Copied from api_gateway, removed REST handlers (port 8080), added tonic-web + CORS for grpc-web browser access. Binary renamed: api-gateway → api Changes: - Package name: api-gateway → api - Deleted src/handlers/ (REST ML endpoints on port 8080) - Added tonic-web 0.13 + tower-http CORS layer - Server::builder().accept_http1(true) for grpc-web - CORS_ORIGINS env var (default http://localhost:5173) - Metrics server on port 9091 (axum) preserved - All 95 lib tests pass, 0 clippy warnings - Added services/api to workspace members Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
449 lines
14 KiB
Rust
449 lines
14 KiB
Rust
//! Throughput Benchmark - Concurrent Authenticated Requests
|
|
//!
|
|
//! Measures maximum requests per second:
|
|
//! - TARGET: >100,000 req/s single-threaded
|
|
//! - Multi-threaded scaling
|
|
//! - Different authentication workloads
|
|
//! - Realistic traffic patterns
|
|
|
|
use criterion::{black_box, criterion_group, criterion_main, BenchmarkId, Criterion, Throughput};
|
|
use std::sync::atomic::{AtomicU64, Ordering};
|
|
use std::sync::Arc;
|
|
use std::time::{Duration, Instant};
|
|
use tokio::runtime::Runtime;
|
|
|
|
/// Lightweight auth simulator
|
|
struct AuthSimulator {
|
|
success_rate: f64,
|
|
counter: Arc<AtomicU64>,
|
|
}
|
|
|
|
impl AuthSimulator {
|
|
fn new(success_rate: f64) -> Self {
|
|
Self {
|
|
success_rate,
|
|
counter: Arc::new(AtomicU64::new(0)),
|
|
}
|
|
}
|
|
|
|
async fn authenticate(&self, _request_id: u64) -> bool {
|
|
let count = self.counter.fetch_add(1, Ordering::Relaxed);
|
|
// Simulate success rate
|
|
(count as f64 / 100.0) % 1.0 < self.success_rate
|
|
}
|
|
|
|
fn requests_processed(&self) -> u64 {
|
|
self.counter.load(Ordering::Relaxed)
|
|
}
|
|
}
|
|
|
|
/// Request handler
|
|
struct RequestHandler {
|
|
auth: Arc<AuthSimulator>,
|
|
}
|
|
|
|
impl RequestHandler {
|
|
fn new(auth: Arc<AuthSimulator>) -> Self {
|
|
Self { auth }
|
|
}
|
|
|
|
async fn handle_request(&self, request_id: u64) -> bool {
|
|
self.auth.authenticate(request_id).await
|
|
}
|
|
}
|
|
|
|
/// Benchmark 1: Single-threaded throughput (TARGET: >100K req/s)
|
|
fn bench_single_threaded_throughput(c: &mut Criterion) {
|
|
let rt = Runtime::new().unwrap();
|
|
let auth = Arc::new(AuthSimulator::new(0.95)); // 95% success rate
|
|
let handler = RequestHandler::new(auth.clone());
|
|
|
|
let mut group = c.benchmark_group("single_threaded_throughput");
|
|
group.throughput(Throughput::Elements(1));
|
|
|
|
group.bench_function("100k_req_target", |b| {
|
|
b.iter_custom(|iters| {
|
|
let start = Instant::now();
|
|
rt.block_on(async {
|
|
for i in 0..iters {
|
|
black_box(handler.handle_request(i).await);
|
|
}
|
|
});
|
|
start.elapsed()
|
|
});
|
|
});
|
|
|
|
group.finish();
|
|
}
|
|
|
|
/// Benchmark 2: Multi-threaded throughput
|
|
fn bench_multi_threaded_throughput(c: &mut Criterion) {
|
|
let mut group = c.benchmark_group("multi_threaded_throughput");
|
|
|
|
for num_threads in &[1, 2, 4, 8, 16] {
|
|
let rt = tokio::runtime::Builder::new_multi_thread()
|
|
.worker_threads(*num_threads)
|
|
.build()
|
|
.unwrap();
