Jedox GPU Accelerator on Azure

image_pdfimage_print

Jedox GPU Accelerator can be tested on Azure NC instances in combination with the freely available Jedox for Azure BYOL virtual machine image (currently providing Jedox 7.0 SR2 on CentOS 7.3). The following instructions will assist you in setting up the virtual machine, installing the GPU driver, and enabling the Jedox GPU Accelerator.

Jedox for Azure BYOL virtual machine image

The BYOL (Bring Your Own License) version is a pre-configured Jedox installation. It requires a valid license for Jedox.

Note that while the Jedox for Azure BYOL virtual machine image is provided for free, running a virtual machine in Azure will be charged by the host. Pricing details will be displayed in the Azure portal when selecting the actual Azure instance.

Setting up the virtual machine

  1. In the Azure portal, create a virtual machine and search for “Jedox” in the marketplace. Select the image Jedox for Azure 7.0 SR2 BYOL and continue.
  2. Select the HDD option in the next step (SSD hard disks are not yet supported for NC instances) and choose any of the currently supported regions, which include the United States (EAST US, NORTH CENTRAL US, SOUTH CENTRAL US, WEST US 2), Europe (WEST EUROPE, NORTH EUROPE, UK SOUTH) and Australia (AUSTRALIA EAST). Note: this is the location of the host machine, which does not necessarily match your location. More regions are likely to be added soon. The official information site can be found here.
  3. Click on “Show all” in the instance selection panel to view the full list of available instances. Select the NC instance that best fits your needs with respect to RAM, CPU cores, and GPUs; e.g. NC6 or NC12. Pricing details for each instance can be found at the bottom of its feature panel. If Azure NC instances are listed but grayed out (i.e., not able to be selected), check the details of your Azure subscription. For example, some instances might not be available via MSDN or MPN subscriptions.
  4. After some deployment time, the machine will be ready to use.

Installing the GPU driver

When connected to the virtual machine via ssh client (e.g. putty), the following steps need to be performed:

  1. Install system updates

sudo yum update

  1. Install the latest Linux Integration Services for Hyper-V

wget http://download.microsoft.com/download/6/8/F/68FE11B8-FAA4-4F8D-8C7D-74DA7F2CFC8C/lis-rpms-4.2.3-2.tar.gz
tar xvzf lis-rpms-4.2.3-2.tar.gz
cd LISISO
sudo ./install.sh
sudo reboot

  1. Install kernel sources of current kernel and GCC (c/c++)

sudo yum -y install kernel-devel-$(uname -r) gcc-c++

  1. Download NVIDIA driver for Tesla K80 (currently version 384.66)

wget http://us.download.nvidia.com/XFree86/Linux-x86_64/384.66/NVIDIA-Linux-x86_64-384.66.run

  1. Start runfile installation and pass kernel source path

sudo ./NVIDIA-Linux-x86_64-384.66.run --kernel-source-path=/usr/src/kernels/$(uname -r) --silent

  1. Append the following option to Jedox OLAP server configuration file (/opt/jedox/ps/Data/palo.ini):

enable-gpu

Configuring Jedox on Azure

  1. To access Jedox from your local web browser, open up port 80 on the VM for your local machine’s IP via Azure Administration.
  2. Start Jedox Suite and activate your license, as described in License Activation in Jedox Web.
  3. Change the Jedox admin user password and deactivate default users, if not needed.

Getting started with Jedox GPU Accelerator

To activate GPU acceleration for a specific cube:

  1. Open the Modeler and select the cube
  2. Open the “Advanced” panel, and check the box for “Activate GPU acceleration”.

For more information, see Using the Jedox GPU Accelerator.

image_pdfimage_print
Was this post helpful?
NoYes (+1 rating, 1 votes)
Loading...