chore(cubestore): Add simple allocator tracking stats for benchmarks

ovr · ovr · commit 28de77a41ba4 · 2025-08-18T18:04:19.000+02:00
diff --git a/rust/cubestore/Cargo.toml b/rust/cubestore/Cargo.toml
@@ -9,3 +9,6 @@ members = [
     "cuberpc",
     "cuberockstore",
 ]
+
+[profile.bench]
+debug = true
diff --git a/rust/cubestore/cubestore/benches/cachestore_queue.rs b/rust/cubestore/cubestore/benches/cachestore_queue.rs
@@ -1,4 +1,4 @@
-use criterion::{criterion_group, criterion_main, BenchmarkId, Criterion};
+use criterion::{criterion_group, BenchmarkId, Criterion};
 use cubestore::cachestore::{
     CacheStore, QueueAddPayload, QueueItemStatus, QueueKey, RocksCacheStore,
 };
@@ -7,6 +7,13 @@ use cubestore::CubeError;
 use std::sync::Arc;
 use tokio::runtime::{Builder, Runtime};
 
+mod tracking_allocator;
+
+use tracking_allocator::TrackingAllocator;
+
+#[global_allocator]
+static ALLOCATOR: TrackingAllocator = TrackingAllocator::new();
+
 fn prepare_cachestore(name: &str) -> Result<Arc<RocksCacheStore>, CubeError> {
     let config = Config::test(&name).update_config(|mut config| {
         // disable periodic eviction
@@ -188,6 +195,7 @@ fn do_get_bench(
 }
 
 fn do_benches(c: &mut Criterion) {
+    ALLOCATOR.reset_stats();
     let runtime = Builder::new_multi_thread().enable_all().build().unwrap();
 
     do_insert_bench(c, &runtime, 512, 64);
@@ -201,4 +209,8 @@ fn do_benches(c: &mut Criterion) {
 }
 
 criterion_group!(benches, do_benches);
-criterion_main!(benches);
+
+fn main() {
+    benches();
+    ALLOCATOR.print_stats();
+}
diff --git a/rust/cubestore/cubestore/benches/metastore.rs b/rust/cubestore/cubestore/benches/metastore.rs
@@ -1,4 +1,4 @@
-use criterion::{criterion_group, criterion_main, BenchmarkId, Criterion};
+use criterion::{criterion_group, BenchmarkId, Criterion};
 use cubestore::config::Config;
 use cubestore::metastore::{BaseRocksStoreFs, Column, ColumnType, MetaStore, RocksMetaStore};
 use cubestore::remotefs::LocalDirRemoteFs;
@@ -8,6 +8,13 @@ use std::fs;
 use std::sync::Arc;
 use tokio::runtime::{Builder, Runtime};
 
+mod tracking_allocator;
+
+use tracking_allocator::TrackingAllocator;
+
+#[global_allocator]
+static ALLOCATOR: TrackingAllocator = TrackingAllocator::new();
+
 fn prepare_metastore(name: &str) -> Result<Arc<RocksMetaStore>, CubeError> {
     let config = Config::test(name);
 
@@ -97,118 +104,86 @@ async fn bench_get_tables_with_path(
     }
 }
 
-fn benchmark_get_tables_with_path_small(c: &mut Criterion, runtime: &Runtime) {
+fn do_get_tables_with_path_bench(
+    c: &mut Criterion,
+    runtime: &Runtime,
+    num_schemas: usize,
+    tables_per_schema: usize,
+    iterations: usize,
+) {
+    let total_tables = num_schemas * tables_per_schema;
     let metastore = runtime.block_on(async {
-        let metastore = prepare_metastore("get_tables_with_path_small").unwrap();
-        populate_metastore(&metastore, 10, 10).await.unwrap(); // 100 tables
+        let metastore = prepare_metastore(&format!("get_tables_with_path_{}", total_tables)).unwrap();
+        populate_metastore(&metastore, num_schemas, tables_per_schema).await.unwrap();
         metastore
     });
 
