diff --git a/hack/load-test/README.md b/hack/load-test/README.md
index dc85dc049..18682495d 100644
--- a/hack/load-test/README.md
+++ b/hack/load-test/README.md
@@ -41,6 +41,10 @@ Set `CLUSTER=3` will inject the `CLUSTER` variable to the app template. You can
 
 > To make multicluster load testing environment, you can set up multiple k3d instances and register them in the control plane.
 
+#### QPS
+
+By default, there is no rate limit for the client. If you want to set the QPS for each worker, you can use `QPS=2`.
+
 ### Cleanup
 
-Run `SHARD=3 WORKER=4 BEGIN=0 SIZE=1000 bash cleanup.sh` to delete `app-0` to `app-999` from namespace `load-test-0` to `load-test-2` in 4 threads.
\ No newline at end of file
+Run `SHARD=3 WORKER=4 BEGIN=0 SIZE=1000 bash cleanup.sh` to delete `app-0` to `app-999` from namespace `load-test-0` to `load-test-2` in 4 threads.
diff --git a/hack/load-test/deploy.sh b/hack/load-test/deploy.sh
index 3dcd6db29..1837987b4 100644
--- a/hack/load-test/deploy.sh
+++ b/hack/load-test/deploy.sh
@@ -5,6 +5,7 @@ SIZE=${SIZE:-1000}
 WORKER=${WORKER:-6}
 VERSION=${VERSION:-1}
 CLUSTER=${CLUSTER:-4}
+QPS=${QPS:-1}
 
 SHARD=${SHARD:-3}
 
@@ -12,6 +13,8 @@ TEMPLATE=${TEMPLATE:-"light"}
 
 END=$(expr $BEGIN + $SIZE - 1)
 
+waitTime=$(expr 1000 / $QPS)e-3
+
 run() {
   for i in $(seq $1 $3 $2); do
     sid=$(expr $i % $SHARD)
@@ -24,6 +27,7 @@ run() {
       sed 's/CLUSTER/'$c'/g' | \
       kubectl apply -f -
     echo "worker $4: apply app $i to $sid"
+    sleep $waitTime
   done
   echo "worker $4: done"
 }
@@ -32,4 +36,4 @@ for i in $(seq 1 $WORKER); do
   run $(expr $BEGIN + $i - 1) $END $WORKER $i &
 done
 
-wait
\ No newline at end of file
+wait