diff --git a/python/examples/imagenet/README.md b/python/examples/imagenet/README.md
index ad8b12b5b..7edcd0201 100644
--- a/python/examples/imagenet/README.md
+++ b/python/examples/imagenet/README.md
@@ -47,3 +47,13 @@ client send inference request
 python resnet50_rpc_client.py ResNet50_vd_client_config/serving_client_conf.prototxt
 ```
 *the port of server side in this example is 9696
+
+### Launch Paddle Serving on Kubernetes
+
+Paddle Serving support deployment on Kubernetes (K8S) clusters. From `imagenet_k8s_rpc.yaml` we define Serving as K8S Deployment and Service. User can deploy Serving on containers and expose internal or external service.
+
+We strongly recommend [Baidu Cloud CCE Cluster](https://cloud.baidu.com/search.html?q=CCE)
+
+```
+kubectl apply -f imagenet_k8s_rpc.yaml
+```
diff --git a/python/examples/imagenet/README_CN.md b/python/examples/imagenet/README_CN.md
index 8650d51a6..026690103 100644
--- a/python/examples/imagenet/README_CN.md
+++ b/python/examples/imagenet/README_CN.md
@@ -47,3 +47,13 @@ client端进行预测
 python resnet50_rpc_client.py ResNet50_vd_client_config/serving_client_conf.prototxt
 ```
 *server端示例中服务端口为9696端口
+
+### K8S启动
+
+还可以运用K8S启动，在`imagenet_k8s_rpc.yaml`中定义了Serving的Deployment和Service，用户可以在K8S集群上启动Deployment并且通过Service对外暴露服务，用户可以在此基础上进行二次开发。
+
+推荐百度云的[CCE（K8S）集群](https://cloud.baidu.com/search.html?q=CCE)
+
+```
+kubectl apply -f imagenet_k8s_rpc.yaml
+```
diff --git a/python/examples/imagenet/imagenet_k8s_rpc.yaml b/python/examples/imagenet/imagenet_k8s_rpc.yaml
new file mode 100644
index 000000000..da93b42cb
--- /dev/null
+++ b/python/examples/imagenet/imagenet_k8s_rpc.yaml
@@ -0,0 +1,55 @@
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+apiVersion: apps/v1beta1
+kind: Deployment
+metadata:
+  name: paddleserving
+  labels:
+    app: paddleserving
+spec:
+  replicas: 1
+  template:
+    metadata:
+      name: paddleserving
+      labels:
+        app: paddleserving
+    spec:
+      containers:
+      - name: paddleserving
+        image: hub.baidubce.com/paddlepaddle/serving:latest
+        imagePullPolicy: Always
+        workingDir: /
+        command: ['/bin/bash', '-c']
+        args: ['pip install -U paddle-serving-server paddle-serving-client paddle-serving-app && \
+                python -m paddle_serving_app.package --get_model resnet_v2_50_imagenet && \
+                tar xf resnet_v2_50_imagenet.tar.gz && \
+                python -m paddle_serving_server.serve --model resnet_v2_50_imagenet_model/ --port 9696']
+        ports: 
+        - containerPort: 9696
+          name: serving
+
+---
+
+apiVersion: v1
+kind: Service
+metadata:
+  name: paddleserving
+spec:
+  ports:
+    - name: paddleserving
+      port: 9696
+      targetPort: 9696
+  selector:
+    app: paddleserving
diff --git a/python/examples/pipeline/faster_rcnn/000000570688.jpg b/python/examples/pipeline/faster_rcnn/000000570688.jpg
new file mode 100644
index 000000000..cb304bd56
Binary files /dev/null and b/python/examples/pipeline/faster_rcnn/000000570688.jpg differ
diff --git a/python/examples/pipeline/faster_rcnn/benchmark.py b/python/examples/pipeline/faster_rcnn/benchmark.py
new file mode 100644
