Researching a performance issue, we came to a startling discovery:
MySQL 5.7 + binlogs is by default 37-45% slower than MySQL 5.6 + binlogs when otherwise using the default MySQL settings.
Test server MySQL versions used:
i7, 8 threads, SSD, Centos 7.2.1511
--no-defaults --log-bin=mysql-bin --server-id=2
Sysbench version 0.5, 4 threads, socket file connection
sysbench --test=/usr/share/doc/sysbench/tests/db/parallel_prepare.lua --oltp-auto-inc=off --mysql-engine-trx=yes --mysql-table-engine=innodb --oltp_table_size=1000000 --oltp_tables_count=1 --mysql-db=test --mysql-user=root --db-driver=mysql --mysql-socket=/path_to_socket_file/your_socket_file.sock prepare
sysbench --report-interval=10 --oltp-auto-inc=off --max-time=50 --max-requests=0 --mysql-engine-trx=yes --test=/usr/share/doc/sysbench/tests/db/oltp.lua --init-rng=on --oltp_index_updates=10 --oltp_non_index_updates=10 --oltp_distinct_ranges=15 --oltp_order_ranges=15 --oltp_tables_count=1 --num-threads=4 --oltp_table_size=1000000 --mysql-db=test --mysql-user=root --db-driver=mysql --mysql-socket=/path_to_socket_file/your_socket_file.sock run
5.6.30: transactions: 7483 (149.60 per sec.)
5.7.12: transactions: 4689 (93.71 per sec.) — That is a 37.36% decrease!
Note: on high-end systems with premium IO (think Fusion-IO, memory-only, high-end SSD with good caching throughput), the difference would be much smaller or negligible.
A helpful comment from Shane Bester on a related bug report made me realize what was happening. Note the following in the MySQL Manual:
“Prior to MySQL 5.7.7, the default value of
sync_binlogwas 0, which configures no synchronizing to disk—in this case, the server relies on the operating system to flush the binary log’s contents from time to time as for any other file. MySQL 5.7.7 and later use a default value of 1, which is the safest choice, but as noted above can impact performance.” — https://dev.mysql.com/doc/refman/5.7/en/replication-options-binary-log.html#sysvar_sync_binlog
The culprit is thus the --sync_binlog=1 change which was made in 5.7.7 (in 5.6 it is 0 by default). While this may indeed be “the safest choice,” one has to wonder why Oracle chose to implement this default change in 5.7.7. After all, there are many other options which aid crash safety.
A related blog post from the MySQL HA team states;
“Indeed, [with sync_binlog=1,] it increases the total number of fsyncs called, but since MySQL 5.6, the server groups transactions and fsync’s them together, which minimizes greatly a potential performance hit.” — http://mysqlhighavailability.com/replication-defaults-in-mysql-5-7-7/ (ref item #4)
This seems incorrect given our findings, unless perhaps it requires tuning some other option.
This raises some actions points/questions for Oracle’s team: why change this now? Was 5.6 never crash-safe in terms of binary logging? How about other options that aid crash safety? Is anything [before 5.7.7] really ACID compliant by default?
In 2009 my colleague Peter Zaitsev had already posted on performance matters in connection with sync_binlog issues. More than seven years later, the questions asked in his post may still be valid today;
“May be opening binlog with O_DSYNC flag if sync_binlog=1 instead of using fsync will help? Or may be binlog pre-allocation would be good solution.” — PZ
Testing the same setup again, but this time with sync_binlog=0 and sync_binlog=1 synchronized/setup on both servers, we see;
Results for sync_binlog=0:
5.6.30: transactions: 7472 (149.38 per sec.)
5.7.12: transactions: 6594 (131.86 per sec.) — A 11.73% decrease
Results for sync_binlog=1:
5.6.30: transactions: 3854 (77.03 per sec.)
5.7.12: transactions: 4597 (91.89 per sec.) — A 19.29% increase
Note: the increase here is to some extent negated by the fact that enabling sync_binlog is overall still causes a significant (30% on 5.7 and 48% on 5.6) performance drop. Also interesting is that this could be the effect of “tuning the defaults” of/in 5.7, and it also makes one think about the possibility o further defaults tuning/optimization in this area.
Results for sync_binlog=100:
5.6.30: transactions: 7564 (151.12 per sec.)
5.7.12: transactions: 6515 (130.22 per sec.) — A 13.83% decrease
Thus, while 5.7.12 made some improvements when it comes to --sync_binlog=1, when --sync_binlog is turned off or is set to 100, we still see a ~11% decrease in performance. This is the same when not using binary logging at all, as a test with only --no-defaults (i.e. 100% vanilla out-of-the-box MySQL 5.6.30 versus MySQL 5.7.12) shows;
Results without binlogs enabled:
5.6.30: transactions: 7891 (157.77 per sec.)
5.7.12: transactions: 6963 (139.22 per sec.) — A 11.76% decrease
This raises another question for Oracle’s team: with four threads, there is a ~11% decrease in performance for 5.7.12 versus 5.6.30 (both vanilla)?
Discussing this internally, we were interested to see whether the arbitrary low number of four threads skewed the results and perhaps only showed a less realistic use case. However, testing with more threads, the numbers became worse still:
Results with 100 threads:
5.6.30. transactions: 20216 (398.89 per sec.)
5.7.12. transactions: 11097 (218.43 per sec.) — A 45.24% decrease
Results with 150 threads:
5.6.30. transactions: 11852 (233.01 per sec.)
5.7.12. transactions: 6606 (129.80 per sec.) — A 44.29% decrease
The findings in this article were compiled from a group effort.