diff --git a/doc/run_with_hive.md b/doc/run_with_hive.md
new file mode 100644
index 0000000000..e9f81e6942
--- /dev/null
+++ b/doc/run_with_hive.md
@@ -0,0 +1,82 @@
+# How SQLFlow connects with Hive
+
+This document is a tutorial on how SQLFlow connects Hive via [HiveServer2](https://cwiki.apache.org/confluence/display/Hive/HiveServer2+Overview).
+
+## Connect Existing Hive server
+
+To connect an existing Hive server instance, we only need to configure a `datasource` string in the format of
+
+``` text
+hive://user:password@ip:port/dbname[?auth=<auth_mechanism>&session.<cfg_key1>=<cfg_value1>...&session<cfg_keyN>=valueN]
+```
+
+In the above format,
+
+- `user:password` is the username and password of hiveserver2.
+- `ip:port` is the endpoint which the hiveserver2 instance listened on.
+- `dbname` is the default database name.
+- `auth_mechanism` is the authentication mechanism of hiveserver2, can be `NOSASL` for unsecurest transport or `PLAIN` for SASL transport.
+- parameters with prefix `session.` is the session confiuration of Hive Thrift API, such as `session.mapreduce_job_queuename=mr` implies `mapreduce.job.queuename=mr`.
+
+You can find more details at [gohive](https://sql-machine-learning.github.io/doc_index/gohive.html).
+
+Using the `datasource` string, you can launch an all-in-one Docker container by running:
+
+``` bash
+docker run --rm -p 8888:8888 sqlflow/sqlflow bash -c \
+"sqlflowserver --datasource='hive://root:root@localhost:10000/iris' &
+SQLFLOW_SERVER=localhost:50051 jupyter notebook --ip=0.0.0.0 --port=8888 --allow-root --NotebookApp.token=''"
+```
+
+Then you can open a web browser and go to `localhost:8888`. There are many SQLFlow tutorials, e.g. `tutorial_dnn_iris.ipynb`. You can follow the tutorials and substitute the data for your own use.
+
+## Connect standalone Hive server for testing
+
+We also pack a standalone Hive server Docker image for testing.
+
+### Connect Hive server with NOSASL Transport
+
+Launch your standalone hive server Docker container by running:
+
+``` bash
+> docker run -d -p 8888:8888 --name=hive sqlflow/gohive:dev
+```
+
+This implies settings in `hive-site.xml`:
+
+``` text
+hive.server2.authentication=NOSASL
+```
+
+Test SQLFlow by running the tutorials in Jupyter Notebook:
+
+``` bash
+> docker run --rm --net=container:hive sqlflow/sqlflow \
+bash -c "sqlflowserver --datasource='hive://root:root@localhost:10000/' &
+SQLFLOW_SERVER=localhost:50051 jupyter notebook --ip=0.0.0.0 --port=8888 --allow-root --NotebookApp.token=''"
+```
+
+## Connect Hive Server with PLAIN SASL Transport
+
+This section would use the [PAM](https://cwiki.apache.org/confluence/display/Hive/Setting+Up+HiveServer2#SettingUpHiveServer2-PluggableAuthenticationModules(PAM)) authentication to do the demonstration.
+
+Launch your standalone hive server Docker container with enable the PAM authentication:
+
+``` bash
+> docker run -d -e WITH_HS2_PAM_AUTH=ON -p 8888:8888 --name=hive sqlflow/gohive:dev
+```
+
+This implies settings in `hive-site.xml`:
+
+``` text
+hive.server2.authentication=PAM
+hive.server2.authentication.pam.services=login,sshd
+```
+
+Test SQLFlow by running the tutorials in Jupyter Notebook:
+
+``` bash
+> docker run --rm --net=container:hive sqlflow/sqlflow \
+bash -c "sqlflowserver --datasource='hive://sqlflow:sqlflow@localhost:10000/?auth=PLAIN' &
+SQLFLOW_SERVER=localhost:50051 jupyter notebook --ip=0.0.0.0 --port=8888 --allow-root --NotebookApp.token=''"
+```