Comments on: I wish .. (1)

By: I wish (2) « Oracle Scratchpad

I wish (2) « Oracle Scratchpad — Sun, 22 Aug 2010 14:23:11 +0000

[...] under: SQL Server — Jonathan Lewis @ 6:13 am UTC Jun 18,2010 Continuing my little list of things in SQL Server that I’d like to see in [...]

By: Jason Bucata

Jason Bucata — Sat, 05 Jun 2010 21:35:58 +0000

I thought I read somewhere, maybe 6 months ago, that Oracle 11g and/or Exadata has merry-go-round tablescans.

Of course, I can’t find it now…

I know for sure I’ve heard of that concept before–and I never pay attention to SQL Server news so it couldn’t have been from there.

By: Jonathan Lewis

Jonathan Lewis — Fri, 04 Jun 2010 08:40:55 +0000

Noons,

It is interesting to note, though, that some of the very large databases that use SQL Server deal with the scalability problem by partitioning their main tables by using separate databases for separate partitions – so you get, for example, one database per day and have to write code that queries across databases.

By: Noons

Noons — Fri, 04 Jun 2010 01:45:58 +0000

@Garry:
very simply, I don’t let it happen.
Programmers trying to access data cross-dbs or cross-schemas can only succeed if they are authorized to do so. That’s why I said the security becomes critical in these environments: relax it and you are in for a nasty surprise.
SS has extensive control of security at both database and server level. Oracle allows this: “alter session set current_schema=xxxx” in a login trigger, just as an example. And a few other things, as of course you are aware. So rather than excluding dbs/schemas by segregating into instances, I exclude them by security. Works very well indeed.

By: joel garry

joel garry — Thu, 03 Jun 2010 22:17:47 +0000

How do you avoid the risk of programmer-controlled locking strategies making redo contention look like a walk in the park?

By: Parthiban Nagarajan

Parthiban Nagarajan — Thu, 03 Jun 2010 12:33:18 +0000

IMHO … Oracle should implement the feature “Filtered Indexes” of SQL Server …

By: Igor

Igor — Wed, 02 Jun 2010 17:53:18 +0000

Hi,

And why not separate dictionary data about “database”. So, that things are self contained, independent. To have transportable “database” (including code, sequences etc). Availability – you don’t care if SYSTEM tablespace of “another database” is corrupted.

Restore/recovery on “database” level. As consequence simplified / faster procedure for PITR.

Of course, risk is to have costly dictionary queries, security concerns…

Just my 2cents.

Regards,
Igor

By: Tom

Tom — Wed, 02 Jun 2010 15:44:55 +0000

ExaData does have that giant flash cache though.

By: Tom

Tom — Wed, 02 Jun 2010 15:44:05 +0000

They added Consistency in 2005 but it wasn’t the default. I tried it back then and I found the performance to be atrocious. Could be one of the reasons why they neglected to turn it on. Another could be the fact that many apps are written against SQL Server with that in mind and changing the way the DB works could have a drastic affect on those applications. I haven’t tried 2008 yet, guess I probably could (time permitting).

By: Tom

Tom — Wed, 02 Jun 2010 15:41:52 +0000

I’ve played with both SQL Server and MySQL quite a bit and to me, MySQL uses a similar model to SQL Server. You have one set of processes managing multiple databases. I haven’t peaked lower to see what it does at the log level. I guess I probably should now that MySQL is part of the Oracle family.