PostGIS数据测试-一百万点要素

小小测试一下，看看单表百万数据的情况

服务器配置

系统版本：Centos7.9.2009
CPU：两颗Intel® Xeon® Gold 6226R CPU @ 2.90GHz处理器，共32核心
内存：DDR4 256G
硬盘：Raid5 共24T
数据库版本：PostgreSQL13+PostGIS3.0

创建表

不添加GIS索引

create table geo_point_one_million_without_index
(
    uid  uuid default uuid_generate_v4() not null
        constraint geo_point_one_million_without_index_pk
            primary key,
    geom geometry
);

comment on table geo_point_one_million_without_index is '点测试(一百万_没有geo索引)表';

comment on column geo_point_one_million_without_index.geom is '地理坐标';

alter table geo_point_one_million_without_index
    owner to postgres;

添加GIS索引

create table geo_point_one_million
(
    uid  uuid default uuid_generate_v4() not null
        constraint geo_point_one_million_pk
            primary key,
    geom geometry
);

comment on table geo_point_one_million is '点测试(一百万)表';

comment on column geo_point_one_million.geom is '地理坐标';

alter table geo_point_one_million
    owner to postgres;

create index idx_geom_geo_point_one_million
    on geo_point_one_million using gist (geom);

ST_GeneratePoints

这里我们使用PostGIS的ST_GeneratePoints函数来创建随机点。

ST_GeneratePoints官方介绍

格式

geometry ST_GeneratePoints( g geometry , npoints integer );

geometry ST_GeneratePoints( geometry g , integer npoints , integer seed );

描述

ST_GeneratePoints生成位于输入区域内的给定数量的伪随机点。可选的 seed 用于重新生成确定性的点序列，并且必须大于零。

可用性：2.3.0

增强：3.0.0，添加种子参数

示例

使用随机种子值1996生成了覆盖在原始面顶部的12个点

SELECT ST_GeneratePoints(geom, 12, 1996)
FROM (
    SELECT ST_Buffer(
        ST_GeomFromText(
        'LINESTRING(50 50,150 150,150 50)'),
        10, 'endcap=round join=round') AS geom
) AS s;

生成数据

不添加索引的数据

以江苏省范围为边界，随机生成一百一十万个点

insert into geo_point_one_million_without_index (geom)
SELECT (
           ST_Dump(
                   ST_GeneratePoints(bound.geom, 1100000)
               )).geom AS geom
FROM (
         select geom
         from province
     ) bound;

查看数据库总数如下：

加载到地图上如下所示，

添加索引的数据

以江苏省范围为边界，随机生成一百一十万个点

insert into geo_point_one_million (geom)
SELECT (
           ST_Dump(
                   ST_GeneratePoints(bound.geom, 1100000)
               )).geom AS geom
FROM (
         select geom
         from province
     ) bound;

效果如上图所示

数据查询效率比较

使用统一范围进行查询，查询范围内的点数据

查询没有索引的表

select geom
from geo_point_one_million_without_index
where ST_Intersects(geom,
                    st_transform(st_setsrid(st_geomfromewkt(
                                                    'POLYGON ((119.72706893295168 33.459312148850884, 119.72395052277068 33.44938433824096, 119.72271659870063 33.4495696575886, 119.7178056612509 33.424791447676355, 119.74576996445631 33.42143309794773, 119.74810254818226 33.42089897286719, 119.75342279140625 33.44529968061062, 119.75459464757762 33.45337738499248, 119.72706893295168 33.459312148850884))'),
                                            4326)
                        , 4490));

如图所示：

经过多次测试，查询时间分布在460ms~550ms左右

查询有索引的表

select geom
from geo_point_one_million
where ST_Intersects(geom,
                    st_transform(st_setsrid(st_geomfromewkt(
                                                    'POLYGON ((119.72706893295168 33.459312148850884, 119.72395052277068 33.44938433824096, 119.72271659870063 33.4495696575886, 119.7178056612509 33.424791447676355, 119.74576996445631 33.42143309794773, 119.74810254818226 33.42089897286719, 119.75342279140625 33.44529968061062, 119.75459464757762 33.45337738499248, 119.72706893295168 33.459312148850884))'),
                                            4326)
                        , 4490));