diff --git a/Cargo.lock b/Cargo.lock
index 130eaaeb61..68f3939c7e 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -1336,9 +1336,13 @@ checksum = "1fd0f2584146f6f2ef48085050886acf353beff7305ebd1ae69500e27c67f64b"
 [[package]]
 name = "bytes"
 version = "1.10.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d71b6127be86fdcfddb610f7182ac57211d4b18a3e9c82eb2d17662f2227ad6a"
+source = "git+https://github.com/discord9/bytes?rev=1572ab22c3cbad0e9b6681d1f68eca4139322a2a#1572ab22c3cbad0e9b6681d1f68eca4139322a2a"
 dependencies = [
+ "backtrace",
+ "crossbeam-channel",
+ "inferno 0.12.2",
+ "papaya",
+ "quanta",
  "serde",
 ]
 
@@ -8862,6 +8866,16 @@ dependencies = [
  "unicode-width 0.1.14",
 ]
 
+[[package]]
+name = "papaya"
+version = "0.2.3"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f92dd0b07c53a0a0c764db2ace8c541dc47320dad97c2200c2a637ab9dd2328f"
+dependencies = [
+ "equivalent",
+ "seize",
+]
+
 [[package]]
 name = "parking"
 version = "2.2.1"
@@ -10095,6 +10109,21 @@ dependencies = [
  "variadics",
 ]
 
+[[package]]
+name = "quanta"
+version = "0.12.6"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f3ab5a9d756f0d97bdc89019bd2e4ea098cf9cde50ee7564dde6b81ccc8f06c7"
+dependencies = [
+ "crossbeam-utils",
+ "libc",
+ "once_cell",
+ "raw-cpuid",
+ "wasi 0.11.1+wasi-snapshot-preview1",
+ "web-sys",
+ "winapi",
+]
+
 [[package]]
 name = "query"
 version = "0.18.0"
@@ -10394,6 +10423,15 @@ dependencies = [
  "thiserror 1.0.69",
 ]
 
+[[package]]
+name = "raw-cpuid"
+version = "11.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "498cd0dc59d73224351ee52a95fee0f1a617a2eae0e7d9d720cc622c73a54186"
+dependencies = [
+ "bitflags 2.9.1",
+]
+
 [[package]]
 name = "rawpointer"
 version = "0.2.1"
@@ -11334,6 +11372,16 @@ dependencies = [
  "libc",
 ]
 
+[[package]]
+name = "seize"
+version = "0.5.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5b55fb86dfd3a2f5f76ea78310a88f96c4ea21a3031f8d212443d56123fd0521"
+dependencies = [
+ "libc",
+ "windows-sys 0.52.0",
+]
+
 [[package]]
 name = "semver"
 version = "1.0.26"
diff --git a/Cargo.toml b/Cargo.toml
index 3ce3ff48dc..5ff2ceaf3c 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -333,6 +333,7 @@ datafusion-datasource = { git = "https://github.com/GreptimeTeam/datafusion.git"
 datafusion-sql = { git = "https://github.com/GreptimeTeam/datafusion.git", rev = "fd4b2abcf3c3e43e94951bda452c9fd35243aab0" }
 datafusion-substrait = { git = "https://github.com/GreptimeTeam/datafusion.git", rev = "fd4b2abcf3c3e43e94951bda452c9fd35243aab0" }
 sqlparser = { git = "https://github.com/GreptimeTeam/sqlparser-rs.git", rev = "4b519a5caa95472cc3988f5556813a583dd35af1" }                           # branch = "v0.58.x"
+bytes = { git = "https://github.com/discord9/bytes", rev = "1572ab22c3cbad0e9b6681d1f68eca4139322a2a" }
 
 [profile.release]
 debug = 1
diff --git a/docs/how-to/how-to-profile-memory.md b/docs/how-to/how-to-profile-memory.md
index b4bc00093a..b28085bde7 100644
--- a/docs/how-to/how-to-profile-memory.md
+++ b/docs/how-to/how-to-profile-memory.md
@@ -92,6 +92,9 @@ curl -X POST localhost:4000/debug/prof/mem > greptime.hprof
 curl -X POST "localhost:4000/debug/prof/mem?output=flamegraph" > greptime.svg
 # or output pprof format
 curl -X POST "localhost:4000/debug/prof/mem?output=proto" > greptime.pprof
+
+
+curl -X POST "localhost:4000/debug/prof/bytes" > greptime.svg
 ```
 
 You can periodically dump profiling data and compare them to find the delta memory usage.
diff --git a/src/servers/src/http.rs b/src/servers/src/http.rs
index 8fa658b6bb..70ba323cfe 100644
--- a/src/servers/src/http.rs
+++ b/src/servers/src/http.rs
@@ -913,6 +913,7 @@ impl HttpServer {
                         Router::new()
                             .route("/cpu", routing::post(pprof::pprof_handler))
                             .route("/mem", routing::post(mem_prof::mem_prof_handler))
+                            .route("/bytes", routing::post(mem_prof::bytes_prof_handler))
                             .route(
                                 "/mem/activate",
                                 routing::post(mem_prof::activate_heap_prof_handler),
diff --git a/src/servers/src/http/mem_prof.rs b/src/servers/src/http/mem_prof.rs
index e6362aef3f..a3d3e76915 100644
--- a/src/servers/src/http/mem_prof.rs
+++ b/src/servers/src/http/mem_prof.rs
@@ -39,6 +39,14 @@ pub struct MemPprofQuery {
     output: Output,
 }
 
+pub async fn bytes_prof_handler() -> crate::error::Result<impl IntoResponse> {
+    let dump = bytes::GLOBAL_TRACER
+        .get_or_init(|| bytes::BytesTracer::new().0)
+        .collector
+        .render_flamegraph()?;
+    Ok((StatusCode::OK, dump))
+}
+
 #[cfg(feature = "mem-prof")]
 #[axum_macros::debug_handler]
 pub async fn mem_prof_handler(