Re: SQL Server on Linux

From: Mladen Gogala <>
Date: Tue, 31 May 2016 14:02:49 +0000 (UTC)
Message-ID: <>

On Tue, 31 May 2016 15:23:46 +1000, Noons wrote:

> Funny how they never grew as much since then. Wonder why, with all the
> geniuses helping?...

I used to like Pete Sharman, especially his work on oracle-l, but he seems to have switched to marketing and is now trying to sell OEM. OEM is an attempt to replace Nagios, Zabbix, Cacti and alike, but I am not very optimistic. I would always select Nagios over OEM, because of two things:

  1. MUCH larger installed base and the true open source community.
  2. It doesn't have WebLogic. I don't use WebLogic, if it is up to me.

That would probably put me at odds with Pete, had he not changed the profession.

> Since around 98 they started to abuse licensing and since 2009 they've
> been a joke, around here. To the point where they can't even get enough
> attendees Australia-wide to run a yearly conference!...

I used to be a regular on NYOUG, but not any more. There is nothing useful for me as a techie on NYOUG. Oracle closed the technical information and is only dispatching the much needed technical info to chosen few, and not on the conferences. Now that you mentioned 12c, it has multi-threaded log writer, based on polling, rather than post-wait event model. Of course, there is no documentation. The only documentation that I've found is how to turn it off, because of a bug which may cause the DB to hang.
As for Oracle 12c, it is an old wisdom to wait for the .2 version. Oracle tried its best to get people to switch, but I would not do that, if the decision was up to me. Oracle 10.1 was disastrously unstable, just like the early releases of 10.2. I wonder if you remember the bug in the client which would hang after 200 days, necessitating a reboot? The same goes for 11.1 and the early releases of 11.2. I was having fun with the process burning CPU like crazy in and calibrate_io turning on parallel query for almost everything in Oracle even advised the customers to delete the results from the dictionary table containing the results of calibrate_io, which I found hilarious. I am steering clear of automatic degree of parallelism ever since, especially because I don't see what would it buy me, except the need to license more CPU cores. I have always considered parallel query to be reserved for very special queries, with the decision to use parallelism always warranting a careful consideration.
The changes in the 12c optimizer are very substantial, the foremost one being the introduction of the most popular values. There have been other changes with the sharing cursors and switching the join method amid execution, if the optimizer finds discrepancies between the calculated and the actual values. Bugs are to be expected, with such an extensive rewrite. And bugs and the lack of stability in the optimizer plans equal possible performance disaster in the critical applications. Personally, I would not switch before

Mladen Gogala
The Oracle Whisperer
Received on Tue May 31 2016 - 16:02:49 CEST

Original text of this message