结论是存储在 remote 端，这其实也很好理解。在远端能高效率地计算，当然应当在远端完成缓存和检索。 ■ Before query execution via DBLINK: =========== DBLINK destination (remote side): conn u2 / u2 create table tab002 as select * from dba_objects; shu immediate startup select o.owner, o.object_name, o.object_type, s.buffer_pool, count (*) buffer from dba_objects o, v $ bh b, dba_segments s where b.objd = o.data_object_id and o.owner = 'U 2' and o.object_name = s.segment_name and o.object_type = s.segment_type group by o.owner, o.object_name, o.object_type, o.status, s.buffer_pool order by buffers; A record was not selected. =========== Local side: conn u3 / u3 select o.owner, o.object_name, o.object_type, s.buffer_pool, count (*) buffer from dba_objects o, v $ bh b, dba_segments s where b.objd = o.data_object_id and o.owner = 'U 2' and o.object_name = s.segment_name and o.object_type = s.segment_type group by o.owner, o.object_name, o.object_type, o.status, s.buffer_pool order by buffers; A record was not selected. select o.owner, o.object_name, o.object_type, s.buffer_pool, count (*) buffer from dba_objects o, v $ bh b, dba_segments s where b.objd = o.data_object_id and o.owner = 'U3' and o.object_name = s.segment_name and o.object_type = s.segment_type group by o.owner, o.object_name, o.object_type, o.status, s.buffer_pool order by buffers; A record was not selected. ■ Query execution via DBLINK: SQL> select count (*) from tab002 @ dbk; COUNT (*) ---------- 69263 conn u3 / u3 select o.owner, o.object_name, o.object_type, s.buffer_pool, count (*) buffer from dba_objects o, v $ bh b, dba_segments s where b.objd = o.data_object_id and o.owner = 'U 2' and o.object_name = s.segment_name and o.object_type = s.segment_type group by o.owner, o.object_name, o.object_type, o.status, s.buffer_pool order by buffers; A record was not selected. <<<<<<<<<<< Cache has no related information select o.owner, o.object_name, o.object_type, s.buffer_pool, count (*) buffer from dba_objects o, v $ bh b, dba_segments s where b.objd = o.data_object_id and o.owner = 'U3' and o.object_name = s.segment_name and o.object_type = s.segment_type group by o.owner, o.object_name, o.object_type, o.status, s.buffer_pool order by buffers; A record was not selected. <<<<<<<<<<< Cache has no related information =========== DBLINK destination (remote side): select o.owner, o.object_name, o.object_type, s.buffer_pool, count (*) buffer from dba_objects o, v $ bh b, dba_segments s where b.objd = o.data_object_id and o.owner = 'U 2' and o.object_name = s.segment_name and o.object_type = s.segment_type group by o.owner, o.object_name, o.object_type, o.status, s.buffer_pool order by buffers; OWNER -------------------------------------------------------------------- ------------------------------ OBJECT_NAME -------------------------------------------------------------------- ------------------------------ OBJECT_TYPE BUFFER_POOL -------------------------------------------------------------------- ------- --------------------- BUFFERS ---------- U2 TAB 002 TABLE DEFAULT 2042 <<<<<<<<<<<<<<<<<< data got into the cache. select o.owner, o.object_name, o.object_type, s.buffer_pool, count (*) buffer from dba_objects o, v $ bh b, dba_segments s where b.objd = o.data_object_id and o.owner = 'U3' and o.object_name = s.segment_name and o.object_type = s.segment_type group by o.owner, o.object_name, o.object_type, o.status, s.buffer_pool order by buffers; A record was not selected.
本文转自健哥的数据花园博客园博客，原文链接：http://www.cnblogs.com/gaojian/p/7575226.html，如需转载请自行联系原作者