Oracle's letter to Russian IT companies • Hacker News Board

Re: Oracle's letter to Russian IT companies

#91

Post by TheLogothete » Mon, Mar 21, 2016, 9:35 AM UTC

chris_wot wrote: post #90

TheLogothete wrote: post #84

Earlier quoted context omitted.

Postgres MVCC is anything but decent. It will take over control of the main heap. Good luck tackling that.

What exactly does that mean? MVCC is definitely different in Oracle and Postgres, but that doesn't mean it "takes over the heap" it just means it stores a new version of the row in the same heap. Later on an auto-vacuum process clears out the old heap page, or sometimes on demand when a select, update or delete accesses the row. Oracle stores the old version of the row in a rollback segment. According to the followin…

The very next sentence:

Both Oracle and SQL Server has some way to restrict the growth of version information whereas PostgreSQL/PPAS doesn't have any way.

Re: Oracle's letter to Russian IT companies

#92

Post by chris_wot » Mon, Mar 21, 2016, 9:56 AM UTC

TheLogothete wrote: post #91

chris_wot wrote: post #90

Earlier quoted context omitted.

What exactly does that mean? MVCC is definitely different in Oracle and Postgres, but that doesn't mean it "takes over the heap" it just means it stores a new version of the row in the same heap. Later on an auto-vacuum process clears out the old heap page, or sometimes on demand when a select, update or delete accesses the row. Oracle stores the old version of the row in a rollback segment. According to the followin…

The very next sentence: Both Oracle and SQL Server has some way to restrict the growth of version information whereas PostgreSQL/PPAS doesn't have any way.

True, this can be an issue. But the auto-vacuum and auto-cleanups should resolve this issue. Have you experienced this as an issue? Genuinely curious.

(Don't speak to me about SQL Server. Of all the brain dead ideas, implementing storage in the TempDB for snapshot isolation!)

Re: Oracle's letter to Russian IT companies

#93

Post by chris_wot » Mon, Mar 21, 2016, 1:12 PM UTC

jl6 wrote: post #85

chris_wot wrote: post #59

Earlier quoted context omitted.

Yeah, but on the bright side once your excised Oracle and replaced it with something better Oracle won't be coming back any time soon :-)

Unfortunately I didn't say we're moving to something better cough SAP :(

My sympathies.

Re: Oracle's letter to Russian IT companies

#94

Post by needusername » Tue, Mar 22, 2016, 7:07 AM UTC

chris_wot wrote: post #87

needusername wrote: post #86

Earlier quoted context omitted.

> What part of parent has not been your experience ? We have several people in house (I know of 6) who can troubleshoot performance issues. For them the CBO is not a black box and they can explain what is happening and why. I know of a local consulting company that has several very good people that you can "order". Said consulting company also offers very good CBO courses/workshops.

Really? I have a book on CBO performance troubleshooting from the foremost expert and even he says that he can't fully understand why the CBO does the things it does. I doubt your in house experts fully understand the Oracle CBO. They may claim they do, but unless they have access to material that isn't public they can't possibly be able to diagnose the CBO fully. You realise there are over a thousand hidden database…

> Really? I have a book on CBO performance troubleshooting from the foremost expert and even he says that he can't fully understand why the CBO does the things it does.

> I doubt your in house experts fully understand the Oracle CBO. They may claim they do, but unless they have access to material that isn't public they can't possibly be able to diagnose the CBO fully.

Maybe. Let me put it this way, so far they could explain all CBO decisions where we had performance issues.