index 000000000..9fdb48f7e
--- /dev/null
+++ b/python/examples/pipeline/faster_rcnn/benchmark.py
@@ -0,0 +1,107 @@
+import sys
+import os
+import yaml
+import requests
+import time
+import json
+import cv2
+import base64
+try:
+    from paddle_serving_server_gpu.pipeline import PipelineClient
+except ImportError:
+    from paddle_serving_server.pipeline import PipelineClient
+import numpy as np
+from paddle_serving_client.utils import MultiThreadRunner
+from paddle_serving_client.utils import benchmark_args, show_latency
+
+def cv2_to_base64(image):
+    return base64.b64encode(image).decode('utf8')
+
+def parse_benchmark(filein, fileout):
+    with open(filein, "r") as fin:
+        res = yaml.load(fin)
+        del_list = []
+        for key in res["DAG"].keys():
+            if "call" in key:
+                del_list.append(key)
+        for key in del_list:
+            del res["DAG"][key]
+    with open(fileout, "w") as fout:
+        yaml.dump(res, fout, default_flow_style=False)
+
+def gen_yml(device):
+    fin = open("config.yml", "r")
+    config = yaml.load(fin)
+    fin.close()
+    config["dag"]["tracer"] = {"interval_s": 10}
+    if device == "gpu":
+        config["op"]["bert"]["local_service_conf"]["device_type"] = 1
+        config["op"]["bert"]["local_service_conf"]["devices"] = "2"        
+    with open("config2.yml", "w") as fout: 
+        yaml.dump(config, fout, default_flow_style=False)
+
+def run_http(idx, batch_size):
+    print("start thread ({})".format(idx))
+    url = "http://127.0.0.1:18082/faster_rcnn/prediction"
+    with open(os.path.join(".", "000000570688.jpg"), 'rb') as file:
+        image_data1 = file.read()
+    image = cv2_to_base64(image_data1)
+
+    start = time.time()
+    for i in range(10):
+        data = {"key": [], "value": []}
+        for j in range(batch_size):
+            data["key"].append("image_" + str(j))
+            data["value"].append(image)
+        r = requests.post(url=url, data=json.dumps(data))
+        print("done")
+    end = time.time()
+    return [[end - start]]
+
+def multithread_http(thread, batch_size):
+    multi_thread_runner = MultiThreadRunner()
+    result = multi_thread_runner.run(run_http , thread, batch_size)
+
+def run_rpc(thread, batch_size):
+    client = PipelineClient()
+    client.connect(['127.0.0.1:9998'])
+    with open("data-c.txt", 'r') as fin:
+        start = time.time()
+        lines = fin.readlines()
+        start_idx = 0
+        while start_idx < len(lines):
+            end_idx = min(len(lines), start_idx + batch_size)
+            feed = {}
+            for i in range(start_idx, end_idx):
+                feed[str(i - start_idx)] = lines[i]
+            ret = client.predict(feed_dict=feed, fetch=["res"])
+            start_idx += batch_size
+            if start_idx > 1000:
+                break
+        end = time.time()
+    return [[end - start]]
+
+
+def multithread_rpc(thraed, batch_size):
+    multi_thread_runner = MultiThreadRunner()
+    result = multi_thread_runner.run(run_rpc , thread, batch_size)
+
+if __name__ == "__main__":
+    if sys.argv[1] == "yaml":
+        mode = sys.argv[2] # brpc/  local predictor
+        thread = int(sys.argv[3])
