P106-90 PG-Strom

3275 단어
자세히 연구하지 않고 대충 도망갔어요...PGSQL의 conf도 최적화되지 않고 플러그인을 불러오기만 했습니다
이 결과는...P106-90 3G보다 G3260???1+1<2 ?????
 
또한 CPU 1개에 100% GPU 1개에 80% 다른 1개에 30%-40%...왜 그랬을까?
 
explain analyze  SELECT 
 a.*
FROM 
  public.bmsql_oorder a, public.bmsql_oorder b where a. o_c_id<>b.o_c_id;


"Custom Scan (GpuJoin) on bmsql_oorder a  (cost=9011757.37..9011757.37 rows=899700000 width=36) (actual time=2213.061..191736.416 rows=899700000 loops=1)"
"  Outer Scan: bmsql_oorder a  (cost=0.00..542.00 rows=30000 width=36) (actual time=1.181..1.181 rows=603792 loops=1)"
"  Depth 1: GpuNestLoop  (plan nrows: 30000...899700000, actual nrows: 603792...919287067)"
"           JoinQuals: (a.o_c_id <> b.o_c_id)"
"           KDS-Heap (size plan: 2226.64KB, exec: 1289.14KB)"
"  GPU Preference: GPU0 (P106-090)"
"  ->  Seq Scan on bmsql_oorder b  (cost=0.00..542.00 rows=30000 width=4) (actual time=0.013..3.063 rows=30000 loops=1)"
"Planning time: 0.478 ms"
"Execution time: 216690.748 ms"

"Nested Loop  (cost=0.00..13501159.00 rows=899700000 width=36) (actual time=0.018..125453.434 rows=899700000 loops=1)"
"  Join Filter: (a.o_c_id <> b.o_c_id)"
"  Rows Removed by Join Filter: 300000"
"  ->  Seq Scan on bmsql_oorder a  (cost=0.00..542.00 rows=30000 width=36) (actual time=0.010..6.958 rows=30000 loops=1)"
"  ->  Materialize  (cost=0.00..692.00 rows=30000 width=4) (actual time=0.000..1.074 rows=30000 loops=30000)"
"        ->  Seq Scan on bmsql_oorder b  (cost=0.00..542.00 rows=30000 width=4) (actual time=0.003..3.702 rows=30000 loops=1)"
"Planning time: 0.947 ms"
"Execution time: 148207.334 ms"


"Custom Scan (GpuJoin) on bmsql_oorder a  (cost=9011757.37..9011757.37 rows=899700000 width=36) (actual time=2204.191..169322.990 rows=899700000 loops=1)"
"  Outer Scan: bmsql_oorder a  (cost=0.00..542.00 rows=30000 width=36) (actual time=1.199..1.199 rows=633792 loops=1)"
"  Depth 1: GpuNestLoop  (plan nrows: 30000...899700000, actual nrows: 633792...1816994118)"
"           JoinQuals: (a.o_c_id <> b.o_c_id)"
"           KDS-Heap (size plan: 2226.64KB, exec: 1289.14KB)"
"  GPU Preference: GPU0 (P106-090)"
"  ->  Seq Scan on bmsql_oorder b  (cost=0.00..542.00 rows=30000 width=4) (actual time=0.039..5.516 rows=30000 loops=1)"
"Planning time: 10.279 ms"
"Execution time: 192661.638 ms"

"Nested Loop  (cost=0.00..13501159.00 rows=899700000 width=36) (actual time=0.020..124936.745 rows=899700000 loops=1)"
"  Join Filter: (a.o_c_id <> b.o_c_id)"
"  Rows Removed by Join Filter: 300000"
"  ->  Seq Scan on bmsql_oorder a  (cost=0.00..542.00 rows=30000 width=36) (actual time=0.012..6.534 rows=30000 loops=1)"
"  ->  Materialize  (cost=0.00..692.00 rows=30000 width=4) (actual time=0.000..1.051 rows=30000 loops=30000)"
"        ->  Seq Scan on bmsql_oorder b  (cost=0.00..542.00 rows=30000 width=4) (actual time=0.002..3.823 rows=30000 loops=1)"
"Planning time: 0.699 ms"
"Execution time: 147828.463 ms"

 

좋은 웹페이지 즐겨찾기