|
|
|
|
let auth = Arc::new(AuthSimulator::new(0.95));
|
|
let handler = Arc::new(RequestHandler::new(auth.clone()));
|
|
|
|
group.throughput(Throughput::Elements(1000));
|
|
|
|
group.bench_with_input(
|
|
BenchmarkId::new("concurrent_requests", num_threads),
|
|
num_threads,
|
|
|b, &threads| {
|
|
b.iter_custom(|iters| {
|
|
let start = Instant::now();
|
|
rt.block_on(async {
|
|
let mut handles = vec![];
|
|
let requests_per_thread = iters / threads as u64;
|
|
|
|
for _ in 0..threads {
|
|
let handler_clone = handler.clone();
|
|
let handle = tokio::spawn(async move {
|
|
for i in 0..requests_per_thread {
|
|
black_box(handler_clone.handle_request(i).await);
|
|
}
|
|
});
|
|
handles.push(handle);
|
|
}
|
|
|
|
for handle in handles {
|
|
handle.await.unwrap();
|
|
}
|
|
});
|
|
start.elapsed()
|
|
});
|
|
},
|
|
);
|
|
}
|
|
|
|
group.finish();
|
|
}
|
|
|
|
/// Benchmark 3: Different success rates
|
|
fn bench_success_rate_impact(c: &mut Criterion) {
|
|
let rt = Runtime::new().unwrap();
|
|
let mut group = c.benchmark_group("success_rate_impact");
|
|
|
|
for success_rate in &[0.5, 0.8, 0.95, 0.99, 1.0] {
|
|
let auth = Arc::new(AuthSimulator::new(*success_rate));
|
|
let handler = RequestHandler::new(auth.clone());
|
|
|
|
group.throughput(Throughput::Elements(1000));
|
|
|
|
group.bench_with_input(
|
|
BenchmarkId::new("throughput", format!("{}%", (success_rate * 100.0) as u32)),
|
|
success_rate,
|
|
|b, _rate| {
|
|
b.iter_custom(|iters| {
|
|
let start = Instant::now();
|
|
rt.block_on(async {
|
|
for i in 0..iters {
|
|
black_box(handler.handle_request(i).await);
|
|
}
|
|
});
|
|
start.elapsed()
|
|
});
|
|
},
|
|
);
|
|
}
|
|
|
|
group.finish();
|
|
}
|
|
|
|
/// Benchmark 4: Burst traffic patterns
|
|
fn bench_burst_patterns(c: &mut Criterion) {
|
|
let rt = Runtime::new().unwrap();
|
|
let auth = Arc::new(AuthSimulator::new(0.95));
|
|
let handler = Arc::new(RequestHandler::new(auth.clone()));
|
|
|
|
let mut group = c.benchmark_group("burst_patterns");
|
|
|
|
// Constant load
|
|
group.bench_function("constant_load_1000_req", |b| {
|
|
b.iter_custom(|iters| {
|
|
let start = Instant::now();
|
|
rt.block_on(async {
|
|
for i in 0..iters.min(1000) {
|
|
black_box(handler.handle_request(i).await);
|
|
}
|
|
});
|
|
start.elapsed()
|
|
});
|
|
});
|
|
|
|
// Burst pattern: All at once
|
|
group.bench_function("burst_1000_concurrent", |b| {
|
|
b.iter_custom(|iters| {
|
|
let start = Instant::now();
|
|
rt.block_on(async {
|
|
let mut handles = vec![];
|
|
for i in 0..iters.min(1000) {
|
|
let handler_clone = handler.clone();
|
|
let handle = tokio::spawn(async move { handler_clone.handle_request(i).await });
|
|
handles.push(handle);
|
|
}
|
|
for handle in handles {
|
|
black_box(handle.await.unwrap());
|
|
}
|
|
});
|
|
start.elapsed()
|
|
});
|
|
});
|
|
|
|
group.finish();
|
|
}
|
|
|
|
/// Benchmark 5: Request size impact on throughput
|
|
fn bench_request_size_throughput(c: &mut Criterion) {
|
|
let rt = Runtime::new().unwrap();
|
|
|
|
struct RequestProcessor {
|
|
auth: Arc<AuthSimulator>,
|
|
}
|
|
|
|
impl RequestProcessor {
|
|