+        device = sys.argv[4]
+        gen_yml(device)
+    elif sys.argv[1] == "run":
+        mode = sys.argv[2] # http/ rpc
+        thread = int(sys.argv[3])
+        batch_size = int(sys.argv[4])
+        if mode == "http":
+            multithread_http(thread, batch_size)
+        elif mode == "rpc":
+            multithread_rpc(thread, batch_size)
+    elif sys.argv[1] == "dump":
+        filein = sys.argv[2]
+        fileout = sys.argv[3]
+        parse_benchmark(filein, fileout)
+    
diff --git a/python/examples/pipeline/faster_rcnn/benchmark.sh b/python/examples/pipeline/faster_rcnn/benchmark.sh
new file mode 100644
index 000000000..7ff22a8b7
--- /dev/null
+++ b/python/examples/pipeline/faster_rcnn/benchmark.sh
@@ -0,0 +1,60 @@
+export FLAGS_profile_pipeline=1
+alias python3="python3.6"
+modelname="bert"
+# HTTP
+ps -ef | grep web_service | awk '{print $2}' | xargs kill -9 
+sleep 3
+python3 benchmark.py yaml local_predictor 1 cpu 
+rm -rf profile_log_$modelname
+for thread_num in 1  
+do
+  for batch_size in 1 2
+  do
+    echo "----FasterRCNN thread num: $thread_num batch size: $batch_size mode:http ----" >>profile_log_$modelname
+    rm -rf PipelineServingLogs
+    rm -rf cpu_utilization.py
+    python3 web_service.py >web.log 2>&1 &
+    sleep 3
+    nvidia-smi --id=2 --query-compute-apps=used_memory --format=csv -lms 100 > gpu_use.log 2>&1 &
+    nvidia-smi --id=2 --query-gpu=utilization.gpu --format=csv -lms 100 > gpu_utilization.log 2>&1 &
+    echo "import psutil\ncpu_utilization=psutil.cpu_percent(1,False)\nprint('CPU_UTILIZATION:', cpu_utilization)\n" > cpu_utilization.py
+    python3 benchmark.py run http $thread_num $batch_size
+    python3 cpu_utilization.py >>profile_log_$modelname
+    ps -ef | grep web_service | awk '{print $2}' | xargs kill -9
+    python3 benchmark.py dump benchmark.log benchmark.tmp
+    mv benchmark.tmp benchmark.log
+    awk 'BEGIN {max = 0} {if(NR>1){if ($modelname > max) max=$modelname}} END {print "MAX_GPU_MEMORY:", max}' gpu_use.log >> profile_log_$modelname
+    awk 'BEGIN {max = 0} {if(NR>1){if ($modelname > max) max=$modelname}} END {print "GPU_UTILIZATION:", max}' gpu_utilization.log >> profile_log_$modelname
+    cat benchmark.log >> profile_log_$modelname
+    #rm -rf gpu_use.log gpu_utilization.log
+  done
+done
+# RPC
+exit
+ps -ef | grep web_service | awk '{print $2}' | xargs kill -9
+sleep 3
+python3 benchmark.py yaml local_predictor 1 gpu
+
+for thread_num in 1 8 16
+do
+  for batch_size in 1 10 100
+  do
+    echo "----Bert thread num: $thread_num batch size: $batch_size mode:rpc ----" >>profile_log_$modelname
+    rm -rf PipelineServingLogs
+    rm -rf cpu_utilization.py
+    python3 web_service.py >web.log 2>&1 &
+    sleep 3
+    nvidia-smi --id=2 --query-compute-apps=used_memory --format=csv -lms 100 > gpu_use.log 2>&1 &
+    nvidia-smi --id=2 --query-gpu=utilization.gpu --format=csv -lms 100 > gpu_utilization.log 2>&1 &
+    echo "import psutil\ncpu_utilization=psutil.cpu_percent(1,False)\nprint('CPU_UTILIZATION:', cpu_utilization)\n" > cpu_utilization.py
+    python3 benchmark.py run rpc $thread_num $batch_size
+    python3 cpu_utilization.py >>profile_log_$modelname
