Benchmarks

Lotus comes with a benchmarking tool that can be used to test how long each resource-intensive mining operation takes. This guide describes how to install the benchmarking tool, and some basic operations.

Installation #

1. You must have the Lotus repository on your computer. If you do not have an existing copy of the repository, clone it from GitHub:

   git clone https://github.com/filecoin-project/lotus.git ~/lotus
   

   This will output something like:

   Cloning into '/root/lotus'...
   remote: Enumerating objects: 93, done.
   ...
   Resolving deltas: 100% (51531/51531), done.
   

2. The lotus binary must be built and within the ~/lotus repository folder. If you just cloned the repository or have misplaced the lotus binary, build the project:

   cd ~/lotus
   make clean all && make install
   

   This will output something like:

   rm -rf  build/.filecoin-install build/.update-modules  lotus lotus-miner lotus-worker lotus-shed lotus-gateway lotus-seed lotus-pond lotus-townhall lotus-fountain lotus-chainwatch lotus-bench lotus-stats lotus-pcr lotus-health lotus-wallet testground
   make -C extern/filecoin-ffi/ clean
   ...
   install -C ./lotus /usr/local/bin/lotus
   install -C ./lotus-miner /usr/local/bin/lotus-miner
   install -C ./lotus-worker /usr/local/bin/lotus-worker
   

3. Call make lotus-bench to build the Lotus benchmark binary:

   make lotus-bench
   

   This will output something like:

   rm -f lotus-bench
   go build -o lotus-bench ./cmd/lotus-bench
   ...
   go run github.com/GeertJohan/go.rice/rice append --exec lotus-bench -i ./build
   

   This will produce a lotus-bench binary in the current folder.

4. You can now run lotus-bench against your system.

Usage #

Use the self-documenting feature of the tool to explore the different commands.

    ./lotus-bench --help

This will output something like:

  NAME:
  lotus-bench - Benchmark performance of lotus on your hardware

  USAGE:
  lotus-bench [global options] command [command options] [arguments...]

  VERSION:
  1.15.1

  COMMANDS:
  prove    Benchmark a proof computation
  sealing
  import   benchmark chain import and validation
  help, h  Shows a list of commands or help for one command

  GLOBAL OPTIONS:
  --help, -h     show help (default: false)
  --version, -v  print the version (default: false)

Commands #

Prove #

Benchmark a proof computation using lotus-bench prove [command options] [arguments...]. For example:

./lotus-bench prove

Available options:

Sealing #

Benchmark a sealing computation using lotus-bench sealing [command options] [arguments...]. For example:

./lotus-bench sealing

This will output something like:

2022-04-13T18:37:41.141+0200	INFO	lotus-bench	lotus-bench/main.go:103	Starting lotus-bench
...
----
results (v28) SectorSize:(536870912), SectorNumber:(1)
seal: addPiece: 1.815989865s (281.9 MiB/s)
seal: preCommit phase 1: 23.512164789s (21.78 MiB/s)
seal: preCommit phase 2: 11.748902504s (43.58 MiB/s)
seal: commit phase 1: 18.391743ms (27.19 GiB/s)
seal: commit phase 2: 5.698575481s (89.85 MiB/s)
seal: verify: 1.536285ms
unseal: 26.136081682s  (19.59 MiB/s)

generate candidates: 101.562µs (4.808 TiB/s)
compute winning post proof (cold): 665.669815ms
compute winning post proof (hot): 624.667643ms
verify winning post proof (cold): 17.949129ms
verify winning post proof (hot): 1.26867ms

compute window post proof (cold): 446.549819ms
compute window post proof (hot): 450.785891ms
verify window post proof (cold): 8.572963ms
verify window post proof (hot): 1.517569ms

Available options:

Import #

Benchmark chain import and validation using lotus-bench import command [command options] [arguments...]. For example:

./lotus-bench import analyze import.car

Available commands:

Available options:

GPUs #

The list of known-to-work supported GPUs is in the hardware-requirements.

Enabling a custom GPU #

If you want to test a GPU that is not explicitly supported, set the following environment variable:

export BELLMAN_CUSTOM_GPU="<NAME>:<NUMBER_OF_CORES>"

Here is an example of trying a GeForce GTX 1660 Ti with 1536 cores:

export BELLMAN_CUSTOM_GPU="NVIDIA GeForce 3090 Ti:10752"

Testing whether the GPU is used #

First, to watch GPU utilization run nvtop in one terminal, then in a separate terminal, run the benchmarking tool to simulate sealing of a sector of small size:

./lotus-bench sealing --sector-size=2KiB

This process uses a fair amount of GPU, and generally takes ~4 minutes to complete. If you do not see any activity in nvtop from lotus during the entire process, it is likely something is misconfigured with your GPU.

Edit this page on GitHub