async fn process(&self, _data: &[u8]) -> bool {
|
|
self.auth.authenticate(0).await
|
|
}
|
|
}
|
|
|
|
let auth = Arc::new(AuthSimulator::new(0.95));
|
|
let processor = RequestProcessor { auth: auth.clone() };
|
|
|
|
let mut group = c.benchmark_group("request_size_throughput");
|
|
|
|
for size in &[100, 1_000, 10_000, 100_000] {
|
|
let data = vec![0u8; *size];
|
|
|
|
group.throughput(Throughput::Bytes(*size as u64));
|
|
|
|
group.bench_with_input(
|
|
BenchmarkId::new("process_request", format!("{}B", size)),
|
|
&data,
|
|
|b, request_data| {
|
|
b.iter_custom(|iters| {
|
|
let start = Instant::now();
|
|
rt.block_on(async {
|
|
for _ in 0..iters {
|
|
black_box(processor.process(request_data).await);
|
|
}
|
|
});
|
|
start.elapsed()
|
|
});
|
|
},
|
|
);
|
|
}
|
|
|
|
group.finish();
|
|
}
|
|
|
|
/// Benchmark 6: Sustained throughput over time
|
|
fn bench_sustained_throughput(c: &mut Criterion) {
|
|
let rt = Runtime::new().unwrap();
|
|
let auth = Arc::new(AuthSimulator::new(0.95));
|
|
let handler = RequestHandler::new(auth.clone());
|
|
|
|
c.bench_function("sustained_1_second", |b| {
|
|
b.iter_custom(|_iters| {
|
|
let start = Instant::now();
|
|
let mut count = 0u64;
|
|
|
|
rt.block_on(async {
|
|
let end_time = Instant::now() + Duration::from_secs(1);
|
|
while Instant::now() < end_time {
|
|
black_box(handler.handle_request(count).await);
|
|
count += 1;
|
|
}
|
|
});
|
|
|
|
let elapsed = start.elapsed();
|
|
let rps = count as f64 / elapsed.as_secs_f64();
|
|
println!("Sustained throughput: {:.0} req/s", rps);
|
|
elapsed
|
|
});
|
|
});
|
|
}
|
|
|
|
/// Benchmark 7: Request rate limits
|
|
fn bench_rate_limited_throughput(c: &mut Criterion) {
|
|
let rt = Runtime::new().unwrap();
|
|
|
|
struct RateLimitedHandler {
|
|
auth: Arc<AuthSimulator>,
|
|
limit: AtomicU64,
|
|
max_rps: u64,
|
|
}
|
|
|
|
impl RateLimitedHandler {
|
|
fn new(auth: Arc<AuthSimulator>, max_rps: u64) -> Self {
|
|
Self {
|
|
auth,
|
|
limit: AtomicU64::new(0),
|
|
max_rps,
|
|
}
|
|
}
|
|
|
|
async fn handle(&self, request_id: u64) -> bool {
|
|
let count = self.limit.fetch_add(1, Ordering::Relaxed);
|
|
if count >= self.max_rps {
|
|
return false; // Rate limited
|
|
}
|
|
self.auth.authenticate(request_id).await
|
|
}
|
|
}
|
|
|
|
let mut group = c.benchmark_group("rate_limited_throughput");
|
|
|
|
for limit in &[1_000, 10_000, 100_000] {
|
|
let auth = Arc::new(AuthSimulator::new(0.95));
|
|
let handler = RateLimitedHandler::new(auth.clone(), *limit);
|
|
|
|
group.bench_with_input(BenchmarkId::new("max_rps", limit), limit, |b, _| {
|
|
b.iter_custom(|iters| {
|
|
let start = Instant::now();
|
|
rt.block_on(async {
|
|
for i in 0..iters {
|
|
black_box(handler.handle(i).await);
|
|
}
|
|
});
|
|
start.elapsed()
|
|
});
|
|
});
|
|
}
|
|
|
|
group.finish();
|
|
}
|
|
|
|
/// Benchmark 8: HFT scenario (TARGET: 100K req/s minimum)
|
|
fn bench_hft_scenario(c: &mut Criterion) {
|
|
let rt = Runtime::new().unwrap();
|
|
let auth = Arc::new(AuthSimulator::new(0.99)); // 99% success (HFT quality)
|
|
let handler = RequestHandler::new(auth.clone());
|
|
|
|
