diff --git a/server/client.go b/server/client.go
index 7ba86ec..47d8f85 100644
--- a/server/client.go
+++ b/server/client.go
@@ -64,7 +64,7 @@ type client struct {
 
 	lastLogID uint64
 
-	reqErr chan error
+	// reqErr chan error
 
 	buf bytes.Buffer
 
@@ -82,7 +82,7 @@ func newClient(app *App) *client {
 	c.ldb = app.ldb
 	c.db, _ = app.ldb.Select(0) //use default db
 
-	c.reqErr = make(chan error)
+	// c.reqErr = make(chan error)
 
 	return c
 }
diff --git a/server/client_resp.go b/server/client_resp.go
index 00d0e95..a7ec21c 100644
--- a/server/client_resp.go
+++ b/server/client_resp.go
@@ -77,26 +77,31 @@ func (c *respClient) run() {
 		c.app.removeSlave(c.client, handleQuit)
 	}()
 
-	// done := make(chan error)
+	done := make(chan error)
 	for {
-		// go func() {
-		reqData, err := c.readRequest()
+		// I still don't know why use goroutine can improve performance
+		// if someone knows and benchamrks with another different result without goroutine, please tell me
+		go func() {
+			reqData, err := c.readRequest()
+			if err == nil {
+				c.handleRequest(reqData)
+			}
+
+			done <- nil
+		}()
+
+		// reqData, err := c.readRequest()
+		// if err == nil {
+		// 	c.handleRequest(reqData)
+		// }
+
+		err := <-done
 		if err != nil {
-			// done <- err
 			return
 		}
-
-		c.handleRequest(reqData)
-		// done <- nil
-		// }()
-
-		// err := <-done
-		// if err != nil {
-		// 	return
-		// }
-		// if c.conn == nil {
-		// 	return
-		// }
+		if c.conn == nil {
+			return
+		}
 	}
 }