Skip to content

Latest commit

 

History

History
45 lines (30 loc) · 1.38 KB

benchmark_results_v0.4.0_vs_v0.3.1.rst

File metadata and controls

45 lines (30 loc) · 1.38 KB
Raw

Benchmarks measured using benchmark/simple_consume_bench.py and simple_produce_bench.py scripts.

Kafka environment:

  • Broker 0.10.2.1
  • Scala 2.11

Machine:

  • OS Mac OSX 10.13 (17A405)
  • 2.7 GHz Intel Core i7
  • 8 GB 1600 MHz DDR3

0.3.1 (vanila asyncio): Total consumed 1001600 messages in 24.43 second(s). Avg 40998.0 m/s. Total produced 500000 messages in 24.38 second(s). Avg 20505.0 m/s

0.3.1 (with uvloop): Total consumed 1001600 messages in 24.46 second(s). Avg 40953.0 m/s. Total produced 500000 messages in 23.36 second(s). Avg 21400.0 m/s

master Pure Python (vanila asyncio): Total consumed 1001600 messages in 18.93 second(s). Avg 52918.0 m/s. Total produced 500000 messages in 19.75 second(s). Avg 25311.0 m/s

master Pure Python (with uvloop): Total consumed 1001600 messages in 18.19 second(s). Avg 55057.0 m/s. Total produced 500000 messages in 15.79 second(s). Avg 31675.0 m/s

master C ext (vanila asyncio): Total consumed 1001600 messages in 4.99 second(s). Avg 200829.0 m/s. Total produced 500000 messages in 13.48 second(s). Avg 37103.0 m/s

master C ext (with uvloop): Total consumed 1001600 messages in 4.72 second(s). Avg 212248.0 m/s. Total produced 500000 messages in 10.24 second(s). Avg 48828.0 m/s

We see an overal boost in speed. With C extension it's ~4.9X speedup on read and ~1.8X on write. Without C extension we still have a good 29% boost on read and 23% on write.