Use monotonic clock for interval measurement (#6211)

Signed-off-by: Lifu Huang <lifu.hlf@gmail.com>

python/sglang/srt/mem_cache/hiradix_cache.py

@@ -335,13 +335,13 @@ class HiRadixCache(RadixCache):
             return value, last_node
 
     def _match_prefix_helper(self, node: TreeNode, key: List):
-        node.last_access_time = time.time()
+        node.last_access_time = time.monotonic()
         child_key = self.get_child_key_fn(key)
         value = []
 
         while len(key) > 0 and child_key in node.children.keys():
             child = node.children[child_key]
-            child.last_access_time = time.time()
+            child.last_access_time = time.monotonic()
             prefix_len = self.key_match_fn(child.key, key)
             if prefix_len < len(child.key):
                 new_node = self._split_node(child.key, child, prefix_len)
@@ -386,7 +386,7 @@ class HiRadixCache(RadixCache):
         return new_node
 
     def _insert_helper(self, node: TreeNode, key: List, value):
-        node.last_access_time = time.time()
+        node.last_access_time = time.monotonic()
         if len(key) == 0:
             return 0
 
@@ -395,7 +395,7 @@ class HiRadixCache(RadixCache):
 
         while len(key) > 0 and child_key in node.children.keys():
             node = node.children[child_key]
-            node.last_access_time = time.time()
+            node.last_access_time = time.monotonic()
             prefix_len = self.key_match_fn(node.key, key)
 
             if prefix_len == len(node.key):

python/sglang/srt/mem_cache/radix_cache.py

@@ -45,7 +45,7 @@ class TreeNode:
         self.key = None
         self.value = None
         self.lock_ref = 0
-        self.last_access_time = time.time()
+        self.last_access_time = time.monotonic()
 
         self.hit_count = 0
         # indicating the node is loading KV cache from host
@@ -322,14 +322,14 @@ class RadixCache(BasePrefixCache):
     ##### Internal Helper Functions #####
 
     def _match_prefix_helper(self, node: TreeNode, key: List):
-        node.last_access_time = time.time()
+        node.last_access_time = time.monotonic()
 
         child_key = self.get_child_key_fn(key)
 
         value = []
         while len(key) > 0 and child_key in node.children.keys():
             child = node.children[child_key]
-            child.last_access_time = time.time()
+            child.last_access_time = time.monotonic()
             prefix_len = self.key_match_fn(child.key, key)
             if prefix_len < len(child.key):
                 new_node = self._split_node(child.key, child, prefix_len)
@@ -361,7 +361,7 @@ class RadixCache(BasePrefixCache):
         return new_node
 
     def _insert_helper(self, node: TreeNode, key: List, value):
-        node.last_access_time = time.time()
+        node.last_access_time = time.monotonic()
         if len(key) == 0:
             return 0
 
@@ -370,7 +370,7 @@ class RadixCache(BasePrefixCache):
         total_prefix_length = 0
         while len(key) > 0 and child_key in node.children.keys():
             node = node.children[child_key]
-            node.last_access_time = time.time()
+            node.last_access_time = time.monotonic()
             prefix_len = self.key_match_fn(node.key, key)
             total_prefix_length += prefix_len
             key = key[prefix_len:]