The problem with many single-statement parallel updates/inserts is that each statement creates an individual implicit transaction, and this transaction will stuck on commit, waiting for its changes to be persistent, i.e written into Innodb redo log (by default also followed with fsync). There is a group commit feature n Innodb, which aims to fix that, but I’m not too familiar with it, not sure how efficient it is. Innodb_flush_log_at_trx_commit=2 mentioned by Guillaume can do wonders, because it omits the slowest part of commit, the fsync.
On the other hand, large multivalued updates , or transactions will do a single write on commit for many updates, instead of multiple tiny writes. That’s the whole trick, which for example makes mysqldump-made logical backups fast to restore , without doing any work in parallel on the client side (Alright, mysqldump has couple more tricks, such as disabling indexes, but multivalued updates are essential). You can experiment with the number of connections, and check whether increasing them brings much more throughput, but my guess more than say number_of_cores *2 would not give improvement (once you start combining the updates). Definitely not thousands of connections. Perhaps a single client connection will suffice. From: JCA Sent: Monday, 7 October 2019 20:52 To: Vladislav Vaintroub Cc: [email protected] Subject: Re: [Maria-discuss] Performance tuning sought for MariaDB On Mon, Oct 7, 2019 at 11:06 AM Vladislav Vaintroub <[email protected]> wrote: You do not share many details how exactly your application exactly interacts with the server. Do you work with large batches, I.e generate big (say 1MB) multi-valued statements like INSERT INTO t(a,b) VALUES(a1,b1),(a2,b2)......,(aN, bN) ON DUPLICATE KEY UPDATE counter=counter+1 It will be mostly individual insertions, amounting to a small amount of data every time. The mass-deletion is straightforward DELETE FROM t WHERE id in (id1,.......idN) This I do every so often, but far less frequently than insertions and modifications. Make sure your multi-valued inserts/deletes do not exceed the value of “max_allowed_packet” session variable. And work preferably with a single connection or small amount of connections. Thanks. I do the latter, actually -- if I have have several insertions/modifications to do, I strive to submit them in a single connection. I do have a separate connection per thread though. Do you think it might be worth the while using a single connection, or a small connection pool, throughout? If this sounds complicated, you can combine multiple updates in large transactions instead, though this could be slightly less efficient, since there is more interaction between the application and DB. That, in fact, has been my observation. Rather than parallelizing single updates, it is usually better to combine updates in large-ish transactions. Most of the update-related work will happen in background anyway, at least for innodb. There is some info in the documentation that mentions multi-value inserts https://mariadb.com/kb/en/library/how-to-quickly-insert-data-into-mariadb/ Thanks; I'll check that out. From: JCA Sent: Monday, 7 October 2019 18:17 To: [email protected] Subject: [Maria-discuss] Performance tuning sought for MariaDB I am running MariaDB 10.0.34 on a Slackware 14.2 system. I have a C application that interacts with MariaDB in the following way: 1. Read data from a FIFO. 2. Insert the data into a table in a MariaDB database, if absent, or modify it according to certain specific criteria otherwise. This works as expected. The problem that I have is that data are being written to the FIFO at a fast rate. In order to be able to keep up, at any given time my application reads the data available at the FIFO, and spawns a thread to process the chunk of data just read. It is in this thread that all the database interaction takes place. In order to deal with this, I have the following entries in my /etc/my.cnf file: # this is read by the standalone daemon and embedded servers [server] # this is only for the mysqld standalone daemon [mysqld] # thread_handling=pool-of-threads # log=/var/log/mysqld.log # this is only for embedded server [embedded] # This group is only read by MariaDB servers, not by MySQL. # If you use the same .cnf file for MySQL and MariaDB, # you can put MariaDB-only options here [mariadb] # log=/var/log/mysqld.log general_log_file = /var/lib/mysql/mysql.log # general_log = 1 # transaction-isolation = READ-COMMITTED # key_buffer = 1280M # 128MB for every 1GB of RAM # sort_buffer_size = 1M # 1MB for every 1GB of RAM # read_buffer_size = 1M # 1MB for every 1GB of RAM # read_rnd_buffer_size = 1M # 1MB for every 1GB of RAM # thread_concurrency = 24 # Based on the number of CPUs # so make it CPU*2 # thread-handling=pool-of-threads # innodb_flush_log_at_trx_commit != 1 # open_files_limit = 50000 thread-handling=pool-of-threads max_connections = 1000 table_open_cache = 800 query_cache_type = 0 innodb_buffer_pool_size = 512M innodb_buffer_pool_instances = 10 innodb_adaptive_hash_index_partitions = 20 innodb_lock_wait_timeout = 5000 With this, my application can keep up with the FIFO writer, but - depending on the circumstances - my database can't. As I am writing this, there are over 1300 threads connected to my database; any command that I issue at the mysql CLI takes over one minute to return. I am keeping track on how long each thread takes to complete, and that is of the order of hundreds of seconds - sometimes thousands. Each thread is itself simple, in that it just issues a couple of simple MariaDB commands. Currently my table consists of 1.6 million entries, and growing - on this basis, I expect that things will get only worse. Each entry,however, will never require more than a couple of hundred bytes of storage. The operations that can be undertaken on entries are insertion, deletion and modification, the latter being straightforward - like e.g. incrementing a counter or replacing a short string. My system has 24 GB of RAM and 12 cores. Occasionally all the cores are fully busy with MariaDB activity, but most of the time barely one or two are. I am a newbie when it comes to interacting with MariaDB - please, bear with me. I know I must use a single database and a single table. I also know - because of the nature of the data that are being written to the FIFO - that the probability for two different threads to be operating on the same entry in the table at the same time is negligible - i.e. for all practical purposes, that will not happen. What I need is advice on how to configure my instance of MariaDB to perform optimally in the scenario above. In particular, I would like for it to make better use of all the cores available - in essence, to parallelize the database operations as much as possible. Feedback from the experts will be much appreciated.
_______________________________________________ Mailing list: https://launchpad.net/~maria-discuss Post to : [email protected] Unsubscribe : https://launchpad.net/~maria-discuss More help : https://help.launchpad.net/ListHelp
