Details
-
Sub-task
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
-
None
Description
OM Echo throughput is affected by this Hadoop configuration.
The default is 1. Increasing it to 10 improves throughput by about 4x.
default:
ozone freon ome --clients=10 -n 500000 -t 100 rpc-payload count = 500000 mean rate = 35199.67 calls/second 1-minute rate = 34028.60 calls/second 5-minute rate = 33769.20 calls/second 15-minute rate = 33724.26 calls/second min = 0.72 milliseconds max = 19.81 milliseconds mean = 2.61 milliseconds stddev = 1.32 milliseconds median = 2.50 milliseconds 75% <= 2.86 milliseconds 95% <= 3.54 milliseconds 98% <= 4.28 milliseconds 99% <= 9.44 milliseconds 99.9% <= 19.81 milliseconds
after:
ozone freon ome --clients=30 -n 5000000 -t 100 rpc-payload count = 5000000 mean rate = 133982.86 calls/second 1-minute rate = 111649.80 calls/second 5-minute rate = 94971.80 calls/second 15-minute rate = 91514.17 calls/second min = 0.18 milliseconds max = 13.46 milliseconds mean = 0.65 milliseconds stddev = 0.64 milliseconds median = 0.61 milliseconds 75% <= 0.70 milliseconds 95% <= 0.85 milliseconds 98% <= 0.94 milliseconds 99% <= 1.96 milliseconds 99.9% <= 13.46 milliseconds
Attachments
Issue Links
- is a child of
-
HDDS-10755 [hsync] Investigate why DataNode Echo throughput is so low
- Resolved
- relates to
-
HDDS-11302 Ozone wrapper configurations to increase ipc.server.read.threadpool.size for SCM and Datanode
- Resolved
- links to