+    ps -ef | grep web_service | awk '{print $2}' | xargs kill -9
+    python3 benchmark.py dump benchmark.log benchmark.tmp
+    mv benchmark.tmp benchmark.log
+    awk 'BEGIN {max = 0} {if(NR>1){if ($modelname > max) max=$modelname}} END {print "MAX_GPU_MEMORY:", max}' gpu_use.log >> profile_log_$modelname
+    awk 'BEGIN {max = 0} {if(NR>1){if ($modelname > max) max=$modelname}} END {print "GPU_UTILIZATION:", max}' gpu_utilization.log >> profile_log_$modelname
+    #rm -rf gpu_use.log gpu_utilization.log
+    cat benchmark.log >> profile_log_$modelname
+  done
+done
diff --git a/python/examples/pipeline/faster_rcnn/config.yml b/python/examples/pipeline/faster_rcnn/config.yml
new file mode 100644
index 000000000..50f9298e0
--- /dev/null
+++ b/python/examples/pipeline/faster_rcnn/config.yml
@@ -0,0 +1,17 @@
+dag:
+  is_thread_op: false
+  tracer:
+    interval_s: 10
+http_port: 18082
+op:
+  faster_rcnn:
+    local_service_conf:
+      client_type: local_predictor
+      concurrency: 2
+      device_type: 1
+      devices: '2'
+      fetch_list:
+      - save_infer_model/scale_0.tmp_1
+      model_config: serving_server/
+rpc_port: 9998
+worker_num: 20
diff --git a/python/examples/pipeline/faster_rcnn/label_list.txt b/python/examples/pipeline/faster_rcnn/label_list.txt
new file mode 100644
index 000000000..941cb4e13
--- /dev/null
+++ b/python/examples/pipeline/faster_rcnn/label_list.txt
@@ -0,0 +1,80 @@
+person
+bicycle
+car
+motorcycle
+airplane
+bus
+train
+truck
+boat
+traffic light
+fire hydrant
+stop sign
+parking meter
+bench
+bird
+cat
+dog
+horse
+sheep
+cow
+elephant
+bear
+zebra
+giraffe
+backpack
+umbrella
+handbag
+tie
+suitcase
+frisbee
+skis
+snowboard
+sports ball
+kite
+baseball bat
+baseball glove
+skateboard
+surfboard
+tennis racket
+bottle
+wine glass
+cup
+fork
+knife
+spoon
+bowl
+banana
+apple
+sandwich
+orange
+broccoli
+carrot
+hot dog
+pizza
+donut
+cake
+chair
+couch
+potted plant
+bed
+dining table
+toilet
+tv
+laptop
+mouse
+remote
+keyboard
+cell phone
+microwave
+oven
+toaster
+sink
+refrigerator
+book
+clock
+vase
+scissors
+teddy bear
+hair drier
+toothbrush
diff --git a/python/examples/pipeline/faster_rcnn/pipeline_http_client.py b/python/examples/pipeline/faster_rcnn/pipeline_http_client.py
new file mode 100644
index 000000000..7037afc2f
--- /dev/null
+++ b/python/examples/pipeline/faster_rcnn/pipeline_http_client.py
@@ -0,0 +1,35 @@
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# from paddle_serving_server.pipeline import PipelineClient
+import numpy as np
+import requests
+import json
+import cv2
+import base64
+import os
+
+
+def cv2_to_base64(image):
+    return base64.b64encode(image).decode('utf8')
+
+
+url = "http://127.0.0.1:18082/faster_rcnn/prediction"
+with open(os.path.join(".", "000000570688.jpg"), 'rb') as file:
+    image_data1 = file.read()
+image = cv2_to_base64(image_data1)
+
+for i in range(1):
+    data = {"key": ["image"], "value": [image]}
+    r = requests.post(url=url, data=json.dumps(data))
+    print(r.json())
diff --git a/python/examples/pipeline/faster_rcnn/web_service.py b/python/examples/pipeline/faster_rcnn/web_service.py
new file mode 100644
index 000000000..1f483a0e6
