.. _ARA_Edge_Experiment:

ARA Edge Usage
---------------------------------------------

**Platform:** ARA edge node equipped with NVIDIA GPU.

**Resources Needed:** Any edge compute node in ARA.

**Description:** The experiment is designed to demonstrate the use of
dedicated edge node in ARA which is equipped with NVIDIA RTX 5000 GPU. 


**Detailed Steps**

#. Login to `ARA portal <https://portal.arawireless.org/>`_ with your
   username and password. 

   .. note:: If you are a first time user, we highly recommended you
		to try the :ref:`Hello World experiment
		<ARA_Hello_World>` experiment first to get
		familiarized with the portal and experiment workflow.

#. Create a reservation using the *Project -> Reservations ->
   Leases* tab from the dashboard with the following attributes:

      * *Site*: AgronomyFarm
      * *Resource Type*: Compute Node
      * *Device Type*: Edge 
      * *Device ID*: 000 

#. Launch a container using the information provided below. You can
   keep the remaining attributes default or set depending on your
   choice.

      
       * *Container Image*: ``arawirelesshub/ara-edge:base``
       * *CPU*: 8
       * *Memory*: 8192
       * *Network*: ARA_Shared_Net

#. Once the container is launched, take a note on the floating IP and
   SSH to the container via the ARA jumpbox. Detailed instructions for
   accessing the container via jumpbox can be found :ref:`here
   <ARA_Jumpbox>`. 

#. Check the NVIDIA GPU information using the following command.

   .. code-block:: console

      # nvidia-smi

   The command will provide the GPU information as shown below.

   .. image:: images/GPU_Info.png
      :align: center

#. In the container, we provide an example program **test_gpu.py** to
   compare the execution time of CPU and GPU in computing squares of
   random numbers in an array of size 100000000. Run the following
   steps to execute the program.

   .. code-block:: console

      # cd
      # python3 test_gpu.py

   The program gives the execution time of same task by CPU and GPU as
   follows: 

   .. image:: images/GPU_Time.png
      :align: center

.. note:: You can install the required libraries and use the GPU for
	  executing your own programs for data analytics at the edge
	  node using the container.