Adds benchmarks comparing performance to JVM based client using fs-hdfs3. Read and write throughput is lower for some reason, need to look into why that is. One thought is DataNode connection caching, but that doesn't seem to happen on the write path for the JVM client either so can't explain the full difference. RPC benchmark shows about 3x faster on native client for getting file status, so that's good
Adds benchmarks comparing performance to JVM based client using fs-hdfs3. Read and write throughput is lower for some reason, need to look into why that is. One thought is DataNode connection caching, but that doesn't seem to happen on the write path for the JVM client either so can't explain the full difference. RPC benchmark shows about 3x faster on native client for getting file status, so that's good