let mut group = c.benchmark_group("hft_scenario");
|
|
group.throughput(Throughput::Elements(100_000));
|
|
|
|
group.bench_function("hft_100k_target", |b| {
|
|
b.iter_custom(|iters| {
|
|
let start = Instant::now();
|
|
rt.block_on(async {
|
|
for i in 0..iters {
|
|
black_box(handler.handle_request(i).await);
|
|
}
|
|
});
|
|
let elapsed = start.elapsed();
|
|
|
|
// Calculate actual throughput
|
|
let rps = iters as f64 / elapsed.as_secs_f64();
|
|
if rps < 100_000.0 {
|
|
println!("⚠️ Below target: {:.0} req/s (target: 100K)", rps);
|
|
} else {
|
|
println!("✓ Target met: {:.0} req/s", rps);
|
|
}
|
|
|
|
elapsed
|
|
});
|
|
});
|
|
|
|
group.finish();
|
|
}
|
|
|
|
/// Benchmark 9: Latency under load
|
|
fn bench_latency_under_load(c: &mut Criterion) {
|
|
let rt = Runtime::new().unwrap();
|
|
let auth = Arc::new(AuthSimulator::new(0.95));
|
|
let handler = Arc::new(RequestHandler::new(auth.clone()));
|
|
|
|
let mut group = c.benchmark_group("latency_under_load");
|
|
|
|
for load in &[100, 1_000, 10_000, 100_000] {
|
|
group.bench_with_input(
|
|
BenchmarkId::new("requests_in_flight", load),
|
|
load,
|
|
|b, &n| {
|
|
b.iter_custom(|_iters| {
|
|
let start = Instant::now();
|
|
rt.block_on(async {
|
|
let mut handles = vec![];
|
|
for i in 0..n {
|
|
let handler_clone = handler.clone();
|
|
let handle =
|
|
tokio::spawn(async move { handler_clone.handle_request(i).await });
|
|
handles.push(handle);
|
|
}
|
|
for handle in handles {
|
|
black_box(handle.await.unwrap());
|
|
}
|
|
});
|
|
start.elapsed()
|
|
});
|
|
},
|
|
);
|
|
}
|
|
|
|
group.finish();
|
|
}
|
|
|
|
/// Benchmark 10: Request batching efficiency
|
|
fn bench_batching_efficiency(c: &mut Criterion) {
|
|
let rt = Runtime::new().unwrap();
|
|
let auth = Arc::new(AuthSimulator::new(0.95));
|
|
let handler = Arc::new(RequestHandler::new(auth.clone()));
|
|
|
|
let mut group = c.benchmark_group("batching_efficiency");
|
|
|
|
for batch_size in &[1, 10, 100, 1000] {
|
|
group.throughput(Throughput::Elements(*batch_size as u64));
|
|
|
|
group.bench_with_input(
|
|
BenchmarkId::new("batch_processing", batch_size),
|
|
batch_size,
|
|
|b, &n| {
|
|
b.iter_custom(|iters| {
|
|
let start = Instant::now();
|
|
rt.block_on(async {
|
|
for batch in 0..(iters / n as u64) {
|
|
let mut handles = vec![];
|
|
for i in 0..n {
|
|
let handler_clone = handler.clone();
|
|
let request_id = batch * n as u64 + i as u64;
|
|
let handle = tokio::spawn(async move {
|
|
handler_clone.handle_request(request_id).await
|
|
});
|
|
handles.push(handle);
|
|
}
|
|
for handle in handles {
|
|
black_box(handle.await.unwrap());
|
|
}
|
|
}
|
|
});
|
|
start.elapsed()
|
|
});
|
|
},
|
|
);
|
|
}
|
|
|
|
group.finish();
|
|
}
|
|
|
|
criterion_group!(
|
|
throughput_benches,
|
|
bench_single_threaded_throughput,
|
|
bench_multi_threaded_throughput,
|
|
bench_success_rate_impact,
|
|
bench_burst_patterns,
|
|
bench_request_size_throughput,
|
|
bench_sustained_throughput,
|
|
bench_rate_limited_throughput,
|
|
bench_hft_scenario,
|
|
bench_latency_under_load,
|
|
bench_batching_efficiency
|
|
);
|
|
|
|
criterion_main!(throughput_benches);
|