     c.bench_with_input(
-        BenchmarkId::new("get_tables_with_path_small_include_non_ready_true", 100),
-        &100,
+        BenchmarkId::new("get_tables_with_path_include_non_ready_true", total_tables),
+        &iterations,
         |b, &iterations| {
             b.to_async(runtime)
                 .iter(|| bench_get_tables_with_path(&metastore, true, iterations));
         },
     );
 
     c.bench_with_input(
-        BenchmarkId::new("get_tables_with_path_small_include_non_ready_false", 100),
-        &100,
+        BenchmarkId::new("get_tables_with_path_include_non_ready_false", total_tables),
+        &iterations,
         |b, &iterations| {
             b.to_async(runtime)
                 .iter(|| bench_get_tables_with_path(&metastore, false, iterations));
         },
     );
 }
 
-fn benchmark_get_tables_with_path_medium(c: &mut Criterion, runtime: &Runtime) {
-    let metastore = runtime.block_on(async {
-        let metastore = prepare_metastore("get_tables_with_path_medium").unwrap();
-        populate_metastore(&metastore, 50, 20).await.unwrap(); // 1,000 tables
-        metastore
-    });
-
-    c.bench_with_input(
-        BenchmarkId::new("get_tables_with_path_medium_include_non_ready_true", 50),
-        &50,
-        |b, &iterations| {
-            b.to_async(runtime)
-                .iter(|| bench_get_tables_with_path(&metastore, true, iterations));
-        },
-    );
-
-    c.bench_with_input(
-        BenchmarkId::new("get_tables_with_path_medium_include_non_ready_false", 50),
-        &50,
-        |b, &iterations| {
-            b.to_async(runtime)
-                .iter(|| bench_get_tables_with_path(&metastore, false, iterations));
-        },
-    );
+async fn do_cold_cache_test(num_schemas: usize, tables_per_schema: usize) {
+    let fresh_metastore = prepare_metastore("cold_cache_fresh").unwrap();
+    populate_metastore(&fresh_metastore, num_schemas, tables_per_schema).await.unwrap();
+    let result = fresh_metastore.get_tables_with_path(false).await;
+    assert!(result.is_ok());
 }
 
-fn benchmark_get_tables_with_path_large(c: &mut Criterion, runtime: &Runtime) {
-    let metastore = runtime.block_on(async {
-        let metastore = prepare_metastore("get_tables_with_path_large").unwrap();
-        populate_metastore(&metastore, 25, 1000).await.unwrap(); // 25,000 tables
-        metastore
-    });
-
-    c.bench_with_input(
-        BenchmarkId::new("get_tables_with_path_large_include_non_ready_true", 10),
-        &10,
-        |b, &iterations| {
-            b.to_async(runtime)
-                .iter(|| bench_get_tables_with_path(&metastore, true, iterations));
-        },
-    );
-
-    c.bench_with_input(
-        BenchmarkId::new("get_tables_with_path_large_include_non_ready_false", 10),
-        &10,
-        |b, &iterations| {
-            b.to_async(runtime)
-                .iter(|| bench_get_tables_with_path(&metastore, false, iterations));
-        },
-    );
+async fn do_warm_cache_test(metastore: &Arc<RocksMetaStore>) {
+    let result = metastore.get_tables_with_path(false).await;
+    assert!(result.is_ok());
 }
 
-fn cold_vs_warm_cache_benchmark(c: &mut Criterion, runtime: &Runtime) {
+fn do_cold_vs_warm_cache_bench(
+    c: &mut Criterion,
+    runtime: &Runtime,
+    num_schemas: usize,
+    tables_per_schema: usize,
+) {
     let metastore = runtime.block_on(async {
         let metastore = prepare_metastore("cold_warm_cache").unwrap();
-        populate_metastore(&metastore, 20, 50).await.unwrap(); // 1,000 tables
+        populate_metastore(&metastore, num_schemas, tables_per_schema).await.unwrap();
         metastore
     });
 
