标签
PostgreSQL , Oracle
背景
Oracle业务适合用PG去O的一些评判标准：
功能指标
如果评估出来的业务中具备这些特性，非常适合使用PostgreSQL。
1、业务使用的数据类型中出现
IP地址、GIS、数组、范围、全文检索、大对象、字节流、比特流、枚举、几何、自定义复合、UUID、XML、JSON、货币、字符串、数值、时间、加密数据类型
2、业务需求中出现
全文检索、模糊查询、相似查询
3、业务使用的SQL中出现
connect by、多维分析(grouping, grouping sets, rollup, cube)、多表JOIN、窗口查询(over partition by ())、聚合函数
4、业务使用的SQL中出现如下HINT
parallel
hash hint
left join
right join
outer join
merge join
hash agg
group agg
merge sort
skip scan
5、业务使用了存储过程
6、表的数据量
单表过亿
7、业务使用了dblink，外部表功能
8、业务使用了bitmap\btree索引
PostgreSQL 内置多种索引接口(hash, btree, gin, gist, sp-gist, brin, bloom)
9、业务中使用了约束
primary key, unique key, check, not null, default value
10、业务中使用了全局序列、局部序列
sequence
11、业务使用了翻转索引
12、业务使用了分区表
13、业务使用了触发器、规则功能
14、业务使用了混合负载
小事务和分析型事务并存。
PostgreSQL通过多核并行、JIT、算子复用等技术，加速分析事务。
15、业务使用了upsert（不存在则插入，存在则更新）
16、业务大量使用了GIS地理位置数据
17、业务有大量数据透视需求（BI分析）
18、业务大量使用了ORACLE的内置函数，（分析函数、聚合函数、窗口函数、数据处理函数等）
19、业务有任意列，任意条件筛选需求
20、业务有倒排索引需求
21、业务有空间数据检索需求
22、业务使用了物化视图
23、业务有多master需求
24、业务有流式计算需求
25、业务有图式搜索需求
26、业务有读写分离需求
27、业务有并行查询的需求
28、业务有加密数据类型需求
29、业务有数据采样需求
30、业务有链路加密需求
性能指标
单机(32 CORE, SSD, 512GB内存)
1、TPC-H性能
SF=100，100GB 裸数据。
2017-07-13 20:04:29 [1499947469] : running queries defined in TPC-H benchmark
2017-07-13 20:04:29 [1499947469] : running query 1
2017-07-13 20:04:29 [1499947469] : run explain
2017-07-13 20:04:29 [1499947469] : run the query on background
2017-07-13 20:04:47 [1499947487] : query 1 finished OK (17 seconds)
2017-07-13 20:04:47 [1499947487] : running query 2
2017-07-13 20:04:47 [1499947487] : run explain
2017-07-13 20:04:47 [1499947487] : run the query on background
2017-07-13 20:08:13 [1499947693] : query 2 finished OK (206 seconds)
2017-07-13 20:08:13 [1499947693] : running query 3
2017-07-13 20:08:13 [1499947693] : run explain
2017-07-13 20:08:14 [1499947694] : run the query on background
2017-07-13 20:08:55 [1499947735] : query 3 finished OK (41 seconds)
2017-07-13 20:08:55 [1499947735] : running query 4
2017-07-13 20:08:55 [1499947735] : run explain
2017-07-13 20:08:55 [1499947735] : run the query on background
2017-07-13 20:09:02 [1499947742] : query 4 finished OK (6 seconds)
2017-07-13 20:09:02 [1499947742] : running query 5
2017-07-13 20:09:02 [1499947742] : run explain
2017-07-13 20:09:02 [1499947742] : run the query on background
2017-07-13 20:09:16 [1499947756] : query 5 finished OK (14 seconds)
2017-07-13 20:09:16 [1499947756] : running query 6
2017-07-13 20:09:16 [1499947756] : run explain
2017-07-13 20:09:16 [1499947756] : run the query on background
2017-07-13 20:09:21 [1499947761] : query 6 finished OK (4 seconds)
2017-07-13 20:09:21 [1499947761] : running query 7
2017-07-13 20:09:21 [1499947761] : run explain
2017-07-13 20:09:21 [1499947761] : run the query on background
2017-07-13 20:10:06 [1499947806] : query 7 finished OK (35 seconds)
2017-07-13 20:10:06 [1499947806] : running query 8
2017-07-13 20:10:06 [1499947806] : run explain
2017-07-13 20:10:06 [1499947806] : run the query on background
2017-07-13 20:10:38 [1499947838] : query 8 finished OK (31 seconds)
2017-07-13 20:10:38 [1499947838] : running query 9
2017-07-13 20:10:38 [1499947838] : run explain
2017-07-13 20:10:38 [1499947838] : run the query on background
2017-07-13 20:11:32 [1499947892] : query 9 finished OK (54 seconds)
2017-07-13 20:11:32 [1499947892] : running query 10
2017-07-13 20:11:32 [1499947892] : run explain
2017-07-13 20:11:32 [1499947892] : run the query on background
2017-07-13 20:11:49 [1499947909] : query 10 finished OK (16 seconds)
2017-07-13 20:11:49 [1499947909] : running query 11
2017-07-13 20:11:49 [1499947909] : run explain
2017-07-13 20:11:49 [1499947909] : run the query on background
2017-07-13 20:11:56 [1499947916] : query 11 finished OK (7 seconds)
2017-07-13 20:11:56 [1499947916] : running query 12
2017-07-13 20:11:56 [1499947916] : run explain
2017-07-13 20:11:56 [1499947916] : run the query on background
2017-07-13 20:13:37 [1499948017] : query 12 finished OK (100 seconds)
