Percona Server with XtraDB

Add Busy_time, Query_time, and Uptime_hires

Reported by Baron Schwartz on 2010-10-26
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Percona Server
Wishlist
Unassigned

Bug Description

I'd like the following global variables in SHOW STATUS:

Busy_time is the total wall-clock time during which user threads were executing queries. This is not the same as Threads_running > 0, because Threads_running is > 0 when there are slaves or event threads executing.

Query_time is the total query time accumulated by the server, including currently executing queries (this is very important).

Uptime_hires is simply a high-resolution Uptime.

The Busy_time is similar to field 10 in /proc/diskstats. The Query_time is similar to field 11. See http://www.mjmwired.net/kernel/Documentation/iostats.txt for details on this. It is worth reading how the numbers are computed in /proc/diskstats.

This information is amazingly powerful for performance analysis. It lets you do all kinds of things with queueing theory and Little's Law. It lets you apply Neil Gunther's universal scalability law to predict the scaling limits of the server. I have gotten the same information from TCP traffic and validated its usefulness. You can see an 'awk' implementation here: http://aspersa.googlecode.com/svn/trunk/usl Look in the convert_tcpdump_tabulate() function.

This is related to https://bugs.launchpad.net/tcprstat/+bug/634306 but implementing it inside the server is much more powerful.

Changed in percona-server:
status: New → Confirmed
importance: Undecided → Wishlist
Stewart Smith (stewart) on 2011-05-25
Changed in percona-server:
status: Confirmed → Triaged
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers