and enquiry tools and components for implementing
distributed databases. The database engine and tools are usually
referred to collectively as "Ingres" without distinction.
--------------------
Trivium: INGRES is an acronym for INteractive Graphics REtrieval System
(revealing the nature of the project out of which the experiments with
relational databases arose). By happy accident, there was also a
French artist by the same name: Jean Auguste Dominique Ingres
(1780-1867). (A highly placed source who wishes to remain anonymous
confirms that the selection of the name WAS an accident.) Warning to
those with no social life: The book entitled "Ingres's Eroticized
Bodies" by Carol Ockman, ISBN: 0300059612, has nothing to do with
software.
<> 01.002 Where can I get PD Ingres?
University Ingres 8.9 is in the public domain. It is available for
anonymous ftp from many sites around the world. It is also distributed
as unsupported or contributed software by some vendors, such as
Digital Equipment Corporation.
ftp Sites:
==========
s2k-ftp.cs.berkeley.edu pub/ingres
This site also has the `linux' port of Ingres 8.9. Many additional
sites can be located using "archie" and other mechanisms.
Contact: ingres@postgres.berkeley.edu
There are no known ports of University Ingres to Xenix, MS-DOS,
Windows NT, or any other non-Unix operating system. [If you are aware
of counter-examples please let me know at rhann@tnc.com] There have been
various bold announcements of efforts to do some of these other ports,
but so far none has come to fruit.
<> 01.003 Where can I get OpenIngres?
OpenIngres is a commercial product. For information, pricing and
authoritative availability information contact your local Computer
Associates office.
Australia: (61)(2)923-2066
Austria: (43)(1)894-1913
Belgium: (32)(2)773 28 11
Canada: (905)676-6700
Denmark: (45)(42)95 86 00
France: (33)(1)40-97-50-50
Germany: (49)(6151)949-0
Holland: (31)(3402)483 45
Ireland: (353)(1)478 0800
New Zealand: (64)(4)801-7654
UK: (44)(753)5777 33
US: 1-800-225-5224
Or call International Operations at 1-516-342-5224 for the number of
your local office.
<> 01.004 How much does INGRES cost?
University Ingres is free. OpenIngres is not.
The cost varies a lot depending on the class of machine you intend
to use, the number of users, and what components of OpenIngres you want.
It can be a little or a lot. Contact your Computer Associates office.
<> 01.005 Does OpenIngres support SQL?
OpenIngres supports SQL. OpenIngres 1.x is compliant with the ANSI/ISO
Entry-Level SQL92 (SQL2) standard and also has some of the
Intermediate-Level features. Embedded SQL and embedded Dynamic SQL are
also fully supported. (See section 05.003 below for SQL2 references.)
Note that OpenIngres also supports QUEL and embedded QUEL, which it
inherits from University Ingres. Computer Associates de-emphasizes
QUEL for obvious commercial reasons. Because QUEL has considerable
technical advantages over SQL, OpenIngres adopted SQL relatively late
(c. 1986) and perhaps for this reason there is a lingering
misapprehension that OpenIngres still does not support SQL.
University Ingres supports only QUEL and embedded QUEL. (See 03.006
for a description of `onyx' which provides an SQL to QUEL interface for
University Ingres.)
<> 01.006 What are the features of OpenIngres?
DATATYPES
=========
OpenIngres supports the following native datatypes:
i1 1-byte signed integers
i2 2-byte signed integers
i4 4-byte signed integers
f4 4-byte floating point
f8 8-byte floating point
cN N-byte fixed length PRINTING character strings (1<=N<=2000)
char(N) N-byte fixed length ASCII character strings (1<=N<=2000)
varchar(N) N-byte variable length ASCII strings (1<=N<=2000)
text(N) N-byte variable length ASCII, except NUL (1<=N<=2000)
date 12-byte date and/or time (absolute or interval)
money 8-byte money type (locally defined currency)
byte 1-byte unsigned quantity
decimal N-byte fixed decimal
long varchar N-byte variable length ASCII strings (1<=N<=2Gb)
long byte N-byte variable length binary data (1<=N<=2Gb)
Columns of the above types may be made nullable, or may be assigned a
default value. OpenOpenIngres 1.x allows user definable default values.
table_key 8-byte surrogate key (optionally system maintained)
object_key 16-byte surrogate key (optionally system maintained)
In addition, an optional OpenIngres component, the Object Management
Extension, allows the programmer to define abstract datatypes and
operators on those datatypes and the native datatypes. The new types
and operators can be used in any context without restriction.
