diff --git a/docker/modelzooserver/Dockerfile b/docker/modelzooserver/Dockerfile
index d21425e7bf..b589e74dd1 100644
--- a/docker/modelzooserver/Dockerfile
+++ b/docker/modelzooserver/Dockerfile
@@ -1,17 +1,24 @@
 # NOTE: The docker build context directory must be the root of the source tree.
 # NOTE: To build the release image, SQLFlow must be built into ./build
-FROM ubuntu:18.04
+FROM alpine:3.12
 
 ARG FIND_FASTED_MIRROR=true
 
-# Choose faster mirrors for apt-get and pip
+# Choose faster mirrors for alpine and pip
 # Install docker.io to release model defininiation
 COPY docker/dev/find_fastest_resources.sh /usr/local/bin/
-RUN bash -c 'if [ "$FIND_FASTED_MIRROR" == "true" ]; then source find_fastest_resources.sh && \
-             echo "Choose the fastest APT source ..." && \
-             choose_fastest_apt_source; fi' && \
-    apt-get -qq update && \
-    apt-get -qq install -y axel sudo docker.io
+RUN /bin/sh -c 'if [ "$FIND_FASTED_MIRROR" == "true" ]; then source find_fastest_resources.sh && \
+    echo "Choose the fastest Alpine source ..." && \
+    choose_fastest_alpine_source && \
+    echo "Choose the fastest PIP source ..." && \
+    choose_fastest_pip_source && \
+    apk add --no-cache python3 py3-pip sudo bash docker-cli && \
+    wget -q -O /etc/apk/keys/sgerrand.rsa.pub http://cdn.sqlflow.tech/alpine/sgerrand.rsa.pub.txt && \
+    wget -q http://cdn.sqlflow.tech/alpine/glibc-2.31-r0.apk && \
+    apk add glibc-2.31-r0.apk && \
+    rm glibc-2.31-r0.apk && \
+    ln -s /usr/bin/python3 /usr/local/bin/python && \
+    ln -s /usr/bin/pip3 /usr/local/bin/pi; fi'
 
 # Install pre-built SQLFlow components.
 COPY build/modelzooserver /usr/local/bin/modelzooserver