通用场景下tikv的copTask在内存中扫描数据太慢,比mysql、pg等都慢

mysql> explain analyze select sum(L_EXTENDEDPRICE) from lineitem;
+----------------------------+------------+---------+-----------+----------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+------------------------------------------------------+-----------+------+
| id                         | estRows    | actRows | task      | access object  | execution info                                                                                                                                                                                                                                                                                                                                                    | operator info                                        | memory    | disk |
+----------------------------+------------+---------+-----------+----------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+------------------------------------------------------+-----------+------+
| HashAgg_11                 | 1.00       | 1       | root      |                | time:1.37s, loops:2, partial_worker:{wall_time:1.366558591s, concurrency:5, task_num:1, tot_wait:6.832190204s, tot_exec:31.755µs, tot_time:6.832224842s, max:1.366448056s, p95:1.366448056s}, final_worker:{wall_time:0s, concurrency:5, task_num:1, tot_wait:6.832286057s, tot_exec:16.792µs, tot_time:6.832305524s, max:1.366465061s, p95:1.366465061s}         | funcs:sum(Column#18)->Column#17                      | 20.2 KB   | N/A  |
| └─TableReader_12           | 1.00       | 13      | root      |                | time:1.37s, loops:2, cop_task: {num: 13, max: 1.37s, min: 708.7ms, avg: 998.8ms, p95: 1.37s, max_proc_keys: 467326, p95_proc_keys: 467326, tot_proc: 12.9s, tot_wait: 55ms, rpc_num: 13, rpc_time: 13s, copr_cache_hit_ratio: 0.00, distsql_concurrency: 15}                                                                                                      | data:HashAgg_5                                       | 436 Bytes | N/A  |
|   └─HashAgg_5              | 1.00       | 13      | cop[tikv] |                | tikv_task:{proc max:1.36s, min:687ms, avg: 985.2ms, p80:1.07s, p95:1.36s, iters:5869, tasks:13}, scan_detail: {total_process_keys: 6001215, total_process_keys_size: 1166860203, total_keys: 6001228, get_snapshot_time: 59.8ms, rocksdb: {key_skipped_count: 6001215, block: {cache_hit_count: 43, read_count: 18391, read_byte: 554.2 MB, read_time: 877.6ms}}} | funcs:sum(tpch1.lineitem.l_extendedprice)->Column#18 | N/A       | N/A  |
|     └─TableFullScan_10     | 6622334.00 | 6001215 | cop[tikv] | table:lineitem | tikv_task:{proc max:1.28s, min:630ms, avg: 924.2ms, p80:1.02s, p95:1.28s, iters:5869, tasks:13}                                                                                                                                                                                                                                                                   | keep order:false                                     | N/A       | N/A  |
+----------------------------+------------+---------+-----------+----------------+-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+------------------------------------------------------+-----------+------+
4 rows in set (1.37 sec)

block: {cache_hit_count: 43, read_count: 18391, read_byte: 554.2 MB, read_time: 877.6ms} 这种带有read_count才是发生物理读吧,我帖子中的那个只有逻辑读。
另外这种慢的情况并不是特殊情况,是普遍情况,所以你随便自己做个环境测试就能测试出来。