travisperson / benchmarks

A place for community-submitted Filecoin benchmarks

Hardware benchmarks on the Filecoin testnet

This repo is a place to collect benchmarks collected from the Filecoin community. Submissions are reviewed manually, but we try to be quick. You can review the current benchmarks here.

Running a benchmark

IT IS IMPORTANT TO BUILD LOTUS FROM SOURCE TO TAKE ADVANTAGE OF ALL AVAILABLE OPTIMIZATIONS

The information below assumes you are already able to build lotus from source. The instructions below detail how to natively compile the filecoin-ffi when building lotus and its tools.

If you have not previosuly built lotus from source, please see the lotus docs.

To build the filecoin-ffi from source you will need to have rustup installed. You can install it by following the instructions at https://rustup.rs/.

You will need 600GB of free disk space available, replace the value of /storage with an appropriate path in the last command.

Build and Run

$ git clone --branch testnet/3 https://github.com/filecoin-project/lotus.git
$ cd lotus
$ env RUSTFLAGS="-C target-cpu=native -g" FFI_BUILD_FROM_SOURCE=1 make clean deps bench
$ env FIL_PROOFS_MAXIMIZE_CACHING=1 RUST_LOG=info TMPDIR=/storage ./bench sealing --storage-dir /storage/bench --sector-size 32GiB 2>&1 | tee bench.log

Adding a benchmark to the list

• Fork the repo
• Create a new file in the benchmarks/{YOUR_VERSION} folder. Copy the format of benchmarks/template.yaml.
• Submit a pull request

Dev server

If you want to test your addition locally or otherwise play around, yarn dev and you're off to the races.

Build

Running yarn parse-data turns the YAML files into the JSON files used during the build process. Adding a new .yaml file does not trigger a JSON rebuild automatically.

Deploy

Any new push to master triggers a rebuild and deploy.