2017-07-13 20:13:37 [1499948017] : running query 13
2017-07-13 20:13:37 [1499948017] : run explain
2017-07-13 20:13:37 [1499948017] : run the query on background
2017-07-13 20:17:11 [1499948231] : query 13 finished OK (213 seconds)
2017-07-13 20:17:11 [1499948231] : running query 14
2017-07-13 20:17:11 [1499948231] : run explain
2017-07-13 20:17:11 [1499948231] : run the query on background
2017-07-13 20:17:15 [1499948235] : query 14 finished OK (4 seconds)
2017-07-13 20:17:15 [1499948235] : running query 15
2017-07-13 20:17:15 [1499948235] : run explain
2017-07-13 20:17:15 [1499948235] : run the query on background
2017-07-13 20:17:40 [1499948260] : query 15 finished OK (25 seconds)
2017-07-13 20:17:40 [1499948260] : running query 16
2017-07-13 20:17:40 [1499948260] : run explain
2017-07-13 20:17:40 [1499948260] : run the query on background
2017-07-13 20:18:41 [1499948321] : query 16 finished OK (60 seconds)
2017-07-13 20:18:41 [1499948321] : running query 17
2017-07-13 20:18:41 [1499948321] : run explain
2017-07-13 20:18:41 [1499948321] : run the query on background
2017-07-13 20:27:55 [1499948875] : query 17 finished OK (552 seconds)
2017-07-13 20:27:55 [1499948875] : running query 18
2017-07-13 20:27:55 [1499948875] : run explain
2017-07-13 20:27:55 [1499948875] : run the query on background
2017-07-13 20:49:57 [1499950197] : query 18 finished OK (1317 seconds)
2017-07-13 20:49:57 [1499950197] : running query 19
2017-07-13 20:49:57 [1499950197] : run explain
2017-07-13 20:49:57 [1499950197] : run the query on background
2017-07-13 20:50:09 [1499950209] : query 19 finished OK (11 seconds)
2017-07-13 20:50:09 [1499950209] : running query 20
2017-07-13 20:50:09 [1499950209] : run explain
2017-07-13 20:50:09 [1499950209] : run the query on background
2017-07-13 20:56:43 [1499950603] : query 20 finished OK (393 seconds)
2017-07-13 20:56:43 [1499950603] : running query 21
2017-07-13 20:56:43 [1499950603] : run explain
2017-07-13 20:56:43 [1499950603] : run the query on background
2017-07-13 20:58:19 [1499950699] : query 21 finished OK (95 seconds)
2017-07-13 20:58:19 [1499950699] : running query 22
2017-07-13 20:58:19 [1499950699] : run explain
2017-07-13 20:58:19 [1499950699] : run the query on background
2017-07-13 21:00:43 [1499950843] : query 22 finished OK (143 seconds)
2017-07-13 21:00:43 [1499950843] : finished TPC-H benchmark
2、TPC-C性能
3000仓库、256客户端。84.5万 tpmC。
3、GIS(KNN检索)
100亿位置信息，近邻查询。
tps: 7.4万/s
rt: 0.848毫秒
4、模糊查询
前后模糊(like '%????%')
1亿数据量，前后模糊，0.2毫秒。
5、全文检索
10亿随机值，返回2万条匹配记录，26毫秒。
《PostgreSQL 全文检索加速 快到没有朋友 - RUM索引接口(潘多拉魔盒)》
6、多表JOIN
2张1亿记录，10张1000万记录，1张1000记录的表进行JOIN，聚合查询。
23毫秒。
c 1000万
d 1000
e 1亿
postgres=# explain (analyze,verbose,timing,costs,buffers)
select count(t1.*) from
e t1 join e t2 on (t1.id=t2.id and t1.id<=1000)
join c t3 on (t1.id=t3.id)
join c t4 on (t1.id=t4.id)
join c t5 on (t1.id=t5.id)
join c t6 on (t1.id=t6.id)
join c t7 on (t1.id=t7.id)
join c t8 on (t1.id=t8.id)
join c t9 on (t1.id=t9.id)
join c t10 on (t1.id=t10.id)
join c t11 on (t1.id=t11.id)
join c t12 on (t1.id=t12.id)
join d t13 on (t1.id=t13.id) ;
Aggregate (cost=3234.08..3234.09 rows=1 width=8) (actual time=23.665..23.665 rows=1 loops=1)
Output: count(t1.*)
Buffers: shared hit=48059
-> Nested Loop (cost=5.76..3234.08 rows=1 width=28) (actual time=0.083..23.553 rows=1000 loops=1)
Output: t1.*
Join Filter: (t1.id = t13.id)
Buffers: shared hit=48059
............
Planning time: 7.943 ms
Execution time: 23.782 ms
(116 rows)
7、单表聚合性能
单表8亿记录，avg,count,sum,min,max维度聚合查询。
32个并行度
5.3秒
postgres=# select count(*),sum(id),avg(id),min(id),max(id) from e;
count | sum | avg | min | max
-----------+-------------------+-----------------------+-----+-----------
800000000 | 40000000400000000 | 50000000.500000000000 | 1 | 100000000
(1 row)
Time: 5316.490 ms (00:05.316)
8、数据导入速度
并行写入，500万条记录/s 或 每秒1.8GB/s。
《PostgreSQL 如何潇洒的处理每天上百TB的数据增量》
小结
简单来说，PostgreSQL是Oracle的最佳替代产品，而且还有额外惊喜，参考应用案例一文。
