Anthony Lopez

“I’d rather see a sermon than hear one any day;”

Archive for the ‘OpenSQL 2008’ Category

Lightning talks

Posted by lopeza on November 15, 2008

tokutek.com Billion row insert contest.

Monitoring & Graphing by Ronald Bradford (dbcollation project)

Cacti by Baron Scwartz
– building framework to make writing templates easy
– new mysql and memcached templates
– mysqlcactitemplates.googlecode.com

slowtar from percona to throttle copies (also rsync)

Advertisements

Posted in OpenSQL 2008 | Leave a Comment »

MySQL Sandbox

Posted by lopeza on November 15, 2008

used to setup multiple mysql servers on a one physical server.

very simple to use command lines which can create standalone sandboxes or a Master slave set.

download it on launchpad

Posted in OpenSQL 2008 | Tagged: | Leave a Comment »

Percona Full Text Search – notes

Posted by lopeza on November 15, 2008

Why Sphix? – Great indexin, search speed, scalibility, online index rebuilds, and better resource utilization.

Mysql
– Indexing on mysql uses btree which can be very slow.

Sphinx
– Speed is greatly higher on sphinx.
– with sphinx you can perform a query and store different groupby sets at the cost of one scan.
– used on many servers
– result sets can be merged together

How is it organized
– searchd – standalone daemon responsible for answering client queries
– indexer – tool to build indexes

Talking to sphinx
– using apis, php, perl, python, ruby, c, c++, java
– sphinxSE – mysql storage engine dedicated to sphinx (good to use for large scale optimizations instead of letting the app do it)

Top New Features
– added select-list feature w/full expressions support
– arbitrary brackets/negations nesting support to query language
– config reload on SIGHUP
– signed 64bit attrs support
– persistent connections, unix socket, and multi-interface support
– kill-list support (delta index)
– MS SQL source type support
– index_exact_words feture, and exact form operator
– inplace inversion of .spa and .spp

http://sphinxsearch.com

Posted in OpenSQL 2008 | Tagged: | Leave a Comment »

Percona Performance Patches

Posted by lopeza on November 15, 2008

Can’t wait for MySQL to release features as they are taking too long.  Percona is trying to solve problems for customers today.

Sources:
google patches and Percona in house development.

Diagnostic patches:
Microslow – Microsecond resolution for long_query_time and for query timing
– log queries into slow log even executed less than one second.
– log queries from slave thread as well.

Userstats (from google)
– information_schema.index_statisticks;
– innodb_io_pattern shows what part of tables accessed, can be used to determine working set
– innodb_show _hash_memory where memory is allocated to.
– innodb_show_buffer_pool
– split_buffer_pool_mutex
– innodb_io_patches

Replication:
– Mirror binlogs to make switching slaves less of an arduous task.

Resources:
http://www.percona.com/docs/diki/devplan:start
http://www.percona.com/docs/wiki/release:start

Posted in OpenSQL 2008 | Tagged: , | Leave a Comment »