Re: Serious article on comparison between MS SQL Server 2005 Yukon and Oracle 10

Database replication via OracleMonster.com wrote:
> Daniel,
> Please describe the fact that give you more credibility over other people? Is it the one that you do not respect other people, or that you are screaming loudly? Neither one would justify it for me. If you have facts other than your talks, please provide it here. Otherwise it is just a noise. Sorry if I hurt your feelings
>
> *****************************************
> * A copy of the whole thread can be found at:
> * http://oraclemonster.com/Uwe/Forum.aspx/oracle-server/5185
> *
> * Report spam or abuse by clicking the following URL:
> * http://oraclemonster.com/Uwe/Abuse.aspx?aid=220d02d5c5e9418d8a20a4fe8704516d
> *****************************************

I've not paid much attention to this thread, since it seems to involve Daniel just making sweeping statements as usual. I'd certainly criticise the report on a number of points:

Page 6: "In Oracle 10g you can rebuild and move an index or even a table without exclusive lock". No you can't. An online table reorganisation or an online index rebuild both take exclusive locks. That they don't *hold* them for the duration of the rebuild or re-organisation is irrelevant: they still have to take them in the first place.

Page 7: "Then you can startup additional instance without performing a single change to the database schema". Apart from creating additional undo tablespaces when using automatic undo, presumably? Or is that a loose use of the word 'schema'?

Page 8: "Oracle RAC in versions prior to 10g (...8i RAC...)...". There was no RAC in 8i. It was still called OPS.

Page 8" "9i RAC...has been running on top of OS specific cluster solution such as ... Microsoft Cluster..." If there's one O/S specific clustering technology that 9i RAC does NOT run on, it's Microsoft's Clustering Services.

Page 13: "Oracle has the option to rebuild any kind of index or even whole table without exclusive lock". Not true, again. But also not true because of the "any kind of index" claim. Domain indexes, for example, cannot be rebuilt on-line.

Page 17: "MSSQL provides the ability to create a so-called computed column and the ability to create an index on this column. [This is basically equivalent to Oracle's function based index.]" No it's not. The whole point of the FBI is not to have the computed values resident in the table, but to be able to nonetheless treat them as index-searchable. The MS approach merely adds the data into the table (after which it can obviously be index-searched).

Page 26: "Before 10g, you could estimate by a script... [whereas] in 10g you can easily check out V$FAST_START_TRANSACTIONS." That view is definitely available in 9i, of course.

So there are a number of technical errors, which might indicate a degree of less-than masterly understanding of the Oracle product. But those errors are fairly small in number and don't affect the general thrust of the argument being made when they arise. Of course the grammar is hopeless ... worse, it's simply attrocious. But on the whole, the comparisons are fair, and it's as reasonable a stab at evaluating two very different RDBMSes as I've read in a while.

I suppose my biggest gripe is that discussions of beta products are intrinsically dubious.
Regards
HJR Received on Sat Nov 20 2004 - 22:17:54 CST