Statement Store Latency Benchmark

CLI tool for benchmarking statement store latency at scale. Clients form a ring topology where each subscribes to statements from the next client, measuring propagation latency across the network.

This crate produces two binaries:

• setup-allowances — one-shot provisioning of on-chain statement allowances via Sudo
• statement-latency-bench — the actual latency benchmark (can be re-run without re-provisioning)

Building

cargo build --release

Setup: Statement Allowances

Before running the benchmark, each account needs an on-chain statement allowance. Run this once (or whenever you change --num-clients):

setup-allowances \
  --rpc-endpoints ws://localhost:9944 \
  --sudo-seed "//Alice" \
  --num-clients 100

This submits Sudo(batch_all(set_storage(...))) transactions to write allowances for all deterministic benchmark accounts, then verifies each allowance exists at the finalized block.

setup-allowances Arguments

Argument	Description	Default
--rpc-endpoints	Comma-separated WebSocket URLs (required)	-
--sudo-seed	Sudo seed/SURI, e.g. "//Alice" (required)	-
--num-clients	Number of accounts to provision	100
--allowance-batch-size	Accounts per set_storage call	100
--allowance-max-count	Max statements allowed per account	100000
--allowance-max-size	Max total statement bytes per account	1000000
--max-batch-calls	Max calls per batch_all transaction	100

Running the Benchmark

Basic example:

statement-latency-bench \
  --rpc-endpoints ws://localhost:9944,ws://localhost:9945 \
  --num-clients 10 \
  --messages-pattern "5:512"

Multi-round with custom settings:

statement-latency-bench \
  --rpc-endpoints ws://node1:9944,ws://node2:9944 \
  --num-clients 100 \
  --num-rounds 10 \
  --interval-ms 5000 \
  --messages-pattern "5:512,1:5120"

statement-latency-bench Arguments

Argument	Description	Default
--rpc-endpoints	Comma-separated WebSocket URLs (required)	-
--num-clients	Number of clients to spawn	100
--messages-pattern	Message pattern "count:size" (e.g., "5:512,3:1024")	"5:512"
--num-rounds	Number of benchmark rounds	1
--interval-ms	Interval between rounds (ms)	10000
--receive-timeout-ms	Timeout for receiving messages (ms)	5000
--statement-expiry-ms	Statement expiry time (ms)	600000
--skip-sync	Skip time synchronization (for local testing)	false

How It Works

1. Clients are distributed round-robin across RPC endpoints
2. Each client sends statements with unique topics
3. Each client subscribes to statements from the next client in the ring
4. Latency is measured from submission to receipt via subscription

Output

Results are logged with min/avg/max statistics for:

• Send duration
• Receive duration
• Full latency

Example output:

Benchmark Results: send_min=0.045s send_avg=0.123s send_max=0.234s receive_min=2.134s receive_avg=3.456s
receive_max=5.678s latency_min=2.234s latency_avg=3.567s latency_max=5.789s