improve multiprocessing

simulation/app.py

@@ -1,3 +1,4 @@
+import os
 from concurrent.futures import ProcessPoolExecutor, as_completed
 from dataclasses import dataclass
 
@@ -86,41 +87,46 @@ def run_simulation(n_servers=2, lam=40):
     }
 
 
-with ProcessPoolExecutor(max_workers=4) as executor:
+if __name__ == "__main__":
-    n_servers = [n + 1 for n in range(10)]
+    with ProcessPoolExecutor(max_workers=os.cpu_count()) as executor:
-    lam = [(n + 1) * 4 for n in range(10)]
+        n_servers = [n + 1 for n in range(10)]
-    parameter_values = [
+        lam = [(n + 1) * 4 for n in range(10)]
-        {"n_servers": n, "lam": m} for n in n_servers for m in lam for _ in range(10)
+        parameter_values = [
-    ]
+            {"n_servers": n, "lam": m}
-    futures = [
+            for n in n_servers
-        executor.submit(run_simulation, **parameters) for parameters in parameter_values
+            for m in lam
-    ]
+            for _ in range(100)
+        ]
+        futures = [
+            executor.submit(run_simulation, **parameters)
+            for parameters in parameter_values
+        ]
 
-    results = []
+        results = []
-    for future in tqdm(as_completed(futures), total=len(parameter_values)):
+        for future in tqdm(as_completed(futures), total=len(parameter_values)):
-        result = future.result()
+            result = future.result()
-        results.append(result)
+            results.append(result)
 
-    df = (
+        df = (
-        pl.DataFrame(results)
+            pl.DataFrame(results)
-        .fill_null(0)
+            .fill_null(0)
-        .group_by(["n_servers", "lam"])
+            .group_by(["n_servers", "lam"])
-        .agg(pl.all().mean())
+            .agg(pl.all().mean())
-        .sort(["n_servers", "lam"])
+            .sort(["n_servers", "lam"])
-    )
-
-    df_queue_time = df.pivot("lam", index="n_servers", values="queue_time").sort(
-        "n_servers"
-    )
-
-    def stats(column):
-        return (
-            df.pivot("lam", index="n_servers", values=column)
-            .sort("n_servers")
-            .with_columns(pl.all().round(2))
         )
 
-    pl.Config.set_tbl_cols(20)
+        df_queue_time = df.pivot("lam", index="n_servers", values="queue_time").sort(
-    print(df)
+            "n_servers"
-    print(stats("queue_time"))
+        )
-    print(stats("utilization_server_1"))
+
+        def stats(column):
+            return (
+                df.pivot("lam", index="n_servers", values=column)
+                .sort("n_servers")
+                .with_columns(pl.all().round(2))
+            )
+
+        pl.Config.set_tbl_cols(20)
+        print(df)
+        print(stats("queue_time"))
+        print(stats("utilization_server_1"))