视频1 视频21 视频41 视频61 视频文章1 视频文章21 视频文章41 视频文章61 推荐1 推荐3 推荐5 推荐7 推荐9 推荐11 推荐13 推荐15 推荐17 推荐19 推荐21 推荐23 推荐25 推荐27 推荐29 推荐31 推荐33 推荐35 推荐37 推荐39 推荐41 推荐43 推荐45 推荐47 推荐49 关键词1 关键词101 关键词201 关键词301 关键词401 关键词501 关键词601 关键词701 关键词801 关键词901 关键词1001 关键词1101 关键词1201 关键词1301 关键词1401 关键词1501 关键词1601 关键词1701 关键词1801 关键词1901 视频扩展1 视频扩展6 视频扩展11 视频扩展16 文章1 文章201 文章401 文章601 文章801 文章1001 资讯1 资讯501 资讯1001 资讯1501 标签1 标签501 标签1001 关键词1 关键词501 关键词1001 关键词1501 专题2001
PriorityqueueoptimizationforfilesortisnowvisibleinM_MySQL
2020-11-09 19:34:02 责编:小采
文档


TL;DR:Priority queue optimization for filesort with small LIMIT is now visible in MariaDB: there is a status variable and you can also see it in the slow query log (KB page link).

A longer variant:

One of the new optimizations in MySQL 5.6 is ability to use a priority queue instead of sorting for ORDER BY … LIMIT queries. The optimization was ported into MariaDB 10.0 long ago, but we still get questions if/when it will be ported. I guess, the reason for this is that, besides the query speed, you can’t see this optimization. Neither EXPLAIN, nor EXPLAIN FORMAT=JSON or PERFORMANCE_SCHEMA or status variables give any indication whether filesort used priority queue or the regular quicksort+merge algorithm.

MySQL 5.6 has only one way one can check whether filesort used priority queue. You need to enable optimizer_trace (set optimizer_trace=1), and then run the query (not EXPLAIN, but the query itself). Then, you can look into the optimizer trace and find something like this:

..."filesort_priority_queue_optimization": {"limit": 10,"rows_estimate": 198717,"row_size": 215,"memory_available": 262144,"chosen": true},...

MariaDB doesn’t support optimizer_trace at the moment. Even if it did, I think it would be wrong to require one to look into the optimizer trace to find out about the picked query plan.

The natural place to show the optimization would be EXPLAIN output. We could show something like“Using filesort (priority queue)”. This was my initial intent. After looking into the source code, this turned out to be difficult to do. The logic that makes the choice between using quicksort+merge and using priority queue is buried deep inside query execution code. (As if the mess caused by late optimizations of ORDER BY and UNIONs didn’t teach anybody in MySQL team anything).

As for query execution, there are two facilities where one could record execution-time details about the query plan. They are the status variables and the slow query log.

Status variables

We’ve addedSort_priority_queue_sortsstatus variable. Now, the list of sort-related status variables is:

MariaDB [test]> show status like 'Sort%';+---------------------------+-------+| Variable_name | Value |+---------------------------+-------+| Sort_merge_passes | 0 |

下载本文
显示全文
专题