After 10million docs indexed the insert /search operation become slower and slower.
We'are using elasticsearch-0.19
elasticsearch setting:3 nodes, 5 shards, 2 replicas
test case: Each document contains 5 fileds (300 byte)
test setting: 150 threads（One client send indexing request, the other send searching request）
At the beginning, the throughput is high(about 2000-3000/sec),
but after 10million docs indexed the insert/search operation become slower and slower.
(index: 60/sec, search:160/sec ).
Why is there such a big contrast?
Is there any way make the indexing/searching faster?
or is this expected?