Improve efficiency of ETW metrics exporter

Cargo.toml

@@ -19,4 +19,7 @@ opentelemetry-http = "0.27"
 opentelemetry-proto = { version = "0.27", default-features = false }
 opentelemetry_sdk = { version = "0.27", default-features = false }
 opentelemetry-stdout = "0.27"
-opentelemetry-semantic-conventions = { version = "0.27", features = ["semconv_experimental"] }
+opentelemetry-semantic-conventions = { version = "0.27", features = [
+    "semconv_experimental",
+] }
+criterion = "0.5"

opentelemetry-etw-metrics/Cargo.toml

@@ -17,7 +17,8 @@ opentelemetry-proto = { workspace = true, features = ["gen-tonic", "metrics"] }
 async-trait = "0.1"
 prost = "0.13"
 tracelogging = "1.2.1"
-tracing = {version = "0.1", optional = true}
+tracing = { version = "0.1", optional = true }
+criterion = { workspace = true, features = ["html_reports", "async_tokio"] }
 
 [dev-dependencies]
 tokio = { version = "1.0", features = ["full"] }
@@ -28,3 +29,11 @@ default = ["internal-logs"]
 
 [package.metadata.cargo-machete]
 ignored = ["tracing"]
+
+[[bench]]
+name = "etw"
+harness = false
+
+[[bench]]
+name = "exporter"
+harness = false

opentelemetry-etw-metrics/benches/etw.rs

@@ -0,0 +1,29 @@
+//! run with `$ cargo bench --bench etw -- --exact <test_name>` to run specific test for logs
+//! So to run test named "fibonacci" you would run `$ cargo bench --bench etw -- --exact fibonacci`
+//! To run all tests for logs you would run `$ cargo bench --bench etw`
+//!
+/*
+The benchmark results:
+criterion = "0.5.1"
+OS: Windows 11 Enterprise N, 23H2, Build 22631.4460
+Hardware: Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz   2.79 GHz, 16vCPUs
+RAM: 64.0 GB
+| Test                           | Average time|
+|--------------------------------|-------------|
+| write_event                    | 2.0649ns    |
+*/
+
+use criterion::{criterion_group, criterion_main, Criterion};
+use opentelemetry_etw_metrics::etw::write;
+
+fn write_event() {
+    let buffer = "This is a test buffer".as_bytes();
+    write(buffer);
+}
+
+fn criterion_benchmark(c: &mut Criterion) {
+    c.bench_function("write_event", |b| b.iter(|| write_event()));
+}
+
+criterion_group!(benches, criterion_benchmark);
+criterion_main!(benches);

opentelemetry-etw-metrics/benches/exporter.rs

@@ -0,0 +1,80 @@
+//! run with `$ cargo bench --bench exporter -- --exact <test_name>` to run specific test for logs
+//! So to run test named "fibonacci" you would run `$ cargo bench --bench exporter -- --exact fibonacci`
+//! To run all tests for logs you would run `$ cargo bench --bench exporter`
+//!
+/*
+The benchmark results:
+criterion = "0.5.1"
+OS: Windows 11 Enterprise N, 23H2, Build 22631.4460
+Hardware: Intel(R) Xeon(R) Platinum 8370C CPU @ 2.80GHz   2.79 GHz, 16vCPUs
+RAM: 64.0 GB
+| Test                           | Average time|
+|--------------------------------|-------------|
+| exporter                       | 847.38µs    |
+*/
+
+use opentelemetry::{InstrumentationScope, KeyValue};
+use opentelemetry_etw_metrics::MetricsExporter;
+
+use opentelemetry_sdk::{
+    metrics::{
+        data::{DataPoint, Metric, ResourceMetrics, ScopeMetrics, Sum},
+        exporter::PushMetricExporter,
+        Temporality,
+    },
+    Resource,
+};
+
+use criterion::{criterion_group, criterion_main, Criterion};
+
+async fn export(mut resource_metrics: ResourceMetrics) {
+    let exporter = MetricsExporter::new();
+    exporter.export(&mut resource_metrics).await.unwrap();
+}
+
+fn create_resource_metrics() -> ResourceMetrics {
+    let data_point = DataPoint {
+        attributes: vec![KeyValue::new("datapoint key", "datapoint value")],
+        start_time: Some(std::time::SystemTime::now()),
+        time: Some(std::time::SystemTime::now()),
+        value: 1.0_f64,
+        exemplars: vec![],
+    };
+
+    let sum: Sum<f64> = Sum {
+        data_points: vec![data_point.clone(), data_point.clone(), data_point],
+        temporality: Temporality::Delta,
+        is_monotonic: true,
+    };
+
+    let resource_metrics = ResourceMetrics {
+        resource: Resource::new(vec![KeyValue::new("service.name", "my-service")]),
+        scope_metrics: vec![ScopeMetrics {
+            scope: InstrumentationScope::default(),
+            metrics: vec![Metric {
+                name: "metric_name".into(),
+                description: "metric description".into(),
+                unit: "metric unit".into(),
+                data: Box::new(sum),
+            }],
+        }],
+    };
+
+    resource_metrics
+}
+
+fn criterion_benchmark(c: &mut Criterion) {
+    let runtime = tokio::runtime::Builder::new_multi_thread()
+        .worker_threads(1)
+        .enable_all()
+        .build()
+        .unwrap();
+
+    c.bench_function("export", |b| {
+        b.to_async(&runtime)
+            .iter(|| export(create_resource_metrics()))
+    });
+}
+
+criterion_group!(benches, criterion_benchmark);
+criterion_main!(benches);