背景
因为需要进行代码优化。所以进行数据表的整表COPY
一直很好奇,多进程对于copy是否有优化呢?于是做了一些实验。
实验环境:32核I5的服务器。内存200G
实验一: 两个进程 分开copy两个表
Python代码如下
dsn = 'postgresql://postgres:pset123456@192.168.10.10/CHN_NAVINFO_2016Spr_0082_0002_108'
conn1 = psycopg2.connect(dsn=dsn)
conn2 = psycopg2.connect(dsn=dsn)
io1 = open('rdb_node.csv', 'w')
io2 = open('rdb_node_with_all_attri_view.csv', 'w')
sql1 = """copy (select * from rdb_node order by node_id_t, node_id) to STDOUT delimiter '|' csv header"""
sql2 = """copy (select * from rdb_node_with_all_attri_view order by node_id_t, node_id) to STDOUT delimiter '|' csv header"""
def table_size(table_name, c):
cur = c.cursor()
cur.execute("select pg_size_pretty(pg_relation_size('%s'));" % table_name)
s = cur.fetchone()[0]
cur.close()
return s
print 'rdb_node size:', table_size('rdb_node', conn1)
print 'rdb_node_with_all_attri_view:', table_size('rdb_node_with_all_attri_view', conn1)
def work(conn, sql, io):
ss = time.time()
cur = conn.cursor()
cur.copy_expe