TABLE ORGANIZATION
==================
OpenIngres allows base tables (and secondary indices) to be organized as:
ISAM
compressed ISAM
hash tables
compressed hash tables
heaps
compressed heaps
sorted heaps
compressed sorted heaps
B-trees
compressed B-trees
Tables may have up to 300 columns, and a row may be up to 2000 bytes.
Tables may have up to 8 million pages of 2kb each (16Gb). The number
of rows that will fit in 8 million pages depends on the row size.
There is no practical limit on the number of tables in a database.
There is no practical limit on the number of databases on a system.
Tables may be reorganized, relocated and re-indexed on-line.
A table may be distributed over multiple OpenIngres locations.
OpenIngres locations can be defined on an unlimited number of physical
drives.
OpenIngres uses the native file system of the platform on which it is
running. A "database" is (one or more) directories, and tables
are files within the directory.
OPERATORS
=========
OpenIngres supports all the SQL92 operators, and QUEL.
DISTRIBUTED DATABASE CAPABILITIES
=========== ======== ============
OpenIngres applications can connect over a network to remote databases
using OpenIngres/Net. That is to say that the application does not
necessarily have to be running on the same machine as the database.
An application can use OpenIngres/Net to connect to more than one remote
database simultaneously (BUT NOT INTEROPERATIVELY; ie. it can't join
together two tables, one from each database.) More importantly changes
made to the databases cannot be guaranteed to have referential
integrity. To achieve this, special coding is required, or the
application should use OpenIngres/Star.
Using OpenIngres/STAR databases can be distributed over multiple host
systems. Using OpenIngres/STAR a table in database `A' CAN be joined with
a table in database `B'.
OpenIngres also provides gateway products that allow OpenIngres to access
"foreign" databases. Gateways exist for a number of products such
as Rdb, DB2 and IMS.
OpenIngres can act as an ODBC server.
SECURITY
========
Using SQL OpenIngres provides the usual per-user security constraints.
Users can be selectively permitted to read, update, append and
delete, on a table by table basis.
Using the optional OpenIngres/Knowledge Management Extensions it
is also possible to define and manage groups of users and user roles.
Both of these make security management much simpler.
QUEL users can additionally control access by time of day, day of
the week, terminal used, and the content of the record. These
extra QUEL constraints ARE NOT IMPOSED ON SQL USERS.
For users with special security needs, there is a secure version
of OpenIngres called OpenIngres/Enhanced Security.
See section 01.013 for further information.
INTEGRITY
=========
Integrity constraints on the data are imposed at the level of the
database engine and cannot be subverted, except that integrity
constraints are not enforced during table load operations using the
`copy' command. Integrity constraints can also be imposed in form
definitions, but these are imposed only if the user happens to use
form which includes the constraints (or validations as they are
called in `vifred'.)
Declarative referential integrity constraints are supported.
Checkpoints of the database can be done on-line and can be written
to disc or tape.
<> 01.007 What platform should I run OpenIngres on?
There is no single good answer to this question, although it pops up
regularly. It depends on dozens, maybe even hundreds of factors.
The best you can probably hope for is some advice on the KIND of system
to run. You can then contemplate factors like the vendor's quality of
service, the operating costs, the kind of deal you can cut, and a host
of intangibles--like do you feel comfortable dealing with the company?
No one can help you with these.
If you are going out to buy hardware to run OpenIngres, and you don't
have an installed base or any other factors affecting your decision
other than just finding the "best" OpenIngres platform, then you should
first decide how you want to run OpenIngres.
You have three or four choices: (1) you can run the OpenIngres server
and the OpenIngres applications on a central system in the traditional
mainframe style. (2) You can run the OpenIngres server on a central
database host and have one or more OpenIngres application servers
accessing the database through OpenIngres/NET. (3) You can have the
OpenIngres server run on a database host and have the applications run
on desktop clients such as PCs or Unix workstations. (4) You can do a
mix of all of the above, with X-terminals thrown in to cloud the issue
still further.