--- /dev/null
+++ b/python/examples/pipeline/faster_rcnn/web_service.py
@@ -0,0 +1,71 @@
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+try:
+    from paddle_serving_server_gpu.web_service import WebService, Op
+except ImportError:
+    from paddle_serving_server.web_service import WebService, Op
+import logging
+import numpy as np
+import sys
+import cv2
+from paddle_serving_app.reader import *
+import base64
+
+class FasterRCNNOp(Op):
+    def init_op(self):
+        self.img_preprocess = Sequential([
+            BGR2RGB(), Div(255.0),
+            Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225], False),
+            Resize((640, 640)), Transpose((2, 0, 1))
+        ])
+        self.img_postprocess = RCNNPostprocess("label_list.txt", "output")
+
+    def preprocess(self, input_dicts, data_id, log_id):
+        (_, input_dict), = input_dicts.items()
+        imgs = []
+        print("keys", input_dict.keys())
+        for key in input_dict.keys():
+            data = base64.b64decode(input_dict[key].encode('utf8'))
+            data = np.fromstring(data, np.uint8)
+            im = cv2.imdecode(data, cv2.IMREAD_COLOR)
+            im = self.img_preprocess(im)
+            imgs.append({
+              "image": im[np.newaxis,:],
+              "im_shape": np.array(list(im.shape[1:])).reshape(-1)[np.newaxis,:],
+              "scale_factor": np.array([1.0, 1.0]).reshape(-1)[np.newaxis,:],
+            })
+        feed_dict = {
+            "image": np.concatenate([x["image"] for x in imgs], axis=0),
+            "im_shape": np.concatenate([x["im_shape"] for x in imgs], axis=0),
+            "scale_factor": np.concatenate([x["scale_factor"] for x in imgs], axis=0)
+        }
+        for key in feed_dict.keys():
+            print(key, feed_dict[key].shape)
+        return feed_dict, False, None, ""
+
+    def postprocess(self, input_dicts, fetch_dict, log_id):
+        #print(fetch_dict)
+        res_dict = {"bbox_result": str(self.img_postprocess(fetch_dict))}
+        return res_dict, None, ""
+
+
+class FasterRCNNService(WebService):
+    def get_pipeline_response(self, read_op):
+        faster_rcnn_op = FasterRCNNOp(name="faster_rcnn", input_ops=[read_op])
+        return faster_rcnn_op
+
+
+fasterrcnn_service = FasterRCNNService(name="faster_rcnn")
+fasterrcnn_service.prepare_pipeline_config("config2.yml")
+fasterrcnn_service.run_service()
diff --git a/tools/Dockerfile.runtime_template b/tools/Dockerfile.runtime_template
new file mode 100644
index 000000000..abdca4eaa
--- /dev/null
+++ b/tools/Dockerfile.runtime_template
@@ -0,0 +1,56 @@
+# Dockerfile template
+FROM <<base_image>>
+
+RUN apt-get update && \
+    apt-get install -y make build-essential
+
+RUN apt-get update && \
+    apt-get install -y wget tar xz-utils bzip2  libcurl4-openssl-dev \
+    curl sed grep zlib1g-dev  libreadline-gplv2-dev libncursesw5-dev libssl-dev libsqlite3-dev tk-dev libgdbm-dev libc6-dev libbz2-dev && \
+    apt-get clean -y
+
+WORKDIR /usr/bin
+      COPY tools/dockerfile/build_scripts /build_scripts
+      RUN bash /build_scripts/install_gcc.sh gcc82 && rm -rf /build_scripts
+      RUN cp gcc gcc.bak && cp g++ g++.bak && rm gcc && rm g++