-    // Cold cache benchmark (first call)
     c.bench_function("get_tables_with_path_cold_cache", |b| {
-        b.to_async(runtime).iter(|| async {
-            let fresh_metastore = prepare_metastore("cold_cache_fresh").unwrap();
-            populate_metastore(&fresh_metastore, 20, 50).await.unwrap();
-            let result = fresh_metastore.get_tables_with_path(false).await;
-            assert!(result.is_ok());
-        });
+        b.to_async(runtime).iter(|| do_cold_cache_test(num_schemas, tables_per_schema));
     });
 
-    // Warm cache benchmark (subsequent calls)
     c.bench_function("get_tables_with_path_warm_cache", |b| {
-        b.to_async(runtime).iter(|| async {
-            let result = metastore.get_tables_with_path(false).await;
-            assert!(result.is_ok());
-        });
+        b.to_async(runtime).iter(|| do_warm_cache_test(&metastore));
     });
 }
 
 fn do_benches(c: &mut Criterion) {
+    ALLOCATOR.reset_stats();
     let runtime = Builder::new_multi_thread().enable_all().build().unwrap();
 
-    benchmark_get_tables_with_path_small(c, &runtime);
-    benchmark_get_tables_with_path_medium(c, &runtime);
-    benchmark_get_tables_with_path_large(c, &runtime);
-    cold_vs_warm_cache_benchmark(c, &runtime);
+    do_get_tables_with_path_bench(c, &runtime, 10, 10, 100);
+    do_get_tables_with_path_bench(c, &runtime, 50, 20, 50);
+    do_get_tables_with_path_bench(c, &runtime, 25, 1000, 10);
+    
+    do_cold_vs_warm_cache_bench(c, &runtime, 20, 50);
 }
 