The pros and cons of the various approaches are too numerous to go
through in detail here now (a future version of the FAQ may attempt a
discussion of this subject). For now, suffice it to say that a single
central system and a number of "dumb" character-cell terminals will
probably be the cheapest solution and will not raise any networking
issues. A client/server solution with a dedicated central database
host and intelligent desk-top clients will probably get the best
performance out of any given database host but could be expensive if
the clients aren't already in place.
Neither of these generalizations will always be true of course, but
they may serve to help the hardware selection process. If the database
server and the applications are going to run on a central system, you
may need a more powerful machine. If you just run the database server
on the central system and have smart clients, you may be able to get
away with a more modest machine. Maybe. As always, job mix, work load
peaks, access patterns and a myriad other things can influence what you
need.
Having selected your strategy, you can move on to select specific
hardware. If the decision is unconstrained by any requirement for
compatibility with existing hardware or the need to exploit existing
expertise, you could look at benchmark results. The "benchmark"
benchmark is the suite of TPC benchmarks (TPC-A, -B and -C).
Unfortunately OpenIngres is not submitted to the Transaction Processing
Council for evaluation. Furthermore, these are presently regarded with
some suspicion because -A and -B have been accused of being open to
manipulation. Indeed the same claim has been made about the whole TPC
organization. Eventually these issues will be resolved, but for now,
the benchmarks are probably still quite useful for comparing hardware
platforms running a specific DBMS product, even if they are not
reliable for comparing different DBMS products. Even if the benchmarks
were above suspicion, they should still only figure as one factor among
many.
No matter what platform you choose, you must configure it correctly to
get the best performance out ot it. A good hardware vendor will be
willing to help you with this, indeed their willingness to help could
reasonably be a selection criterion. Proper configuration can include
tuning system software parameters, installing optional software
designed to improve database performance, and choosing the right
hardware components in the first place. It is possible to misconfigure
any system. Note that workstations are generally designed and
configured to optimize number crunching and sequential access to
contiguous regions of disc--for signal and image processing
applications. A database server has very different needs.
A relational database application will be disc bound (usually), so it
pays to buy fast discs. In most relational databases you tend to get
lots of little records spread all over the disc so transfer time will
usually be irrelevant--access time (rotational latency and seek-time)
will dominate. Fast discs are currently able to achieve average seek
times of 8-9msec and average rotational latency of 4.3msec.
A fast disc is wasted without a controller that can take advantage of
its speed. Choose a fast controller too. The controller should not
only be fast with one disc, it should exhibit a linear increase in
throughput as discs are added.
No OpenIngres system should have fewer than 3 physical discs. To
achieve the maximum level of security from disaster you will want to
ensure that you can keep your database, your transaction log and your
journal files physically separate. If this is done, then even in the
event of a catastrophic failure of a disc drive you would be able to
recover the database to the state in which it was at the precise
instant of the failure. A RAID might be a very good alternative.
Some routine database operations require space equal to 3 times the
space occupied by the largest table. Plan to have sufficient free disc
space at ALL times to allow you to do a "modify" on your largest
table.
Extra RAM is never wasted.
A high-capacity tape back-up device would be convenient.
<> 01.008 Can I run OpenIngres on machine ABC?
It is believed that the entire OpenIngres product is presently available
for at least the following machines/operating systems:
Intel machines running Windows NT
Intel machines running OS/2 (strictly Net) --OpenROAD/W4GL in beta
Bull DPX/20
Convex (** see below)
DEC Alpha OSF/1
DEC VAX BSD
DEC VAX Ultrix
DEC VAX VMS
DG AViiON DG/UX
HP 9000 HP/UX
IBM HACMP/6000
IBM RS/6000 AIX
ICL DRS 6000
ICL VME
MOTOROLA 88000 System V
NCR Series 3000
NCR Series 3600 (massively parallel)
PYRAMID OSx BSD
PYRAMID SYSTEM V
SCO Open Desktop
SCO UNIX
SEQUENT DYNIX/PTX
SIEMENS MX
Sun SPARCsystems Solaris 2.x
Sun SPARCsystems SunOS 4.1.x
TANDEM Integrity Nonstop-UX
In addition, the OpenIngres clients are available for:
386/486 machines running MS-DOS
386/486 machines running MS-Windows
DEC AXP OpenVMS
There is NO server product available for MS-DOS or MS-Windows.
There is NO OpenIngres product available for Silicon Graphics, but this
is under development and should be available by late 97/early 98.
