hive-hbase整合后查询缓慢

[chuckpu]

hive和hbase整合后发现
select count(*) from hbase_test
的map阶段一直停留在0% 大概过了20多分钟才开始不为0
hbase数据量在850W条左右
我直接导入到hive的表中的数据做这个查询也就30s左右
不知道是什么情况呢？
谢谢
2014-05-27 16:04:46,122 Stage-1 map = 0%, reduce = 0%, Cumulative CPU 927.95 sec
2014-05-27 16:04:47,128 Stage-1 map = 0%, reduce = 0%, Cumulative CPU 927.95 sec
2014-05-27 16:04:48,134 Stage-1 map = 0%, reduce = 0%, Cumulative CPU 927.95 sec
2014-05-27 16:04:49,140 Stage-1 map = 0%, reduce = 0%, Cumulative CPU 927.95 sec
2014-05-27 16:04:50,146 Stage-1 map = 0%, reduce = 0%, Cumulative CPU 927.95 sec
2014-05-27 16:04:51,153 Stage-1 map = 0%, reduce = 0%, Cumulative CPU 1002.69 sec
2014-05-27 16:04:52,158 Stage-1 map = 0%, reduce = 0%, Cumulative CPU 1002.69 sec
2014-05-27 16:04:53,166 Stage-1 map = 0%, reduce = 0%, Cumulative CPU 1002.69 sec
2014-05-27 16:04:54,172 Stage-1 map = 0%, reduce = 0%, Cumulative CPU 1002.69 sec
2014-05-27 16:04:55,178 Stage-1 map = 0%, reduce = 0%, Cumulative CPU 1002.69 sec
2014-05-27 16:04:56,183 Stage-1 map = 0%, reduce = 0%, Cumulative CPU 1002.69 sec
2014-05-27 16:04:57,189 Stage-1 map = 0%, reduce = 0%, Cumulative CPU 1002.69 sec
2014-05-27 16:04:58,195 Stage-1 map = 0%, reduce = 0%, Cumulative CPU 1002.69 sec
2014-05-27 16:04:59,201 Stage-1 map = 0%, reduce = 0%, Cumulative CPU 1002.69 sec

[vah101]

配置 hbase.client.scanner.caching 建hbase表的时候，注意预分region 最后利用rowkey作为查询条件 另外，hbase的数据冗余、读性能是瓶颈，跟orcfile方式比起来性能差距很大 可以参考这个： http://doc.okbase.net/superlxw1234/archive/48484.html