Re: Testing Database

From: Geoff Bray <geoff_at_tisdec.tis.tandy.com>
Date: 4 Mar 1994 22:08:31 GMT
Message-ID: <2l8bgv$55j_at_tisdec.tis.tandy.com>


In article <2krhid$15oi_at_nwg.nectec.or.th>, pradit_at_nwg.nectec.or.th (Pradit Pitaksathienkul) writes:
|> Dear all,
|> Can anyone tell me the comparision between Oracle and Ingres ? My
|> University is during decide to buy one for Education , and mis .
|> Ingres's distributor came to my office and install Ingres for test and
|> give me document to show the high performance than Oracle .
|> So I need some details about Oracle .
|> Regards,
|> pradit.

Disclaimer: The following opinion is my own and is not necessarily that of the company.

Experience: I have worked with several RDBMS (Oracle, Ingres, DEC Rdb, and Tandem nonstop SQL ) over the past 6 years.

I have been a DBA/programmer/technical support person for the Oracle side of one of our systems for the past 4 1/2 years. About a year ago we looked at converting to Ingres from Oracle for price considerations. I spent about two months testing and in my opinion Oracle beats Ingres hands down. The testing was done on a DEC 5500 running Ultrix 4.?. Most of our code was done in Embedded SQL (ESQL) so that is pretty much what most of this document refers to. My work with Oracle Forms, Menus, Report Writer was after this test and my only work with Ingres' ABF, etc. was 5 years ago so I'm sure any thoughts I have of that would be incredibly obsolete. Below is a summary of my results.

PERFORMANCE On all of the tests I ran Oracle was significantly faster. i.e. Many of our maintenance programs use a select similar to that listed below:

	SELECT sku, desc, qty, etc
	FROM skumaster
	WHERE sku > 'TESTSKU'
	ORDER BY sku;

Basically we open this cursor and display 10 or 15 records at a time and let the user scroll through them. The table has about 30,000 rows with about 25 columns. To select the first 15 rows takes less than 1 second for Oracle and 58 seconds for Ingres. Ingres suggested a work-around - to just retrieve the index value and then the rest of the row. This decreased the time to 8 seconds for just the index value - still totally unacceptable. Actual fetches of the rows were fine but the initial time was prohibitive.

FUNCTIONALITY Oracle handles row level locking. Ingres does page level locking and escalation of locks. Escalation means after a certain amount of contention a user gets a table lock (horrible). Oracle also automatically deals with deadlocks and Ingres does not.

Oracle allows for outer joins whereas Ingres does not and this is a much more workable solution than a union all. Even the Ingres tech notes state that a union all is "slow" so one can imagine how slow it really is!

Oracle has a decode function that we have found very helpful in certain case. This can be used to return values based on columns not retrieved and is a sweet extension.

The only plus for Ingres in functionality is that is allows for structures in ESQL and Oracle does not. We get around this by defining pointers to structure elements but obviously this is inferior to using a structure.

Oracle has extensive date functions and Ingres date functions are much more limited.

The Ingres facility for loading data from flat files is a piece of junk. I used it 6 years ago and hated it with a passion and so I was very hopeful that vast changes had been made to it when we tried it out a year ago but that was not the case. It is still, in my opinion, a piece of junk. The Oracle facility is incredibly more robust, friendly, etc.

MISC As far as the interactive interface that goes with the database Ingres' ISQL is much more user friendly than Oracle's SQL Plus.

Ingres does support group granting functionality for tables/views/etc and Oracle 6 does not. This is a little hassle that has been fixed in Oracle 7.

The Ingres manuals were much briefer and not near the quality of the Oracle manuals. I know that almost all programmers would contend that good software documentation is harder to find that a hen that lays golden eggs but Oracle is way ahead of Ingres in that area.

As you can see, I clearly think that Oracle is superior. That is not political propaganda because I work for Oracle because I don't. This opinion is from my experience with trying to match what users want to what the database can feasibly do.


Geoffrey Bray                       |  The thoughts and views herein do not
Tandy Information Services          |  necessarily reflect that of Tandy
geoff_at_tisdec.tis.tandy.com          |  Corporation.
------------------------------------------------------------------------------
I kinda wish I was the government so my budget wouldn't have to balance.
Received on Fri Mar 04 1994 - 23:08:31 CET

Original text of this message