[** OpenIngres 6.3 runs on the Convex C2 and C3 under Covex OS 10.1
but the current/future status of this product is unknown to me. Check
with your local Computer Associates office.]
The NAIUA web site currently lists platforms where OpenIngres products were
recently made generally available, as well as platforms which are
currently in Beta release. This information is furnished to the NAIUA
by Computer Associates representatives. You can find this information
on the following web page:
http://www.naiua.org/faqs.html
Be sure to contact your local Computer Associates office for
authoritative information.
<> 01.009 Can I run OpenIngres over network ABC?
There are OpenIngres/NET network adaptors for at least:
TCP/IP
DECnet
SNA LU0
SNA LU62
Wollongong TCP/IP
DEC UCX TCP/IP
IBM TCP/IP
KNET TCP/IP
Be sure to contact your Computer Associates office for authoritative
information.
<> 01.010 Where is the INGRES FAQ?
The INGRES FAQ (~300kb) is archived at:
ftp.naiua.org in /pub/ingres
Updates to the FAQ seem to get posted very infrequently at the moment,
especially when I am distracted by actual work.
The following is an example of the dialogue with the FTP server; lines
that include user input are marked with a `->' symbol:
-> ftp ftp.naiua.org
Connected to ftp.naiua.org.
...
-> Name (ftp.naiua.org:): anonymous
331 Guest login ok, send your complete e-mail address as password.
-> Password: juser@somewhere.org
...
230 Guest login ok, access restrictions apply.
-> ftp> cd /pub/ingres
...
250 CWD command successful.
-> ftp> get Ingres-FAQ
200 PORT command successful.
150 Opening ASCII mode data connection for Ingres-FAQ (234847 bytes).
226 Transfer complete.
local: Ingres-FAQ remote: Ingres-FAQ
240460 bytes received in 45 seconds (5.2 Kbytes/s)
-> ftp> quit
221 Goodbye.
<> 01.011 Where can I get the OpenIngres TPC benchmark results?
Ingres is not submitted for TPC benchmarking (or at least the
results are not submitted to TPC for publication). This may soon change;
see below.
You can get the TPC bylaws and the specifications for TPC-A, TPC-B, and
TPC-C, as well as the results for systems that ARE tested from the
Transaction Processing Council web site:
http://www.tpc.org/bench.descrip.html
The Transaction Processing Council (TPC) can be contacted c/o
Shanley Public Relations
777 N. First Street, Suite 600
San Jose, CA 95112-6311
Tel (408) 295-8894
FAX (408) 295-9768
shanley@cup.portal.com
The Transaction Processing Council is a non-profit organization formed
in 1988 to develop standard benchmarks for transaction processing
systems. The council has more than 40 members consisting of RDBMS
vendors and hardware vendors.
Be aware that there is presently considerable controversy surrounding
the validity of the TPC-A and TPC-B benchmarks. Some commentators have
asserted that certain vendors were able to exploit unrealistic features
of the benchmark specifications to achieve results that are better than
they would have obtained with a less imaginative interpretation of what
was required. So far the same criticism has not been made about the
more realistic TPC-C benchmark.
The thing to keep in mind when looking at some of these results is that
a system's performance is expressed (among other ways) as a ratio of
the number of transactions per second per dollar (tps/$). This means
that if you can get tps high enough, you can spend ridiculous amounts
of money and still get a pretty good tps/$ ratio. For instance, if you
have a really fast 64-bit CPU it may be advantageous to spend an
astronomical sum to eliminate disc I/O entirely by buying enough RAM to
load the entire database into memory--say 40Gb of RAM--because the tps
will be so high that it will still yield a reasonable--even
attractive--tps/$ ratio. It remains to be shown that this gives any
insight into the performance to be expected from the same software
running on more pedestrian hardware.
While this may make the TPC-A and -B benchmarks less useful for
comparing different database products, they may still be useful for
comparing different hardware platforms for the same database product,
ie for helping with the question: what is the best hardware to run DBMS
X on?
Interestingly, CA management have twice committed to performing and
publishing TPC benchmarks for OpenIngres 2.0, once at Ingres World
'95, and again at Ingres World '96. They would probably only do this
if they seriously expect to blow someone out of the water.
As of Ingres World '97, CA management is now stating that they expect
TPC benchmarks to be published by late this year (1997). Assuming they
are as good as CA believes they will be, you should certainly hear about
it when they come out.
<> 01.012 Is postgres anything to do with INGRES?
