Mariadb 10.1

Optimistic mode of in-order parallel replication
有序并行复制的乐观模式
尝试并行应用大多数的事务性DML，并处理（回滚、重试）任何发生的冲突。

Enhanced semisync replication; Wait for at least one slave to acknowledge transaction before committing
增强的半同步复制。
直到至少一个从属服务器确认接收到该事务，该事务才变得可见。这意味着，如果主服务器完全丢失，其他连接看到的任何事务都将复制到某个地方，从而避免了潜在的幻像。
Mariadb10.3.3之前以插件形式实现。

Triggers can now be run on the slave for row-based events.
在基于行的复制下，可以在从库上运行主库上不存在的触发器。

Dump Thread Enhancements from Google. Makes multiple slave setups faster by allowing concurrent reading of binary log.
读取binlog事件时，dump thread不会彼此阻塞。
dump thread不会被正在Write Binlog事件的用户会话阻止。
用户会话不受正在读取binlog事件的dump thread阻止。
dump thread和user session都可以获得更好的吞吐量，并且改进非常明显，尤其是在有许多从库服务器的情况下。

Commits in certain instances in parallel replication complete immediately, avoiding losing throughput when many transactions need conflicting locks
如果服务器检测到提交事务T1中的一个持有另一个事务T2正在等待的 InnoDB行锁，则提交将立即完成，而不会进一步延迟。当许多事务需要冲突的锁时，这有助于避免丢失吞吐量。

Doesn’t make stupid choices when several multi-part keys and potential range accesses are present.
优化多字段索引的相关排序问题。例如：select * from tb_page where page_id=0 and page_name>='Sg' order by page_name limit 10;
index(page_id,page_name)

Always uses “range” and (not full “index” scan) when it switches to an index to satisfy ORDER BY … LIMIT.
优化多字段索引的相关排序问题。例如：select * from t where col1<=1412199999 order by col1 desc,col2 desc,col3 desc limit 10;
index(col1,col2,col3)

Tries hard to be smart and use cost/number of records estimates from other parts of the optimizer.
优化单表多个索引，优化器选择最合适的索引，而不是目前的最先创建的索引。
例如：select * from tb_test1 where fd1=1 order by fd_pk limit 1000;
tb_test1.ix_fd_fdpk(fd1,fd_pk),tb_test1.ix_fd1_fd2(fd1,fd2)。
select * from tb_test2 where fd1=1 order by fd_pk limit 1000;
tb_test2.ix_fd1_fd2(fd1,fd2),tb_test2.ix_fd_fdpk(fd1,fd_pk)。

Takes full advantage of InnoDB’s Extended Keys feature when checking if filesort() can be skipped
利用InnoDB的存储特点来优化索引排序。
例如：SELECT pk2 FROM tb_bug1 USE INDEX(key1) WHERE pk1 = 123 AND col1 = 'video' ORDER BY pk2 DESC LIMIT 10;
CREATE TABLE tb_bug1 (
pk1 int(11) NOT NULL,
pk2 varchar(64) NOT NULL,
col1 varchar(16) DEFAULT NULL,
PRIMARY KEY (pk1,pk2),
KEY key1 (pk1,col1)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;

The ORDER BY optimizer takes multiple-equalities into account
排序优化需要考虑等式传播。
例如：select * from a join b on a.id=b.id order by a.id desc limit 10;
select * from a join b on a.id=b.id order by b.id desc limit 10;

Make switching of index due to small limit cost-based: We have made the decision in make_join_select() of whether to switch to a new index in order to support "ORDER BY ... LIMIT N" cost-based.
order by limit小批量数据时，基于成本考虑索引的切换。
例如：select * from t1 where a LIKE '0.5555%' AND b >= now()-interval 100 day ORDER BY b DESC LIMIT 100;
index(a),index(b),a的范围筛选值更少，order by b导致使用了b的索引，而不是基于成本考虑的a的索引。

MAX_STATEMENT_TIME can be used to automatically abort long running queries.
终止超过一定时间（MAX_STATEMENT_TIME，单位为秒，默认0，即无限制）的sql语句（所有sql，不包括存储过程）。

UNION ALL works without usage of a temporary table .
没有order by 子句的union all 可以不发送到临时表，行可以直接发送到客户端。

Consistent support for IF EXISTS, IF NOT EXISTS, and OR REPLACE clauses.
语法支持：database\function\role\server\user\view\event\index\trigger

ANALYZE statement provides output that looks like EXPLAIN output, but also includes data from the query execution (how many rows were actually read, etc).
新的解析支持，ANALYZE explainable_statement,调用查询优化器，查询实际影响行数（真实执行查询）。

EXPLAIN FORMAT=JSON is a re-implementation of similar feature in MySQL 5.6
ANALYZE FORMAT=JSON produces detailed information about the statement execution
以json形式输出相关解析。

Using Compound Statements Outside of Stored Programs
语法支持：可以在存储过程之外使用BEGIN,IF, CASE, LOOP, WHILE, REPEAT 

The number of rows affected by a slow UPDATE or DELETE is now recorded in the slow query log - see also mysql.slow_log Table.
update\delete的慢SQL语句实际影响行数也可以记录在mysql.slow_log了,之前rows为0。

Allow up to 64K pages in InnoDB (old limit was 16K)
InnoDB内部page的容量限制由16K改为64K。支持多个blob等大数据类型的字段，减少分page存储。

The Facebook/Kakao defragmentation patch (see Defragmenting InnoDB Tablespaces) which uses OPTIMIZE TABLE to defragment InnoDB tablespaces).
对InnoDB表空间进行碎片整理，没有创建新表，也不需要将数据从旧表复制到新表。取而代之的是，此功能加载n页面（由innodb-defragment-n-pages确定）并尝试移动记录，以便页面将充满记录，然后在操作后释放完全为空的页面。

New server variable enforce_storage_engine
新的系统参数enforce_storage_engine用于强制对新表使用特定的存储引擎。用于避免使用其他引擎不必要地创建表。

Default size of query_alloc_block_size changed from 8192 to 16384 and query_prealloc_size from 8192 to 24576 to avoid the need for simple queries with one join to call my_malloc
优化query_alloc_block_size、query_prealloc_size参数的默认值，避免简单查询调用my_malloc

SET STATEMENT - set variables for the duration of the query (MDEV-5231). This is a backport of Per-query Variable Statement feature of Percona Server 5.6 (which, in turn, is based in MySQL GSoC 2009 project by Joseph Lukas), with many bugs fixed.
对单个查询设置系统参数的值，执行完查询后将还原之前的值。语法：SET STATEMENT <variable=value> FOR <statement>;

数据加密

InnoDB页面压缩

密码验证插件