diff --git a/doc/design/cluster_train/data_dispatch.md b/doc/design/cluster_train/data_dispatch.md
index 39fbe55c26a61fb6b934e05b1a3b7d99b874787f..c82e7b558ed3d9571ddd5992e73811fb3fbaa205 100644
--- a/doc/design/cluster_train/data_dispatch.md
+++ b/doc/design/cluster_train/data_dispatch.md
@@ -101,7 +101,7 @@ PaddlePaddle提供专用的[data reader creator](https://github.com/PaddlePaddle
 
 ```python
 # ...
-reader = paddle.reader.creator.RecordIO("/home/user_name/random_images-*-of-*")
+reader = paddle.reader.creator.RecordIO("/pfs/datacenter_name/home/user_name/random_images-*-of-*")
 batch_reader = paddle.batch(paddle.dataset.mnist.train(), 128)
 trainer.train(batch_reader, ...)
 ```
@@ -150,7 +150,7 @@ endpoint=datacenter2.paddlepaddle.org
 不用mount的方式来访问数据，而是直接用API的接口远程访问
 
 ```
-f = open('/pfs/datacenter/home/user/test1.dat')
+f = open('/pfs/datacenter_name/home/user_name/test1.dat')
 ```
 
 
diff --git a/doc/design/cluster_train/src/file_storage.graffle b/doc/design/cluster_train/src/file_storage.graffle
index 95ad2758ccae252dd322c497e7b167135cf478a8..5331f407f7ee77fca263be24eb83ed1d44ca2bd4 100644
Binary files a/doc/design/cluster_train/src/file_storage.graffle and b/doc/design/cluster_train/src/file_storage.graffle differ
diff --git a/doc/design/cluster_train/src/file_storage.png b/doc/design/cluster_train/src/file_storage.png
index b744bed48551b4361171bf4502e24cdff69a68fb..d88af97e034aae6fc09a728619e2a56fe8481f80 100644
Binary files a/doc/design/cluster_train/src/file_storage.png and b/doc/design/cluster_train/src/file_storage.png differ