Postgres is an experimental relational database management system
developed at UC Berkeley. It is intended to demonstrate how object
oriented data management can be accomplished without discarding the
relational model. (See Stonebraker's writings on the subject,
under the heading The Third Generation Database System Manifesto.)
The project is directed by Professor Michael Stonebraker, one of
the originators of Ingres. Postgres is a spiritual descendent of
University Ingres but is not truly related to INGRES.
Some very exciting news on the Postgres front is the release of
PostgresSQL (or Postgres95 as it was initially known). PostgresSQL has
many fascinating features, and it is the only public domain database
software with a native SQL interface. (Which is not to suggest that
having an SQL interface is in any way a good thing, but many people
have been asking for a PD RDBMS that supports SQL.) PostgresSQL runs
on AIX, Linux, and possibly NT.
It is now available for ftp as
ftp://ftp.postgreSQL.org/pub/postgresql-v6.1.1.tar.gz
For full information, see:
http://www.postgresql.org
http://s2k-ftp.cs.berkeley.edu:8000/postgres95
Potentially even more exciting is that OpenLink Software offers a native
PostgresSQL JDBC driver to allow Java applications to run against
PostgresSQL databases. See http://www.openlinksw.com for more information.
Mailing Address Purpose
======= ======= =======
pgsql-announce@postgresql.org Postgres discussion and announcements
pgsql-bugs@postgresql.org Postgres bug reports
pgsql-questions@postgresql.org Questions to the developers of Postgres
More info, including how to subscribe-unsubscribe to various Postgres
mailing lists, can be found on this web page:
http://www.postgresql.org/supp-mlists.shtml
[Thanks to Paul M. Aoki and Jolly Chen of UC Berkeley for some of this
information.]
There is a commercial version of postgres called `Illustra' (at one time
called `Montage' and also `Miro'). Illustra was recently acquired by
Informix. You can find more info at the Informix web site:
http://www.illustra.com
-------------------
Trivium: Illustra was originally called `Miro'; evidently the joke of
naming products after artists lived on briefly--Miro was a student of
Ingres'.
<> 01.013 Is INGRES secure?
OpenINGRES claims to provide C2 levels of security functionality.
The multi-level secure variant of INGRES (INGRES/Enhanced Security) has
already been certified to provide F-B1 levels of functionality at the
E3 ITSEC assurance level. OpenINGRES/Enhanced Security (the next
release) will shortly be submitted to the US NCSC for B1 evaluation.
INGRES/Enhanced Security is NOT presently certified to be NCSC B1
secure.
To find out what the NCSC security designations mean, consult the
"Orange Book":
Department of Defense Trusted Computer System Evaluation Criteria
DoD Directive 5200.28-STD
National Computer Security Center,
Fort Meade,
MD 20755-6000,
USA
Attn: Chief, Computer Security Standards.
Office of Standards and Products
There is also a "Red Book" covering network security.
<> 01.014 Does INGRES run on the PC under MS-DOS or MS-Windows?
The server does not run under DOS or Windows (but it DOES run under
Windows NT). The INGRES tools do run under DOS and Windows and use
INGRES/Net to access a remote database on a server.
Relational Technology did briefly market a DOS version of INGRES
Release 5 that included the "engine", ABF, QBF and Report Writer, but
that was discontinued.
<> 01.015 How can I subscribe/unsubscribe to INFO-INGRES?
INFO-INGRES is a mailing list that may be of interest to anyone without
access to a news service (for comp.databases.ingres), or to anyone
whose news service is read-only. INFO-INGRES mailings are cross
posted to comp.databases.ingres, and vice versa, so theoretically
the same articles should be available from both sources.
INFO-INGRES is now maintained using majordomo list software. To subscribe to
the list, send mail to the following address:
majordomo@ams.org
then put the following line in the message body:
subscribe info-ingres
To cancel your subscription to the list, once again email to:
majordomo@ams.org
and put the following line in the message body:
unsubscribe info-ingres
You may also subscribe to a digested version of the list, info-ingres-digest.
To send mail to the mailing list, send mail to:
INFO-INGRES@AMS.ORG
If you can't get onto or off the list using E-mail you can try calling
Tom Blythe at (401) 455-4015.
