dblink与fdw测试 索引区别

以下为使用fdw和dblink精简对比：（具体fdw建立外部服务器等操作不列了，官方文档很详细）
A：explain analyze SELECT t.*
FROM v_wpath_route_new_lv_test t where uuid=‘0ad240d1ca0a48b5b40a24a42ee5d875’

B：explain analyze SELECT t.*
FROM dblink(‘hostaddr=******, ‘select uuid,’‘0’’ as stateflag,wpath_name,wpath_uuid,layer,ems_name,direction,path_no,route_no,link_type,a_device,z_device,a_port,z_port from v_wpath_route_new_lv_test’::text) t(uuid character varying, stateflag integer, wpath_name character varying, wpath_uuid character varying, layer character varying, ems_name character varying, direction character varying, path_no integer, route_no integer, link_type character varying, a_device character varying, z_device character varying, a_port character varying, z_port character varying)
where t.uuid=‘0ad240d1ca0a48b5b40a24a42ee5d875’

其中有个条件uuid=‘0ad240d1ca0a48b5b40a24a42ee5d875’，uuid在远程数据库为索引字段，使用该索引的话查询会快很多；A中v_wpath_route_new_lv_test为远程数据库的v_wpath_route_new_lv_test的外部表（参考B）
A方法外部表是走索引的

当用dblink时，若是把条件放在dblink外面（如B），则查询是在本地数据库执行，看查询计划不走索引，猜测因为dblink只同步了数据没同步索引
但是若把条件放在dblink里面的话，是在远程数据库执行，则快很多，比如：B2：explain analyze SELECT t.*
FROM dblink(‘hostaddr=******, ‘select uuid,’‘0’’ as stateflag,wpath_name,wpath_uuid,layer,ems_name,direction,path_no,route_no,link_type,a_device,z_device,a_port,z_port from v_wpath_route_new_lv_test where t.uuid=‘0ad240d1ca0a48b5b40a24a42ee5d875’'::text) t(uuid character varying, stateflag integer, wpath_name character varying, wpath_uuid character varying, layer character varying, ems_name character varying, direction character varying, path_no integer, route_no integer, link_type character varying, a_device character varying, z_device character varying, a_port character varying, z_port character varying)
因为远程数据库有索引走索引，而外部表则是在远程数据库执行，理解是把所有相关sql放在远程数据库执行就走索引，本人做了试验将相同情况的A和B2对比发现B2更快
但是某些情况下查询条件是灵活的，不能放在dblink里面，也不能在远程数据库端执行走索引，这种情况下就适合用fdw外部表