Comments on: MyISAM concurrent insert http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/ Everything about MySQL Performance Fri, 05 Dec 2008 07:57:47 +0000 http://wordpress.org/?v=2.5.1 By: Marc Pinel http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-337964 Marc Pinel Thu, 31 Jul 2008 15:01:45 +0000 http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-337964 Hi man, i have concurrent-insert=2 set....how would the positively impact the performance? i have a torrent tracker (legal content ofcourse) with heavy php and mysql visits daily...and some spikes. Hi man,

i have concurrent-insert=2 set….how would the positively impact the performance? i have a torrent tracker (legal content ofcourse) with heavy php and mysql visits daily…and some spikes.

]]> By: Online Car Insurance Rate Quotes http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-2865 Online Car Insurance Rate Quotes Thu, 28 Sep 2006 07:36:21 +0000 http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-2865 <strong>Online Car Insurance Rate Quotes...</strong> Online Car Insurance Rate Quotes... Online Car Insurance Rate Quotes…

Online Car Insurance Rate Quotes…

]]> By: Vadim http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-2766 Vadim Mon, 25 Sep 2006 06:16:20 +0000 http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-2766 Jerry, Well... There are several points: 1. Possible negative effect - your INSERTS will be a bit longer. I don't know any bugs with disabled concurrent insert. 2. If you use only point selects, or selects that retieve only several rows - I don't think concurrent=off will help you much - perhaps otherwise. 3. if you use select retrieves many rows - you definitely should use concurrent=off 4. if your database is more write that read - then I again don't think concurrent is important for you - most time tables are in WRITE lock. 5. But better way - test it - for example you can take mysql with patched query log and analyze execution time of queries with concurrent insert = off and on. Jerry,

Well…
There are several points:
1. Possible negative effect - your INSERTS will be a bit longer. I don’t know any bugs with disabled concurrent insert.
2. If you use only point selects, or selects that retieve only several rows - I don’t think concurrent=off will help you much - perhaps otherwise.
3. if you use select retrieves many rows - you definitely should use concurrent=off
4. if your database is more write that read - then I again don’t think concurrent is important for you - most time
tables are in WRITE lock.
5. But better way - test it - for example you can take mysql with patched query log and analyze execution time of queries with concurrent insert = off and on.

]]> By: Jerry http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-2756 Jerry Sun, 24 Sep 2006 16:41:17 +0000 http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-2756 What is the negative impact to turn the concurrent_insert off? Will it make insert waits much longer? Is there any bug or issue related to turn off this default feature? Will turning off the feature help queries which have no insert on the same table at the same time? My system has update:Insert:Select ratio is 3:1.5:1 so the db is write more than read. Thanks. What is the negative impact to turn the concurrent_insert off? Will it make insert waits much longer? Is there any bug or issue related to turn off this default feature?
Will turning off the feature help queries which have no insert on the same table at the same time?

My system has update:Insert:Select ratio is 3:1.5:1 so the db is write more than read.

Thanks.

]]> By: Vadim http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-260 Vadim Fri, 23 Jun 2006 13:31:39 +0000 http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-260 Citrin, I usually use sysbench (http://sysbench.sf.net). Citrin,

I usually use sysbench (http://sysbench.sf.net).

]]> By: citrin http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-259 citrin Fri, 23 Jun 2006 13:22:58 +0000 http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-259 Which soft was used in this benchmark? Which soft was used in this benchmark?

]]> By: Vadim http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-194 Vadim Sat, 17 Jun 2006 05:15:39 +0000 http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-194 I did not test concurrent-insert = 2, but looking at source code - it changes SELECTs code in no way. So the results will be the same. I did not test concurrent-insert = 2, but looking at source code - it changes SELECTs code in no way. So the results will be the same.

]]> By: Apachez http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-187 Apachez Fri, 16 Jun 2006 21:26:24 +0000 http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-187 How does the "concurrent-insert = 2" feature described by Arjen affect the outcome of the above benchmark ? How does the “concurrent-insert = 2″ feature described by Arjen affect the outcome of the above benchmark ?

]]> By: Vadim http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-154 Vadim Tue, 13 Jun 2006 12:11:13 +0000 http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-154 Jay, It happens even with only SELECT queries. The problem is pthread_rwlock_rdlock is called for each scanned rows - that is the more rows we are processing the problem is worse. I don't have exactly number of rows - but I saw the same problem for 10.000 rows. Good idea for benchmarks to check when the problem popups - I will try it soon. Jay,

It happens even with only SELECT queries.
The problem is pthread_rwlock_rdlock is called for each scanned rows - that is the more rows we are processing the problem
is worse.
I don’t have exactly number of rows - but I saw the same problem for 10.000 rows.
Good idea for benchmarks to check when the problem popups - I will try it soon.

]]> By: Jay Pipes http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-153 Jay Pipes Tue, 13 Jun 2006 11:56:39 +0000 http://www.mysqlperformanceblog.com/2006/06/13/myisam-concurrent-insert/#comment-153 Hi Vadim! Great info. Could you by any chance give us a heads up on what the "tipping point" is where you see scalability issues with concurrent insert? Is there a specific % read/write where MyISAM starts to have trouble, or does the situation normally occur when there is a huge range condition like the one above (20000 rows)? Cheers! Jay Hi Vadim!

Great info. Could you by any chance give us a heads up on what the “tipping point” is where you see scalability issues with concurrent insert? Is there a specific % read/write where MyISAM starts to have trouble, or does the situation normally occur when there is a huge range condition like the one above (20000 rows)?

Cheers!

Jay

]]>