Re: native Oracle-port on Linux -- what would it take?

Matthew Hannigan schrieb in Nachricht <67amng$glv$1_at_the-fly.zip.com.au>...
>In article <679vcl$5j1_at_pell.pell.portland.or.us>,
>david parsons <o r c @ p e l l . p o r t l a n d . o r . u s> wrote:
>>In article <01bd0b13$3da5c480$2e5c0c26_at_sfinance3>,
>>S V <sv1_at_mindspring.com> wrote:
>>>Linux is missing many features for any OS to be worth considering
>>>as a viable database platform.
>>>1. Linux has no logical volumes layer.
>>
>> You don't need that for a database.
>>
>>>2. Linux has no transaction-oriented filesystem.
>>
>> You don't care about that for a database.
>
>Yeah, last I heard the db does transactions, not the fs!

In fact there is a difference between db transactions (these are the activities in the database) and os/filesystem transactions. Oracle doesn´t need file system transactions although it makes life more secure. Filesystem transactions prevent the oracle files that are on the filesystems (redo logs, control files, tablespace datafiles) from being destroyed when a system failure arises. Redologs are for storing db transactions in case the system fails before the (db)transactions are written to the datafiles. After a reboot Oracle will apply the completed transactions which are in the redologs but not in the datafiles. Datafiles on filesystems - these are the files where the tables exists - will (in most cases) not profit from this fact because the datafiles will never grow! Oracle does NOT need any raw devices. Using raw devices (unbuffered character special devices - no filesystem buffering, no filesystem overhead) will gain a few performance points but you will buy this with a more complex backup and recovery concept (if you have one ;-) because you can´t manipulate this devices with tar or cpio et al but you have to use dd to dump the whole device (as far as it is used by the raw device data file) to tape or from tape. I am using Oracle since Version 5.1b and except 2 tries with Oracle 7.1 I never used unix raw devices. You can get the performance boost using a separate controller/drive for redologs/controlfiles and one for datafiles. The costs are even less the costs you will have using and administering your raw devices but the performance of a multi controller/drive system is much higher than using raw devices.
>
>>>3. Linux has no support for raw devices - hence NO even remote
 possibility
>>> to run Oracle Parallel Server.

Why for heavens sake will you use the parallel server? If Performance is a problem use more controllers/disks or faster processors (alpha or sparc or power pc ...) or divide your database in two parts running on different linux servers connected with fast ethernet ... If you want to user Parallel Server and/or Parallel Query go and buy an expensive multi processor system (compaq, ibm aix, sun ...) and use it without problems. But we are talking of linux and cheap but reliable systems.
>>

- snip -

Andreas Received on Tue Dec 23 1997 - 00:00:00 CST