Improve benchmarking

- The benchmark addon now manages setting up and tearing down the backend and
traffic processes itself.
- Use wrk instead of hey. I get more consistent results with this tool, and hey
shows a strange tail-latency bump that seems artificial.
- Make termination behaviour simpler. The bencmark revealed a bug where .done
events were not called if the proxy was shut down by an addon.

1 files changed, 9 insertions, 39 deletions

diff --git a/test/bench/README.md b/test/bench/README.md
index 05741c07..3d9e7ef7 100644
--- a/test/bench/README.md
+++ b/test/bench/README.md
@@ -1,7 +1,7 @@
 
-This directory contains a set of tools for benchmarking and profiling mitmproxy.
-At the moment, this is simply to give developers a quick way to see the impact
-of their work. Eventually, this might grow into a performance dashboard with
+This directory contains an addon for benchmarking and profiling mitmproxy. At
+the moment, this is simply to give developers a quick way to see the impact of
+their work. Eventually, this might grow into a performance dashboard with
 historical data, so we can track performance over time.
 
 
@@ -9,48 +9,18 @@ historical data, so we can track performance over time.
 
 Install the following tools:
 
-    go get -u github.com/rakyll/hey
+    https://github.com/wg/wrk
+
     go get github.com/cortesi/devd/cmd/devd
 
 You may also want to install snakeviz to make viewing profiles easier:
 
     pip install snakeviz
 
-In one window, run the devd server:
-
-    ./backend
-
-
-# Running tests
-
-Each run consists of two files - a mitproxy invocation, and a traffic generator.
-Make sure the backend is started, then run the proxy:
-
-    ./simple.mitmproxy
-
-Now run the traffic generator:
-
-    ./simple.traffic
-
-After the run is done, quit the proxy with ctrl-c.
-
-
-# Reading results
-
-Results are placed in the ./results directory. You should see two files - a
-performance log from **hey**, and a profile. You can view the profile like so:
-
-    snakeviz ./results/simple.prof
-
-
-
-
-
-
-
-
-
-
+Now run the benchmark by loading the addon. A typical invocation is as follows:
 
+    mitmdump -p0 -q --set benchmark_save_path=/tmp/foo -s ./benchmark.py
 
+This will start up the backend server, run the benchmark, save the results to
+/tmp/foo.bench and /tmp/foo.prof, and exit.