@LordHunter317 said:
Windows software that can't tolerate spaces in filesnames is in the minority these days.
If you insist.
@LordHunter317 said:
.To me, the fact that so many people have had to write about it tells
me it's very difficult to figure out and use. Oracle is an
extremely complex beast, evidenced by the fact DBAs frequently can't do
development roles, and developers frequently can't do DBA
roles. Which is just silly when you think about it.
It is silly, but it's a cultural thing that occurs across databases,
not just Oracle. Goes the same for DB2, Sybase, or
SQL Server. It's about specializing competence and
interest. It sucks because of the barriers to communication
that are erected between camps. But it's hard to specialize in multiple areas (though some people do it successfully)
@LordHunter317 said:
Save for initial setup and maintaince (which really should be
automatic) all the DBA should ever have to do is change the backup
tapes.
Oh. My. God. Do you have any idea how much work goes into
designing and maintaining a large scale database with hundreds to
thousands TPS and hundreds of gigs to terabytes of data ?
System admins & DBA's need to think about:
I/O Design and Layout - how many disks are needed, what kind of
communications infrastructure (Fibre Channel, SCSI, NAS, etc.) ,
how should the datafiles be spread (RAID 5, 1+0, or extent-based?
Or are you using EMC?), etc
Physical schema design - Should this be a heap or index-organized
table (similar to SQL server's clustered index vs. heap
table)? Should I partition the table? Put it in a
cluster? A hash cluster? What indexes should I
use (b-tree, reverse, function, bitmap, domain specific, full
text)? Should I use materialized views to speed up some
reporting queries? What refresh policy should they have? Do
I need external tables, or just use SQL*Loader for flat file
loads? etc.
Capacity planning - how to build up resources over time, anticipating
breaking points, noting bottlenecks, etc. You can't just
"throw hardware" at a problem without understanding it, for example,
adding CPU's can make a system SLOWER if you're not careful.
Performance tuning & SQL tuning - Should the server be a
shared-threading model or dedicated server model? Should parallel
processing be used for some tables? There's also compensating for
poorly written PL/SQL or SQL code (possible in any language,
really). What kind of locking / isolation mode should be used for
which transactions? Will I need user-level locks at any point?
Backup and recovery design and maintenance -- what's the mean time to
recover for a) component failure, b) media failure, c) backup media
failure, d) catastrophic failure, e) user error, f) block
corruption, etc. How much impact do the hot backups have on
performance & how should they be spread out to
compensate? Develop recovery test scenarios, practise
them, come up with new ways your DB can fail and ensure you're ready
Security integration - Do we use internal roles & users or external
(LDAP)? How do they map? What about data-level
security? Do we use views? fine grained access
control? Do we require certificate-baesd authentication
(x509)? TLS for the database connection? Will I need to set
up synonyms to link schemas? etc
Monitoring and troubleshooting - Are there pending two-phase commit
transactions? Any resources being locked excessively? What
are my hot SQL's, what wait events are causing the most
bottlenecks? Any new patterns in load or data volume that
may impact my capacity or backup & recovery plans?
Patch and feature management -- testing new security patches and
major service packs for weird problems with your application, bringing
new features to developer's attention, etc.
A lot of this is true for Windows and *NIX system administration in
general. You seriously think all of this can be
automated? Some can
be assisted by automotons, but it still needs human judgement.
@LordHunter317 said:
PostgreSQL w/ pg_autovaccum seems completely maintaince void, save for
the occasional need for index recreation in some corner cases.
They're working to eliminate that too, as I understand it.
So it's more than possible.
I'm not sure your point here. Oracle's equivalent locking model
to PostgreSQL (superior, IMHO, because of flashback queries), has never
required vacuuming, the rollback segments are a circle. And it's
gotten easier to design these, with undo tablespaces in 9i, allowing
time-oriented retention policies for older versions.
Could software get more automatic for mundane stuff, or
hard-to-troubleshoot stuff? Absolutely! 10g has a lot of
automatic maintenance stuff for SQL tuning, parameter tuning, and stats
collection. The new web-based OEM has a relatively usable
interface to setup RMAN scheduled backups without resorting to the
command line RMAN.
@LordHunter317 said:
My Free oracle account seems to not be able to access anything.
otn.oracle.com , click documentation or downloads. I have no special privileges.
@LordHunter317 said:
the database download is FREE,
And is severly limited. It works for development, but that's
it. It may not even work for delpoyment testing due to the
connection limitations.
Absolutely NOT. The downloads are for the FULL,
UNRESTRICTED enterprise edition of Oracle's products. You're
expected to purchase a license to deploy it legally. But there is
no technical limitation whatsoever.
@LordHunter317 said:
It's also MS' standard so I'll have to kindly suggest that you're the one on crack..
Fair enough, it's a taste and historical frustration thing, I suppose.
@LordHunter317 said:
Very little military stuff is written in Ada, as it turns out.
No idea, I'll take your word for it.
@LordHunter317 said:
I'm no Oracle expert, but I believe a tablespace is much closer (to a SQL server database).
Fair enough.