Skip to main content

​
Overview

Avail provides comprehensive benchmarking capabilities for measuring pallet performance and Kate RPC operations. Benchmarks help determine runtime weights and identify performance bottlenecks.

​
Pallet Benchmarking

​
Prerequisites

Build the node with runtime benchmarking features enabled: 
cargo build --release --locked --features=runtime-benchmarks

​
Running Pallet Benchmarks

1

Benchmark a specific pallet

./target/release/avail-node benchmark pallet \
  --chain=dev \
  --steps=50 \
  --repeat=20 \
  --pallet="pallet_name" \
  --extrinsic="*" \
  --heap-pages=4096 \
  --output ./output/pallet_name.rs
2

Benchmark all pallets

Use the automated benchmark script:
cd ~/workspace/source
./scripts/run_benchmarks.sh
3

Review results

Benchmark outputs are saved to the ./output directory by default.

​
Benchmark Script Configuration

The run_benchmarks.sh script supports environment variables for customization:
VariableDefaultDescription
STEPS50Number of steps for weight calculation
REPEAT20Number of times to repeat each benchmark
TEMPLATE_PATH./.maintain/frame-weight-template.hbsWeight template file
OUTPUT_PATH./outputOutput directory for benchmark results
PALLETS*Specific pallets to benchmark (or * for all)
OUR_PALLETS-Set to benchmark only custom Avail pallets
Example: 
export STEPS=100
export REPEAT=50
export PALLETS="da_control pallet_vector"
./scripts/run_benchmarks.sh

​
Excluded Pallets

The following pallets are excluded from automatic benchmarking:
  • pallet_election_provider_support_benchmarking (helper pallet)
  • pallet_babe (no automatic benchmarking)
  • pallet_grandpa (no automatic benchmarking)
  • pallet_mmr (no automatic benchmarking)
  • pallet_offences (no automatic benchmarking)

​
Header Kate Commitment Benchmarks

Avail includes specialized benchmarks for the Kate commitment header builder:

​
Time Measurement Benchmarks

cargo bench --bench header_kate_commitment_cri
Uses the Criterion framework for detailed time measurements.
cargo bench --bench header_kate_commitment_divan
Uses the Divan framework for performance measurement.

​
Low-Level Performance Benchmarks

cargo bench --bench header_kate_commitment_iai_callgrind
Measures instructions, cache hits/misses, and main memory accesses using Valgrind’s Callgrind.
cargo bench --bench header_kate_commitment_iai
Alternative IAI benchmark for instruction counts and cache analysis.

​
Kate RPC Benchmarks

​
Running Kate RPC Benchmarks

These benchmarks require a running development node and Deno runtime.
1

Start development node

./avail-node --dev
2

Run RPC benchmarks

deno run -A ./examples/deno/benchmarks/query_proof.ts
deno run -A ./examples/deno/benchmarks/query_rows.ts
deno run -A ./examples/deno/benchmarks/query_block_length.ts
deno run -A ./examples/deno/benchmarks/query_data_proof.ts

​
Available Kate RPC Metrics

When Kate RPC metrics are enabled (--enable-kate-rpc-metrics), the following metrics are exposed:
  • avail_kate_rpc_query_rows_execution_time - Query rows execution time (μs)
  • avail_kate_rpc_query_proof_execution_time - Query proof execution time (μs)
  • avail_kate_rpc_query_block_length_execution_time - Query block length execution time (μs)
  • avail_kate_rpc_query_data_proof_execution_time - Query data proof execution time (μs)

​
Benchmark Subcommands

The avail-node benchmark command supports multiple subcommands: 
avail-node benchmark --help

​
Common Options

OptionDescription
--chain=<CHAIN>Chain specification (dev, turing, mainnet)
--steps=<N>Number of benchmark steps
--repeat=<N>Number of repetitions
--pallet=<NAME>Pallet to benchmark
--extrinsic=<NAME>Specific extrinsic or * for all
--heap-pages=<N>Heap pages for runtime execution
--output=<PATH>Output file path
--template=<PATH>Weight template file
--header=<PATH>License header file

​
Troubleshooting

If benchmarks fail, check the error log at ./output/benchmarking_errors.txt.

​
Common Issues

Build failures: 
# Ensure runtime-benchmarks feature is enabled
cargo build --release --locked --features=runtime-benchmarks
Out of memory errors: 
# Reduce the number of heap pages or steps
--heap-pages=2048 --steps=25
Pallet not found: 
# List all available pallets
./target/release/avail-node benchmark pallet --list --chain=dev

​
Best Practices

  1. Consistent environment - Run benchmarks on the same hardware for comparable results
  2. Minimal load - Close other applications to avoid interference
  3. Multiple runs - Use --repeat=20 or higher for statistical significance
  4. Sufficient steps - Use --steps=50 minimum for accurate weight curves
  5. Documentation - Record hardware specs and benchmark parameters for reproducibility

Build docs developers (and LLMs) love

Get started for freeTalk to us