Bruce Momjian: Postgres Blog

Parallel Query Report from the PGCon Developer Meeting

Thu, 17 May 2012 10:45:01 -0400

As part of yesterday's PGCon Developer Meeting, I hosted a discussion about adding resource parallelism to Postgres, which I blogged about previously. Josh Berkus has kindly summarized the results of that discussion. I am hoping to help track and motivate progress in this area in the coming months; it is a multi-year project to complete.

Template1 Me!

Wed, 16 May 2012 12:00:00 -0400

You might have heard of the template1 database before, or seen it in the output of pg_dumpall:

REVOKE ALL ON DATABASE template1 FROM PUBLIC;
…

Postgres 9.2 Draft Release Notes Published

Thu, 10 May 2012 12:00:00 -0400

I have completed the Postgres 9.2 release notes I started seven days ago. Writing the release notes is always a taxing experience. I have to basically dedicate an entire week to the process of digesting 1100 commit messages to produce 3000 lines of SGML text. Once I am done though, it is rewarding to see the finished product. It is like working on a painting all year, and spending a frustrating week framing it and hanging it on the all — once it is hung, you stand and admire it, and forget much of the sweat it took to produce. I am sure many community members feel the same.

Curious how the 9.2 release item count compares to previous major releases? Here are the results:

Release	Items
9.2	241
9.1	203
9.0	237
8.4	314
8.3	214
8.2	215
8.1	174
8.0	230
7.4	263

Postgres Drinking Game

Mon, 07 May 2012 12:00:00 -0400

My children have traveled to many conferences with me, and have heard many webcasts and training calls at home. I guess, after hearing about Postgres so often, they pick up certain phrases of interest, and the big one for them is "shared buffers". Anytime someone uses those words, they start howling and make up some funny sentence using the words.

There are a variety of beer drinking games that trigger drinking when a word is said — perhaps "shared buffers" is enough on its own to make a drinking game.

The Effectiveness of effective_cache_size

Fri, 04 May 2012 12:00:00 -0400

Having reported the methods for finding the size of the kernel cache on Linux, I wish to highlight the importance of the postgresql.conf setting effective_cache_size.

Unlike other memory settings that control how memory is allocated, effective_cache_size tells the optimizer how much cache is present in the kernel. This is important for determining how expensive large index scans will be. The optimizer knows the size of shared_buffers, but not the kernel cache size, which affects the probability of expensive disk access.

Starting on 9.2 Release Notes

Thu, 03 May 2012 12:00:00 -0400

As in previous years, I have started working on major release notes, this time for Postgres 9.2. I start with this command:

$ src/tools/git_changelog --since '2011-06-11 00:00:00 GMT' --master-only \
> --oldest-first --details-after

Measuring Free Memory and Kernel Cache Size on Linux

Wed, 02 May 2012 12:00:00 -0400

Measuring Linux free memory and kernel cache size can be somewhat tricky. You might look at /proc/meminfo for the answer (commas added):

# cat /proc/meminfo
MemTotal:       24,736,604 kB
MemFree:         3,805,392 kB
Buffers:           743,016 kB
Cached:         18,188,208 kB
...

Caching Levels

Mon, 30 Apr 2012 12:00:00 -0400

There are several levels of caching used in a typical server — here they are, in order of increasing distance from the cpu:

1.	Cpu cache
2.	Random-Access Memory (ram)
3.	Storage controller cache
4.	Storage device (disk) cache

All of these are a fixed size and set at hardware installation time, i.e. you can't move cache from one level to another. For Postgres, there is flexibility in how #2, random-access memory, is allocated, and this provides a never-ending opportunity for administrators to optimize their systems. The three ram allocation possibilities are:

Upcoming Events

Sat, 28 Apr 2012 12:00:00 -0400

During the next two months I will be attending events in the following cities: New York City, Ottawa, Charlotte (North Carolina), and Boston — the details are on my website. I will also be doing training in many of these locations. And I spoke in Philadelphia this week — I guess this is what the New Postgres Era looks like.

View or Post Comments

Why Use Multiple Schemas?

Fri, 27 Apr 2012 12:00:00 -0400

I mentioned that Postgres supports multiple clusters, databases, and schemas, but an open question is why use multiple schemas, rather than placing everything in the "public" schema? By default, search_path places everything in the public schema (assuming a schema matching the current user name does not exist):

test=> SHOW search_path;
  search_path
----------------
 "$user",public
(1 row)