diff --git a/src/zmalloc.c b/src/zmalloc.c
index 0117d8d91..e740a4167 100644
--- a/src/zmalloc.c
+++ b/src/zmalloc.c
@@ -52,6 +52,7 @@ void zlibc_free(void *ptr) {
 #include <string.h>
 #include "zmalloc.h"
 #include <stdatomic.h>
+#include "serverassert.h"
 
 #define UNUSED(x) ((void)(x))
 
@@ -87,10 +88,31 @@ void zlibc_free(void *ptr) {
 #define dallocx(ptr, flags) je_dallocx(ptr, flags)
 #endif
 
-#define update_zmalloc_stat_alloc(__n) atomic_fetch_add_explicit(&used_memory, (__n), memory_order_relaxed)
-#define update_zmalloc_stat_free(__n) atomic_fetch_sub_explicit(&used_memory, (__n), memory_order_relaxed)
+#define update_zmalloc_stat_alloc(__n)                                                                                 \
+    do {                                                                                                               \
+        if (unlikely(thread_index == -1)) zmalloc_register_thread_index();                                             \
+        used_memory_thread[thread_index] += (__n);                                                                     \
+    } while (0)
+
+#define update_zmalloc_stat_free(__n)                                                                                  \
+    do {                                                                                                               \
+        if (unlikely(thread_index == -1)) zmalloc_register_thread_index();                                             \
+        used_memory_thread[thread_index] -= (__n);                                                                     \
+    } while (0)
+
+/* A thread-local storage which keep the current thread's index in the used_memory_thread array. */
+static __thread int thread_index = -1;
+/* MAX_THREADS_NUM = IO_THREADS_MAX_NUM(128) + BIO threads(3) + main thread(1). */
+#define MAX_THREADS_NUM 132
+static unsigned long long used_memory_thread[MAX_THREADS_NUM];
+
+static atomic_int total_active_threads;
 
-static _Atomic size_t used_memory = 0;
+/* Register the thread index in start_routine. */
+void zmalloc_register_thread_index(void) {
+    thread_index = atomic_fetch_add_explicit(&total_active_threads, 1, memory_order_relaxed);
+    assert(total_active_threads < MAX_THREADS_NUM);
+}
 
 static void zmalloc_default_oom(size_t size) {
     fprintf(stderr, "zmalloc: Out of memory trying to allocate %zu bytes\n", size);
@@ -388,7 +410,11 @@ char *zstrdup(const char *s) {
 }
 
 size_t zmalloc_used_memory(void) {
-    size_t um = atomic_load_explicit(&used_memory, memory_order_relaxed);
+    assert(total_active_threads < MAX_THREADS_NUM);
+    size_t um = 0;
+    for (int i = 0; i < total_active_threads; i++) {
+        um += used_memory_thread[i];
+    }
     return um;
 }
 
@@ -604,8 +630,6 @@ size_t zmalloc_get_rss(void) {
 
 #if defined(USE_JEMALLOC)
 
-#include "serverassert.h"
-
 #define STRINGIFY_(x) #x
 #define STRINGIFY(x) STRINGIFY_(x)
 
diff --git a/src/zmalloc.h b/src/zmalloc.h
index a909366c1..a068c671c 100644
--- a/src/zmalloc.h
+++ b/src/zmalloc.h
@@ -142,6 +142,7 @@ size_t zmalloc_get_memory_size(void);
 void zlibc_free(void *ptr);
 void zlibc_trim(void);
 void zmadvise_dontneed(void *ptr);
+void zmalloc_register_thread_index(void);
 
 #ifdef HAVE_DEFRAG
 void zfree_no_tcache(void *ptr);