grpcpulse

A CLI tool and Rust library to benchmark and compare gRPC endpoints side by side. Measures latency, throughput, and stream lag — built for comparing Solana RPC providers (Shyft, Helius, Triton) or any gRPC service.

Install

cargo install grpcpulse

Modes

Latency — round-trip benchmark

Fires N concurrent requests and reports p50/p95/p99 latency per endpoint:

grpcpulse latency \
  --endpoint "shyft=https://grpc.shyft.to|x-token:YOUR_TOKEN" \
  --endpoint "my-node=http://localhost:10000" \
  --requests 100 \
  --concurrency 10 \
  --warmup 5

+----------+---------+---------+---------+---------+--------+
| Endpoint | p50     | p95     | p99     | Max     | Errors |
+============================================================+
| my-node  | 4.12ms  | 8.34ms  | 12.01ms | 25.40ms | 0.0%   |
| shyft    | 93.10ms | 127ms   | 128ms   | 128ms   | 0.0%   |
+----------+---------+---------+---------+---------+--------+

Stream — Solana slot lag comparison

Subscribes to Geyser slot streams on all endpoints simultaneously and measures which provider delivers each slot first:

grpcpulse stream \
  --endpoint "shyft-ams=https://grpc.ams.shyft.to|x-token:YOUR_TOKEN" \
  --endpoint "shyft-sgp=https://grpc.sgp.shyft.to|x-token:YOUR_TOKEN" \
  --slots 50

+-----------+------+-------+--------------+------------+
| Endpoint  | Wins | Win % | Avg Lag (ms) | Slots Seen |
+=======================================================+
| shyft-ams |   36 | 69.2% |        16.53 |         52 |
| shyft-sgp |   16 | 30.8% |         2.38 |         52 |
+-----------+------+-------+--------------+------------+

Fastest: shyft-ams (36/50 slots, avg 0.00ms ahead)

Auth headers

Pass headers per endpoint using |key:value after the URL:

--endpoint "shyft=https://grpc.shyft.to|x-token:abc123"

Or globally with -H (applies to all endpoints):

-H "authorization:Bearer mytoken"

Use as a library

use grpcpulse::bench::{run, BenchConfig};
use grpcpulse::endpoint::Endpoint;
use grpcpulse::report::print_table;

#[tokio::main]
async fn main() {
    let endpoints = vec![
        Endpoint::new("shyft", "https://grpc.shyft.to")
            .with_header("x-token", "YOUR_TOKEN"),
        Endpoint::new("local", "http://localhost:10000"),
    ];

    let results = run(endpoints, BenchConfig::default()).await;
    print_table(&results);
}

Options

grpcpulse latency

grpcpulse stream

License

MIT