diff --git a/sdgx/utils.py b/sdgx/utils.py
index 4e8ba504..b8799ce6 100644
--- a/sdgx/utils.py
+++ b/sdgx/utils.py
@@ -18,7 +18,27 @@
 except ImportError:
     from functools import lru_cache as cache
 
-__all__ = ["download_demo_data", "get_demo_single_table", "cache", "Singleton", "find_free_port"]
+__all__ = [
+    "download_demo_data",
+    "get_demo_single_table",
+    "cache",
+    "Singleton",
+    "find_free_port",
+    "download_multi_table_demo_data",
+    "get_demo_single_table",
+]
+
+MULTI_TABLE_DEMO_DATA = {
+    "rossman": {
+        "parent_table": "store",
+        "child_table": "train",
+        "parent_url": "https://raw.githubusercontent.com/juniorcl/rossman-store-sales/main/databases/store.csv",
+        "child_url": "https://raw.githubusercontent.com/juniorcl/rossman-store-sales/main/databases/train.csv",
+        "parent_primary_keys": ["Store"],
+        "child_primary_keys": ["Store", "Date"],
+        "foreign_keys": ["Store"],
+    }
+}
 
 
 def find_free_port():
@@ -95,6 +115,72 @@ def __call__(cls, *args, **kwargs):
         return cls._instances[cls]
 
 
+def download_multi_table_demo_data(
+    data_dir: str | Path = "./dataset", dataset_name="rossman"
+) -> dict[str, Path]:
+    """
+    Download multi-table demo data "Rossman Store Sales" or "Rossmann Store Sales" if not exist
+
+    Args:
+        data_dir(str | Path): data directory
+
+    Returns:
+        dict[str, pathlib.Path]: dict, the key is table name, value is demo data path
+    """
+    demo_data_info = MULTI_TABLE_DEMO_DATA[dataset_name]
+    data_dir = Path(data_dir).expanduser().resolve()
+    parent_file_name = dataset_name + "_" + demo_data_info["parent_table"] + ".csv"
+    child_file_name = dataset_name + "_" + demo_data_info["child_table"] + ".csv"
+    demo_data_path_parent = data_dir / parent_file_name
+    demo_data_path_child = data_dir / child_file_name
+    # For now, I think it's OK to hardcode the URL for each dataset
+    # In the future we can consider using our own S3 Bucket or providing more data sets through sdg.idslab.io.
+    if not demo_data_path_parent.exists():
+        # make dir
+        demo_data_path_parent.parent.mkdir(parents=True, exist_ok=True)
+        # download parent table from github link
+        logger.info("Downloading parent table from github to {}".format(demo_data_path_parent))
+        parent_url = demo_data_info["parent_url"]
+        urllib.request.urlretrieve(parent_url, demo_data_path_parent)
+    # then child table
+    if not demo_data_path_child.exists():
+        # make dir
+        demo_data_path_child.parent.mkdir(parents=True, exist_ok=True)
+        # download child table from github link
+        logger.info("Downloading child table from github to {}".format(demo_data_path_child))
+        parent_url = demo_data_info["child_url"]
+        urllib.request.urlretrieve(parent_url, demo_data_path_child)
+
+    return {
+        demo_data_info["parent_table"]: demo_data_path_parent,
+        demo_data_info["child_table"]: demo_data_path_child,
+    }
+
+
+def get_demo_multi_table(
+    data_dir: str | Path = "./dataset", dataset_name="rossman"
+) -> dict[str, pd.DataFrame]:
+    """
+    Get multi-table demo data as DataFrame and relationship
+
+    Args:
+        data_dir(str | Path): data directory
+
+    Returns:
+        dict[str, pd.DataFrame]: multi-table data dict, the key is table name, value is DataFrame.
+    """
+    multi_table_dict = {}
+    # download if not exist
+    demo_data_dict = download_multi_table_demo_data(data_dir, dataset_name)
+    # read Data from path
+    for table_name in demo_data_dict.keys():
+        each_path = demo_data_dict[table_name]
+        pd_obj = pd.read_csv(each_path)
+        multi_table_dict[table_name] = pd_obj
+
+    return multi_table_dict
+
+
 def ignore_warnings(category: Warning):
     def ignore_warnings_decorator(func: Callable):
         @functools.wraps(func)
diff --git a/tests/conftest.py b/tests/conftest.py
index ee1b3a00..8f1635f7 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -13,7 +13,7 @@
 from sdgx.data_connectors.csv_connector import CsvConnector
 from sdgx.data_loader import DataLoader
 from sdgx.data_models.metadata import Metadata
-from sdgx.utils import download_demo_data
+from sdgx.utils import download_demo_data, download_multi_table_demo_data
 
 _HERE = os.path.dirname(__file__)
 
@@ -132,3 +132,29 @@ def demo_single_table_data_loader(demo_single_table_data_connector, cacher_kwarg
 @pytest.fixture
 def demo_single_table_metadata(demo_single_table_data_loader):
     yield Metadata.from_dataloader(demo_single_table_data_loader)
+
+
+@pytest.fixture
+def demo_multi_table_path():
+    yield download_multi_table_demo_data(DATA_DIR)
+
+
+@pytest.fixture
+def demo_multi_table_data_connector(demo_multi_table_path):
+    connector_dict = {}
+    for each_table in demo_multi_table_path.keys():
+        each_path = demo_multi_table_path[each_table]
+        connector_dict[each_table] = CsvConnector(path=each_path)
+    yield connector_dict
+
+
+@pytest.fixture
+def demo_multi_table_data_loader(demo_multi_table_data_connector, cacher_kwargs):
+    loader_dict = {}
+    for each_table in demo_multi_table_data_connector.keys():
+        each_connector = demo_multi_table_data_connector[each_table]
+        each_d = DataLoader(each_connector, cacher_kwargs=cacher_kwargs)
+        loader_dict[each_table] = each_d
+    yield loader_dict
+    for each_table in demo_multi_table_data_connector.keys():
+        demo_multi_table_data_connector[each_table].finalize()