veyepar: Towards One MILLION SQL Queries Per Second

Author(s):	Stewart Smith
Location	OGGB F&P
Date	jan Wed 14	Days Raw Files
Start	15:40	First Raw Start	15:00
Duration	0:45:00	Offset	0:39:20
End	16:25	Last Raw End	16:25
Chapters	00:00
Total cuts_time	44 min.

http://lca2015.linux.org.au/schedule/30203/view_talk
raw-playlist raw-mp4-playlist encoded-files-playlist host public tweet mp4 svg png

assets release.pdf Towards_One_MILLION_SQL_Queries_Per_Second.json
logs
Admin: episode episode list cut list raw files day marks day marks day image_files

State:

Locked:

clear this to unlock

Locked by:

user/process that locked.

Start:

initially scheduled time from master, adjusted to match reality

Duration:

length in hh:mm:ss

Name:

Video Title (shows in video search results)

Emails:

email(s) of the presenter(s)

Released:

has someone authorised pubication

Normalise:

Channelcopy:

m=mono, 01=copy left to right, 10=right to left, 00=ignore.

Thumbnail:

filename.png

Description:

markdown

Over the past 10+ years, MySQL has been dragged kicking and screaming
into the multi-core world and is still the most widely used open source
database.

Ten years ago, MySQL AB's own benchark whitepapers were boasting of
HUNDREDS of transactions per second.

In 2001, what would become MySQL Cluster, on a 72 CPU system, could get
1 million reads per second using the NoSQL C++ NDBAPI.

In 2013, the current stable release (MySQL 5.6) could get 350,000
queries per second on a 32 core system with the current development
release (MySQL 5.7) reaching an astounding 500,000 SQL queries/sec.

Joining IBM at the start of 2014 it was only a matter of time before
someone found out about my (not so) secret life of 10 years knee deep in
MySQL internals.

This is the tale of scaling MySQL throughout the years and scaling MySQL
on the new POWER8 processor.

- Whare are global mutexes and what can you do about them?
- Do atomic variables solve all of your problems?
- How do you deal with NUMA?
- What tools exist to find bottlenecks and find out why you can't get any
more performance out of a system with 42% idle CPU?
- and can you *really* get to 1 million SQL queries per second on a dual socket system?

Comment:

production notes