 criterion_group!(benches, do_benches);
-criterion_main!(benches);
+
+fn main() {
+    benches();
+    ALLOCATOR.print_stats();
+}
diff --git a/rust/cubestore/cubestore/benches/tracking_allocator.rs b/rust/cubestore/cubestore/benches/tracking_allocator.rs
@@ -0,0 +1,121 @@
+use std::alloc::{GlobalAlloc, Layout, System};
+use std::sync::atomic::{AtomicUsize, Ordering};
+
+pub struct TrackingAllocator {
+    inner: System,
+    allocations: AtomicUsize,
+    deallocations: AtomicUsize,
+    reallocations: AtomicUsize,
+    current_allocated: AtomicUsize,
+    peak_allocated: AtomicUsize,
+    total_allocated: AtomicUsize,
+}
+
+impl TrackingAllocator {
+    pub const fn new() -> Self {
+        Self {
+            inner: System,
+            allocations: AtomicUsize::new(0),
+            deallocations: AtomicUsize::new(0),
+            reallocations: AtomicUsize::new(0),
+            current_allocated: AtomicUsize::new(0),
+            peak_allocated: AtomicUsize::new(0),
+            total_allocated: AtomicUsize::new(0),
+        }
+    }
+
+    pub fn reset_stats(&self) {
+        self.allocations.store(0, Ordering::Relaxed);
+        self.deallocations.store(0, Ordering::Relaxed);
+        self.reallocations.store(0, Ordering::Relaxed);
+        self.current_allocated.store(0, Ordering::Relaxed);
+        self.peak_allocated.store(0, Ordering::Relaxed);
+        self.total_allocated.store(0, Ordering::Relaxed);
+    }
+
+
+    pub fn print_stats(&self) {
+        let allocations = self.allocations.load(Ordering::Relaxed);
+        let deallocations = self.deallocations.load(Ordering::Relaxed);
+        let reallocations = self.reallocations.load(Ordering::Relaxed);
+        let current_allocated = self.current_allocated.load(Ordering::Relaxed);
+        let peak_allocated = self.peak_allocated.load(Ordering::Relaxed);
+        let total_allocated = self.total_allocated.load(Ordering::Relaxed);
+
+        println!("=== FINAL MEMORY STATISTICS ===");
+        println!("Total allocations: {}", allocations);
+        println!("Total deallocations: {}", deallocations);
+        println!("Total reallocations: {}", reallocations);
+        println!("Current allocated: {} bytes ({:.2} MB)", current_allocated, current_allocated as f64 / 1024.0 / 1024.0);
+        println!("Peak allocated: {} bytes ({:.2} MB)", peak_allocated, peak_allocated as f64 / 1024.0 / 1024.0);
+        println!("Total allocated: {} bytes ({:.2} MB)", total_allocated, total_allocated as f64 / 1024.0 / 1024.0);
+        println!("===============================");
+    }
+
+    fn update_allocated(&self, size: usize, is_allocation: bool) {
+        if is_allocation {
+            self.allocations.fetch_add(1, Ordering::Relaxed);
+            self.total_allocated.fetch_add(size, Ordering::Relaxed);
+            let current = self.current_allocated.fetch_add(size, Ordering::Relaxed) + size;
+            
+            // Update peak if current exceeds it
+            let mut peak = self.peak_allocated.load(Ordering::Relaxed);
+            while current > peak {
+                match self.peak_allocated.compare_exchange_weak(
+                    peak,
+                    current,
+                    Ordering::Relaxed,
+                    Ordering::Relaxed,
+                ) {
+                    Ok(_) => break,
+                    Err(new_peak) => peak = new_peak,
+                }
+            }
+        } else {
+            self.deallocations.fetch_add(1, Ordering::Relaxed);
+            self.current_allocated.fetch_sub(size, Ordering::Relaxed);
+        }
+    }
+}
+
+unsafe impl GlobalAlloc for TrackingAllocator {
+    unsafe fn alloc(&self, layout: Layout) -> *mut u8 {
+        let ptr = self.inner.alloc(layout);
+        if !ptr.is_null() {
+            self.update_allocated(layout.size(), true);
+        }
+        ptr
+    }
+
+    unsafe fn dealloc(&self, ptr: *mut u8, layout: Layout) {
+        self.inner.dealloc(ptr, layout);
+        self.update_allocated(layout.size(), false);
+    }
+
+    unsafe fn realloc(&self, ptr: *mut u8, layout: Layout, new_size: usize) -> *mut u8 {
+        let new_ptr = self.inner.realloc(ptr, layout, new_size);
+        if !new_ptr.is_null() {
+            self.reallocations.fetch_add(1, Ordering::Relaxed);
+            
+            // Update counters: subtract old size, add new size
+            self.current_allocated.fetch_sub(layout.size(), Ordering::Relaxed);
+            self.total_allocated.fetch_add(new_size, Ordering::Relaxed);
+            let current = self.current_allocated.fetch_add(new_size, Ordering::Relaxed) + new_size;
+            
+            // Update peak if current exceeds it
+            let mut peak = self.peak_allocated.load(Ordering::Relaxed);
+            while current > peak {
+                match self.peak_allocated.compare_exchange_weak(
+                    peak,
+                    current,
+                    Ordering::Relaxed,
+                    Ordering::Relaxed,
+                ) {
+                    Ok(_) => break,
+                    Err(new_peak) => peak = new_peak,
+                }
+            }
+        }
+        new_ptr
+    }
+}

Original file line number	Diff line number	Diff line change
`@@ -9,3 +9,6 @@ members = [`
`9`	`9`	`"cuberpc",`
`10`	`10`	`"cuberockstore",`
`11`	`11`	`]`
	`12`	`+`
	`13`	`+[profile.bench]`
	`14`	`+debug = true`