Improved Buffer Pool Scalability

The InnoDB buffer pool is a well known point of contention when many queries are executed concurrently. In XtraDB, the global mutex protecting the buffer pool has been split into several mutexes to decrease contention.

This feature splits the single global InnoDB buffer pool mutex into several mutexes:

Name Protects
flush_state_mutex flushing state of dirty blocks
LRU_list_mutex LRU lists of blocks in buffer pool
flush_list_mutex flush list of dirty blocks to flush
free_list_mutex list of free blocks in buffer pool
zip_free_mutex lists of free area to treat compressed pages
zip_hash_mutex hash table to search compressed pages

The goal of this change is to reduce mutex contention, which can be very impacting when the working set does not fit in memory.

Version Specific Information

Other Information

Detecting Mutex Contention

You can detect when you suffer from mutex contention in the buffer pool by reading the information provided in the SEMAPHORES section of the output of SHOW ENGINE INNODB STATUS:

Under normal circumstances this section should look like this:

SEMAPHORES
----------
OS WAIT ARRAY INFO: reservation count 50238, signal count 17465
Mutex spin waits 0, rounds 628280, OS waits 31338
RW-shared spins 38074, OS waits 18900; RW-excl spins 0, OS waits 0

If you have a high-concurrency workload this section may look like this:

1 ----------
2 SEMAPHORES
3 ----------
4 OS WAIT ARRAY INFO: reservation count 36255, signal count 12675
5 --Thread 10607472 has waited at buf/buf0rea.c line 420 for 0.00 seconds the semaphore:
6 Mutex at 0x358068 created file buf/buf0buf.c line 597, lock var 0
7 waiters flag 0
8 --Thread 3488624 has waited at buf/buf0buf.c line 1177 for 0.00 seconds the semaphore:
9 Mutex at 0x358068 created file buf/buf0buf.c line 597, lock var 0
10 waiters flag 0
11 --Thread 6896496 has waited at btr/btr0cur.c line 442 for 0.00 seconds the semaphore:
12 S-lock on RW-latch at 0x8800244 created in file buf/buf0buf.c line 547
13 a writer (thread id 14879600) has reserved it in mode  exclusive
14 number of readers 0, waiters flag 1
15 Last time read locked in file btr/btr0cur.c line 442
16 Last time write locked in file buf/buf0buf.c line 1797
[...]
17 Mutex spin waits 0, rounds 452650, OS waits 22573
18 RW-shared spins 27550, OS waits 13682; RW-excl spins 0, OS waits 0

Note that in the second case you will see indications that threads are waiting for a mutex created in the file buf/buf0buf.c (lines 5 to 7 or 8 to 10). Such an indication is a sign of buffer pool contention.

Percona Server
Call Us
+1-888-316-9775 (USA - Sales)
+1-208-473-2904 (USA - Sales)
+44-208-133-0309 (UK - Sales)
0-800-051-8984 (UK - Sales)
0-800-181-0665 (GER - Sales)
+1-877-862-4316 (Emergency)
+1-855-55TRAIN (Training)
+1-925-271-5054 (Training)

Table Of Contents

Previous topic

Percona Server In-Place Upgrading Guide: From 5.5 to 5.6

Next topic

Improved InnoDB I/O Scalability

This Page



© Copyright Percona LLC and/or its affiliates 2009-2014.
Except where otherwise noted, this documentation is licensed under the following license:
CC Attribution-ShareAlike 2.0 Generic
Created using Sphinx 1.1.3.
This documentation is developed in Launchpad as part of the Percona Server source code.
If you spotted innacuracies, errors, don't understood it or you think something is missing or should be improved, please file a bug.
]]>