+      RUN ln -s /usr/local/gcc-8.2/bin/gcc /usr/local/bin/gcc
+      RUN ln -s /usr/local/gcc-8.2/bin/g++ /usr/local/bin/g++
+      RUN ln -s /usr/local/gcc-8.2/bin/gcc /usr/bin/gcc
+      RUN ln -s /usr/local/gcc-8.2/bin/g++ /usr/bin/g++
+      ENV PATH=/usr/local/gcc-8.2/bin:$PATH
+
+# install python
+WORKDIR /home
+      COPY tools/dockerfile/build_scripts /build_scripts
+      RUN bash /build_scripts/install_python.sh <<python_version>> && rm -rf /build_scripts
+      # Other 
+
+# install whl and bin
+WORKDIR /home
+      COPY tools/dockerfile/build_scripts /build_scripts
+      RUN bash /build_scripts/install_whl.sh 0.5.0 2.0.0 <<run_env>> <<python_version>> && rm -rf /build_scripts
+
+# install tensorrt
+WORKDIR /home
+      COPY tools/dockerfile/build_scripts /build_scripts
+      RUN bash /build_scripts/install_trt.sh && rm -rf /build_scripts
+
+# install go
+RUN wget -qO- https://dl.google.com/go/go1.14.linux-amd64.tar.gz | \
+    tar -xz -C /usr/local && \
+    mkdir /root/go && \
+    mkdir /root/go/bin && \
+    mkdir /root/go/src && \
+    echo "GOROOT=/usr/local/go" >> /root/.bashrc && \
+    echo "GOPATH=/root/go" >> /root/.bashrc && \
+    echo "PATH=/usr/local/go/bin:/root/go/bin:$PATH" >> /root/.bashrc
+
+RUN wget https://paddle-serving.bj.bcebos.com/others/centos_ssl.tar && \
+    tar xf centos_ssl.tar && rm -rf centos_ssl.tar && \
+    mv libcrypto.so.1.0.2k /usr/lib/libcrypto.so.1.0.2k && mv libssl.so.1.0.2k /usr/lib/libssl.so.1.0.2k && \
+    ln -sf /usr/lib/libcrypto.so.1.0.2k /usr/lib/libcrypto.so.10 && \
+    ln -sf /usr/lib/libssl.so.1.0.2k /usr/lib/libssl.so.10 && \
+    ln -sf /usr/lib/libcrypto.so.10 /usr/lib/libcrypto.so && \
+    ln -sf /usr/lib/libssl.so.10 /usr/lib/libssl.so
+
+EXPOSE 22
diff --git a/tools/dockerfile/build_scripts/install_gcc.sh b/tools/dockerfile/build_scripts/install_gcc.sh
index e75021b2a..bf0dd5f2e 100644
--- a/tools/dockerfile/build_scripts/install_gcc.sh
+++ b/tools/dockerfile/build_scripts/install_gcc.sh
@@ -39,6 +39,7 @@ if [ "$1" == "gcc82" ]; then
   ../gcc-8.2.0/configure --prefix=/usr/local/gcc-8.2 --enable-threads=posix --disable-checking --disable-multilib && \
   make -j8 && make install
   cd .. && rm -rf temp_gcc82
+  rm -rf gcc-8.2.0 gcc-8.2.0.tar.xz
   cp ${lib_so_6} ${lib_so_6}.bak  && rm -f ${lib_so_6} && 
   ln -s /usr/local/gcc-8.2/lib64/libgfortran.so.5 ${lib_so_5} && \
   ln -s /usr/local/gcc-8.2/lib64/libstdc++.so.6 ${lib_so_6} && \
diff --git a/tools/dockerfile/build_scripts/install_python.sh b/tools/dockerfile/build_scripts/install_python.sh
new file mode 100644
index 000000000..80d6554d8
--- /dev/null
+++ b/tools/dockerfile/build_scripts/install_python.sh
@@ -0,0 +1,38 @@
+#!/bin/bash
+
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+# 
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+# 
+#     http://www.apache.org/licenses/LICENSE-2.0
+# 
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+VERSION=$1
+
+if [[ "$VERSION" == "2.7" ]];then
+    wget -q https://www.python.org/ftp/python/2.7.15/Python-2.7.15.tgz && tar -xvf Python-2.7.15.tgz && cd Python-2.7.15 
