diff --git a/tools/aws_benchmarking/README.md b/tools/aws_benchmarking/README.md
index 5fd586cc1537eabe3c45ec9c2a364e53c3e2d000..dfa2a5f478400fb80b438ddb1ebd0ee613671b0f 100644
--- a/tools/aws_benchmarking/README.md
+++ b/tools/aws_benchmarking/README.md
@@ -50,7 +50,10 @@ Training nodes will run your `ENTRYPOINT` script with the following environment
  - `TASK_NAME`: unique name to identify this training process.
  - `TRAINING_ROLE`: current node's role in this training process, either "PSERVER" or "TRAINER"
  - `PSERVER_HOSTS`: comma separated value of pserver end points, I.E. "192.168.1.2:5436,192.168.1.3:5436"
- - `TRAINER_INDEX`: an integer to identify the index of current trainer
+ - `PSERVERS`: same as above
+ - `TRAINERS`: trainer count
+ - `SERVER_ENDPOINT`: current server end point if the node role is a pserver
+ - `TRAINER_INDEX`: an integer to identify the index of current trainer if the node role is a trainer.
 
  Now we have a working distributed training script which takes advantage of node environment variables and docker file to generate the training image. Run the following command:
 
@@ -73,11 +76,15 @@ Training nodes will run your `ENTRYPOINT` script with the following environment
 Now let's start the training process:
 
 ```bash
-docker run -i -v $HOME/.aws:/root/.aws -v <full path to your pem file>:/<key pare name>.pem \
+docker run -i -v $HOME/.aws:/root/.aws -v <full path to your pem file>:/root/<key pare name>.pem \
 putcn/paddle_aws_client \
 --action create \
 --key_name <your key pare name> \
---security_group_id <your security group id>
+--security_group_id <your security group id> \
+--pserver_image_id <your pserver image id> \
+--trainer_image_id <your trainer images id> \
+--pserver_count 2 \
+--trainer_count 2
 ```
 
 Now just wait until you see this:
@@ -91,7 +98,7 @@ That means you can turn off your laptop and your cluster is creating instances,
 To access the master log:
 
 ```bash
-docker run -i -v $HOME/.aws:/root/.aws -v <full path to your pem file>:/<key pare name>.pem \
+docker run -i -v $HOME/.aws:/root/.aws \
 putcn/paddle_aws_client \
 --action status \
 --master_server_public_ip <master ip> \
@@ -101,7 +108,7 @@ putcn/paddle_aws_client \
 To tear down the training setup:
 
 ```bash
-docker run -i -v $HOME/.aws:/root/.aws -v <full path to your pem file>:/<key pare name>.pem \
+docker run -i -v $HOME/.aws:/root/.aws \
 putcn/paddle_aws_client \
 --action cleanup \
 --master_server_public_ip <master ip> \
@@ -111,7 +118,7 @@ putcn/paddle_aws_client \
 To retrieve training logs
 TBD
 
-### Tech details
+### Tech details
 
 *What to expect in this step*
 
diff --git a/tools/aws_benchmarking/diagram.png b/tools/aws_benchmarking/diagram.png
index 9dd656c9b4719fc6a96eb3d68c796daa0aaf7b98..b97909c5fe78b59d0e636ff73c2ed3e63a0be722 100644
Binary files a/tools/aws_benchmarking/diagram.png and b/tools/aws_benchmarking/diagram.png differ
diff --git a/tools/aws_benchmarking/server/logs/master.log b/tools/aws_benchmarking/server/logs/master.log
new file mode 100644
index 0000000000000000000000000000000000000000..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391
diff --git a/tools/aws_benchmarking/server/pserver.sh.template b/tools/aws_benchmarking/server/pserver.sh.template
index fe2360ed2063753805e86114857ede9be66be705..5e46a4246f1a31261654dc52d4ab30ccd89f1957 100644
--- a/tools/aws_benchmarking/server/pserver.sh.template
+++ b/tools/aws_benchmarking/server/pserver.sh.template
@@ -1,2 +1,2 @@
 #!/bin/bash
-nvidia-docker run -i -p {PSERVER_PORT}:{PSERVER_PORT} -e "MASTER_ENDPOINT={MASTER_ENDPOINT}" -e "TASK_NAME={TASK_NAME}" -e "TRAINING_ROLE=PSERVER"  -e "PSERVER_HOSTS={PSERVER_HOSTS}" {DOCKER_IMAGE}
\ No newline at end of file
+nvidia-docker run -i -p {PSERVER_PORT}:{PSERVER_PORT} -e "SERVER_ENDPOINT={SERVER_ENDPOINT}" -e "MASTER_ENDPOINT={MASTER_ENDPOINT}" -e "TASK_NAME={TASK_NAME}" -e "TRAINING_ROLE=PSERVER" -e "TRAINERS={TRAINER_COUNT}" -e "PSERVER_HOSTS={PSERVER_HOSTS}" -e "PSERVERS={PSERVER_HOSTS}" {DOCKER_IMAGE} {COMMAND}
\ No newline at end of file
diff --git a/tools/aws_benchmarking/server/trainer.sh.template b/tools/aws_benchmarking/server/trainer.sh.template
index 89f405811e768f2d8d4b74e75f818b4eb5df160d..56405a8e31d0c89c90168ce5126afc5b3206da0f 100644
--- a/tools/aws_benchmarking/server/trainer.sh.template
+++ b/tools/aws_benchmarking/server/trainer.sh.template
@@ -1,2 +1,2 @@
 #!/bin/bash 
-nvidia-docker run -i  -e "MASTER_ENDPOINT={MASTER_ENDPOINT}" -e "TASK_NAME={TASK_NAME}" -e "TRAINER_INDEX={TRAINER_INDEX}" -e "TRAINING_ROLE=TRAINER"  -e "PSERVER_HOSTS={PSERVER_HOSTS}" {DOCKER_IMAGE}
\ No newline at end of file
+nvidia-docker run -i  -e "MASTER_ENDPOINT={MASTER_ENDPOINT}" -e "TASK_NAME={TASK_NAME}" -e "TRAINER_COUNT={TRAINER_COUNT}" -e "TRAINER_INDEX={TRAINER_INDEX}" -e "TRAINING_ROLE=TRAINER"  -e "PSERVER_HOSTS={PSERVER_HOSTS}" {DOCKER_IMAGE} {COMMAND}
\ No newline at end of file