There is also a Finnish Ingres users mailing list. You can join the
list by mailing to address 'fiua-request@audit.fi' . Leave subject line
empty, and put into the message body a single line containing word
'subscribe'. In return mail, you will receive a subscription confirmation
and further information about how to post to the list. For more
information, contact
Matti Salo
Tel +358-400-901566
Haapalammentie 13 a 27
40801 VAAJAKOSKI FINLAND
mail matti.salo@audit.fi
<> 01.016 Can I get the FAQ by E-mail?
The FAQ is posted irregularly. Between postings it is
available from the archive site using ftp (see section 01.010). If
you do not have ftp access to the archive site, you can obtain the
FAQ by mail using one of the many ftpmail servers.
[The following information is provided by Bruce Horrocks and Tom Northey.]
The ftpmail servers use different software and the commands required
will differ from server to server. Try sending a message consisting of
just the word "help".
For some systems, the body of the message should be something like this:
open address [username [password]]
mode ascii
cd /pub/directory/etc
get filename
where address is an IP address, username and password are the username
and password to log into that machine as. Normally these are omitted
and they default to "anonymous" and your own mail address as the
password. Default transfer mode seems to be binary with files coming
over uuencoded, hence the switch into ascii mode.
In reply you get a confirmation of receipt mail message with an ID
number that lets you cancel your request. Assuming all goes well you
get the file requested as a mail message sometime later. Large files
are blocked up into 64k messages and need to be reassembled before
uudecoding.
Here is a list of ftpmail servers:
USA
===
ftpmail@decwrl.dec.com
ftpmail@sunsite.unc.edu
bitftp@pucc.princeton.edu
Europe
======
bitftp@vm.gmd.de
ftpmail@ftp.uni-stuttgart.de
ftpmail@ftp.inf.tu-dresden.de
ftpmail@grasp.insa-lyon.fr
bitftp@plearn.edu.pl
ftpmail@ftp.inf.tu-dresden.de
ftpmail@ftp.uni-stuttgart.de
ftpmail@doc.ic.ak.uk
Australia
=========
ftpmail@cs.uow.edu.au
<> 01.017 Can I get the FAQ on discette?
Yes. As of release 1.5 (13JUL94) it is available on discette at no
charge from Quest Software or Common Sense Computing. To get a copy, write
to: info@oz.quests.com or
Europe Elsewhere
====== =========
Common Sense Computing Quest Software Pty Ltd
Canada House 10 illowa St
272 Field End Rd East Malvern Vic 3145
Eastcote, Middx. HA4 9NA Australia
UK
Phone: 081 866 4400 Phone: 03 885 5829
FAX: 081 429 2848 FAX: 03 885 0116
NOTE: Common Sense Computing and Quest Software distribute the FAQ at no
charge as a service to the user community and are not in any way responsible
for the content of the FAQ. All corrections and flames should be
directed to ingfaq@naiua.org
<> 01.018 Is the FAQ available on WWW?
Yes, but not in a really convenient form.
You can try looking at
http://www.naiua.org/faqs.html
A much better implementation of the FAQ in hypertext format will
be announced soon.
<> 01.019 Where can I get a comparison of RDBMS products?
In November of 1996 Martin Rennhackkamp published a comparison of Oracle,
Informix, DB2, OpenIngres, Sybase, and Microsoft SQL Server in DBMS
Magazine. See:
http://www.dbms.mfi.com/9611d52.html
<> 01.020 Can I use a RAID with Ingres?
Yes. As far as Ingres is concerned it is just disk space.
RAID stands for Redundant Array of Inexpensive Disks. Just what it means
beyond that is up to you. There are generally held to be 6 non-trivial
variants of RAID, each with different properties. All offer some degree
fault-tolerance and redundancy, but there are performance and price
tradeoffs.
For many people RAID means RAID 5. RAID 5 has many virtues but write
performance is not one of them. Write-intensive applications would do
well to choose another configuration. Also, anecdotal evidence claims
that RAID 5 is not as reliable as theory predicts--possibly just due to
complexity.
RAID 0+1 is comparitively expensive (duplicate drives) but it offers
outstanding performance both reading and writing, and reliability is
claimed to be very good. Those who advocate using large numbers of
smaller drives instead of a few large drives, but who worry about the
the short MTBF (Mean Time Between Failures) when using many drives,
should find RAID 0+1 quite attractive.
For more information on RAID consult:
http://www.compumedia.com/~kkirk/raid.html or
http://www.storage.com