+    ./configure --enable-unicode=ucs4 --enable-shared CFLAGS=-fPIC --prefix=/usr/local/ && make && make install -j8 > /dev/null && make altinstall > /dev/null && ldconfig
+    cd .. && rm -rf Python-2.7.15*
+    wget https://bootstrap.pypa.io/pip/2.7/get-pip.py
+    python2.7 get-pip.py
+    rm -rf get-pip.py
+elif [[ "$VERSION" == "3.6" ]];then
+    wget -q https://www.python.org/ftp/python/3.6.8/Python-3.6.8.tgz && \
+    tar -xzf Python-3.6.8.tgz && cd Python-3.6.8 && \
+    CFLAGS="-Wformat" ./configure --prefix=/usr/local/ --enable-shared > /dev/null && \
+    make -j8 > /dev/null && make altinstall > /dev/null && ldconfig
+    cd .. && rm -rf Python-3.6.8*
+elif [[ "$VERSION" == "3.7" ]];then
+    wget -q https://www.python.org/ftp/python/3.7.0/Python-3.7.0.tgz && \
+    tar -xzf Python-3.7.0.tgz && cd Python-3.7.0 && \
+    CFLAGS="-Wformat" ./configure --prefix=/usr/local/ --enable-shared > /dev/null && \
+    make -j8 > /dev/null && make altinstall > /dev/null && ldconfig
+    cd .. && rm -rf Python-3.7.0*
+fi
diff --git a/tools/dockerfile/build_scripts/install_whl.sh b/tools/dockerfile/build_scripts/install_whl.sh
new file mode 100644
index 000000000..8d27e953b
--- /dev/null
+++ b/tools/dockerfile/build_scripts/install_whl.sh
@@ -0,0 +1,56 @@
+#!/bin/bash
+
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+# 
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+# 
+#     http://www.apache.org/licenses/LICENSE-2.0
+# 
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+SERVING_VERSION=$1
+PADDLE_VERSION=$2
+RUN_ENV=$3 # cpu/10.1 10.2
+PYTHON_VERSION=$4
+
+client_release="paddle-serving-client==$SERVING_VERSION"
+app_release="paddle-serving-app==0.3.1"
+if [[ "$RUN_ENV" == "cpu" ]];then
+    server_release="paddle-serving-server==$SERVING_VERSION"
+    python$PYTHON_VERSION -m pip install $client_release $app_release $server_release
+    python$PYTHON_VERSION -m pip install paddlepaddle==${PADDLE_VERSION}
+    cd /usr/local/
+    wget https://paddle-serving.bj.bcebos.com/bin/serving-cpu-noavx-openblas-${SERVING_VERSION}.tar.gz 
+    tar xf serving-cpu-noavx-openblas-${SERVING_VERSION}.tar.gz
+    echo "export SERVING_BIN=$PWD/serving-cpu-noavx-openblas-${SERVING_VERSION}/serving">>/root/.bashrc
+    rm -rf serving-cpu-noavx-openblas-${SERVING_VERSION}.tar.gz
+    cd -
+elif [[ "$RUN_ENV" == "cuda10.1" ]];then
+    server_release="paddle-serving-server-gpu==$SERVING_VERSION.post101"
+    python$PYTHON_VERSION -m pip install $client_release $app_release $server_release
+    python$PYTHON_VERSION -m pip install paddlepaddle-gpu==${PADDLE_VERSION}
+    cd /usr/local/
+    wget https://paddle-serving.bj.bcebos.com/bin/serving-gpu-101-${SERVING_VERSION}.tar.gz
+    tar xf serving-gpu-101-${SERVING_VERSION}.tar.gz
+    echo "export SERVING_BIN=$PWD/serving-gpu-101-${SERVING_VERSION}/serving">>/root/.bashrc
+    rm -rf serving-gpu-101-${SERVING_VERSION}.tar.gz
+    cd -
+elif [[ "$RUN_ENV" == "cuda10.2" ]];then
+    server_release="paddle-serving-server-gpu==$SERVING_VERSION.post102"
+    python$PYTHON_VERSION -m pip install $client_release $app_release $server_release
+    python$PYTHON_VERSION -m pip install paddlepaddle-gpu==${PADDLE_VERSION}
+    cd /usr/local/
