Performance Optimization Guidance

This is a reference benchmark process document designed to assist in performance analysis and optimization for client. This document provides as general a testing framework as possible, allowing developers with needs to adjust it according to their specific circumstances across various platform.

Flow

Preparation

Step 1: Setup Dragonfly

• Please refer to official doc.

Step 2: Start a file server

• Start with docker:

[object Object], FILE_SERVER_PORT=12345
docker run -d  --,[object Object], -p ,[object Object],:80 --name dragonfly-fs dragonflyoss/file-server:latest

• Check the file server is ready:

[object Object],
curl -s -o /dev/null \
     -w ,[object Object], \
     http://localhost:12345/nano \
| grep -q ,[object Object], \
&& ,[object Object], ,[object Object], \
|| ,[object Object], ,[object Object],

• Optional:

you can build your own image, take a reference from Dockerfile.

Step 3: Install test tools

• Request Generator: oha

brew install oha

• Profiling: flamegraph

cargo install flamegraph

Step 4: Setup Dragonfly Peer

Document: Install with binary.

• Compile the target binary

cargo build --release --bin dfdaemon

• Connect to Dragonfly

[object Object],
./target/release/dfdaemon --config client.yaml -l info --verbose

FlameGraph

Now, let's start benchmark with the following params:

• $FILE_SERVER_ADDRESS
• $CLIENT_PROXY_ADDRESS

Collect Flamegraph

• Capture the flamegraph:

[object Object],
,[object Object], flamegraph -o my_flamegraph.svg --pid 3442

• Make the request:

oha -c 1000 \
    -n 100 \
    --rand-regex-url ,[object Object],/\(nano\|micro\|small\|medium\|large\) \
    -x ,[object Object],