add ALPS

README.md

@@ -145,20 +145,20 @@ cd http_client && ./test.sh
 ### Ablation Study
 The frontend scheduler offer two parameter to control policy fine-tuning. By defauly, policy fine-tuning is enable
 ```
-python3 --exp_result ../experiments/seals main.py
+python3 main.py
 ```
 And users can disable unpredictability finetuning and overload finetuning by
 ```
-python3 main.py --exp_result ../experiments/seals --unpred --overload 
+python3 main.py --unpred --overload 
 ```
 
 ### Sensitivity Analysis
 Users can control policy parameters 
 ```
-python3 main.py --alpha 1 --beta 1 --theta 50 --gamma 1 --exp_result ../experiments/seals
+python3 main.py --alpha 1 --beta 1 --theta 50 --gamma 1
 ```
 Users can also change machine learning methods by
 ```
-python3 main.py --ml avg --exp_result ../experiments/seals
-python3 main.py --ml LR --exp_result ../experiments/seals
+python3 main.py --ml avg
+python3 main.py --ml LR
 ```

frontend/main.py

@@ -78,18 +78,19 @@ def main():
     parser.add_argument("--overload", action='store_false', help="Overload fine-tune")
     args = parser.parse_args()
     try:
-        stop_event = threading.Event()
-        thread = threading.Thread(target=trace_print, args=(stop_event,args.exp_cpu,))
-        thread.start()
+        if args.exp_result != "":
+            stop_event = threading.Event()
+            thread = threading.Thread(target=trace_print, args=(stop_event,args.exp_cpu,))
+            thread.start()
         while True:
             client_socket, addr = server_socket.accept()
             handle_client(client_socket, args)
     except KeyboardInterrupt:
         print("Server is shutting down.")
     finally:
-        stop_event.set()
-        thread.join()
         if args.exp_result != "":
+            stop_event.set()
+            thread.join()
             update_cpuT(args.exp_cpu, args.exp_result)
         client_socket.close()
         server_socket.close()