+    wget https://paddle-serving.bj.bcebos.com/bin/serving-gpu-102-${SERVING_VERSION}.tar.gz
+    tar xf serving-gpu-102-${SERVING_VERSION}.tar.gz
+    echo "export SERVING_BIN=$PWD/serving-gpu-102-${SERVING_VERSION}/serving">>/root/.bashrc
+    rm -rf serving-gpu-102-${SERVING_VERSION}.tar.gz
+    cd -
+fi
+
+
diff --git a/tools/generate_runtime_docker.sh b/tools/generate_runtime_docker.sh
new file mode 100644
index 000000000..75a2baba1
--- /dev/null
+++ b/tools/generate_runtime_docker.sh
@@ -0,0 +1,80 @@
+#!/bin/sh
+
+#abort on error
+set -e
+
+function usage
+{
+    echo "usage: arg_parse_example -a AN_ARG -s SOME_MORE_ARGS [-y YET_MORE_ARGS || -h]"
+    echo "   ";
+    echo "   --env                 : running env, cpu/cuda10.1/cuda10.2/cuda11";
+    echo "   --python              : python version, 2.7/3.6/3.7 ";
+    echo "   --serving             : serving version(0.5.0)";
+    echo "   --paddle              : paddle version(2.0.1)"
+    echo "   --image_name          : image name(default serving_runtime:env-python)"
+    echo "  -h | --help            : helper";
+}
+
+function parse_args
+{
+  # positional args
+  args=()
+
+  # named args
+  while [ "$1" != "" ]; do
+      case "$1" in
+          --env )               env="$2";             shift;;
+          --python )            python="$2";     shift;;
+          --serving )           serving="$2";      shift;;
+          --paddle )            paddle="$2";      shift;;
+      --image_name )          image_name="$2";    shift;;
+          -h | --help )         usage;            exit;; # quit and show usage
+          * )                 args+=("$1")             # if no match, add it to the positional args
+      esac
+      shift # move to next kv pair
+  done
+  # restore positional args
+  set -- "${args[@]}"
+
+  # set positionals to vars
+  positional_1="${args[0]}"
+  positional_2="${args[1]}"
+
+  # validate required args
+  if [[ -z "${paddle}" || -z "${env}" || -z "${python}" || -z "${serving}" ]]; then
+      echo "Invalid arguments"
+      usage
+      exit;
+  fi
+
+  if [[ -z "${image_name}" ]]; then
+      image_name="serving_runtime:$env-$python"
+      echo "image_name is not assigned, so it will be set ($image_name)."
+  fi
+
+}
+
+
+function run
+{
+  parse_args "$@"
+
+  echo "named arg: env: $env"
+  if [ $env == "cpu" ]; then
+      base_image="ubuntu:16.04"
+  elif [ $env == "cuda10.1" ]; then
+      base_image="nvidia\/cuda:10.1-cudnn7-runtime-ubuntu16.04"
+  elif [ $env == "cuda10.2" ]; then
+      base_image="nvidia\/cuda:10.2-cudnn8-runtime-ubuntu16.04"
+  fi
+  echo "base image: $base_image"
+  echo "named arg: python: $python"
+  echo "named arg: serving: $serving"
+  echo "named arg: paddle: $paddle"
+  echo "named arg: image_name: $image_name"
+  
+  sed -e "s/<<base_image>>/$base_image/g" -e "s/<<python_version>>/$python/g" -e "s/<<run_env>>/$env/g" Dockerfile.runtime_template > Dockerfile.tmp
+  docker build --build-arg ftp_proxy=http://172.19.57.45:3128 --build-arg https_proxy=http://172.19.57.45:3128 --build-arg http_proxy=http://172.19.57.45:3128 --build-arg HTTP_PROXY=http://172.19.57.45:3128 --build-arg HTTPS_PROXY=http://172.19.57.45:3128 -t $image_name -f Dockerfile.tmp .
+}
+
+run "$@";