How do you run the Petrel software on AWS?

Lecture de 5 minute(s)
Niveau du contenu : Avancé

AWS prescriptive guidance based on how different customers have been running the Petrel software on AWS

Geoscientists and engineers that analyze subsurface data make use of specialized software for earth modeling. One of such applications is the Petrel software from SLB. Petrel is a software application that runs on Windows and can be deployed on AWS in multiple ways, depending on the use case and user preferences.

Disclaimer: this article does not contain official recommendations from the software vendor. It provides AWS prescriptive guidance based on how different customers have been running the application in AWS. For a list of deployment options currently supported by the vendor, please contact SLB. Petrel is a trademark of Schlumberger Technology Corporation

Option 1: Deployment on Amazon AppStream 2.0

Petrel can be deployed via Amazon Appstream 2.0. This option minimizes operational overhead and costs and provides a streaming application experience.

From a user point of view, it provides dual-monitor support on web browsers with a maximum display resolution of 2560x1440 pixels per monitor. When using the AppStream 2.0 client for Windows, the support extends to 4 monitors and allows for USB peripheral access.

The recommended instance families to use with this application are Graphics G5/G4, based on availability in the selected region.

Follow the tutorial Create a Custom AppStream 2.0 Image by Using the AppStream 2.0 Console and the vendor documentation to create an image that you can use to deploy the application.

Note: the maximum session duration that you can set using the AppStream 2.0 API and CLI is 120 hours.

Option 2: Deployment on Amazon Workspaces

Petrel can be deployed via Amazon Workspaces. This option provides a persistent Desktop experience for integrated Geoscience workflows that involve multiple applications, while lowering operational overhead and costs.

The recommended bundle to use with this application is the Windows Graphics.g4dn bundle. For extremely demanding use cases, usage of GraphicsPro.g4dn can be evaluated.

From a user point of view, it is worth noting that Graphics bundles support only a single monitor configuration with a maximum resolution of 2560x1600. USB redirection can be enabled following the documentation.

Note: some Petrel users many find having a single monitor a limitation.

Option 3: Deployment on Amazon EC2

Petrel can be deployed via Amazon EC2. This option can be useful in the following use cases:

  • users are in a location having high latency with the closest AWS region running AppStream/Workspaces. It has been noted that a latency > 50ms can affect the user experience for 3D applications in geoscience
  • users need a particular instance type that is not offered by AppStream/Workspaces
  • there is a desire to implement a Windows Failover Cluster setup

When running on a EC2, two options are available as far as instance image:

In both cases, install the Petrel application following the installation guide from the vendor.

For remote access of the application from the users, two options are available:

  • For the best user experience using graphic applications, AWS recommends to use NICE DCV:
  • As an alternative measure, you can enable GPU rendering on Microsoft Windows Remote Desktop (this is disabled by default) and launch the application with a regular RDP client. This approach is something that can be used for testing purposes as it does not allow users to benefit from the high level of compression and high resolution support of NICE DCV. The steps to enable GPU rendering are:
    • open the Windows Run dialog from the start menu
    • type gpedit.msc and then click OK to display the Local Group Policy Editor
    • using the navigation panel on the left side, select Administrative Templates > Windows Components > Remote Desktop Services > Remote Desktop Session Host > Remote Session Environment > RemoteFX for Windows Server
    • Under Setting, enable each of the following options: right-click each one, select Edit, and then click Enable
      • Configure RemoteFX
      • Optimize visual experience when using RemoteFX
      • Optimize visual experience for Remote Desktop Service Sessions
    • From the navigation panel on the left side, select Remote Session Environment
    • Under Setting, enable each of the following options: right-click each one, select Edit, and then click Enable
      • Use the hardware default graphics adapter for all Remote Desktop Services sessions
      • Configure compression for RemoteFX data
    • Restart the instance for the changes to take effect

Data storage

If access to Petrel project data is only needed from a Windows environment, a viable option is to utilize Amazon FSX for Windows File Server. This is compatible with all deployment options mentioned above and provides up to 20 GB/s of maximum throughput available to a single client accessing the file system. For many organizations, this is an ideal setup.

In other situations, there could be a need to make Petrel data available to a Linux environment on top of Windows. This could be for instance a Linux HPC cluster that executes Reservoir Simulation jobs against the same user data, using SLB eclrun tool, that's integrated with the most common HPC workload schedulers such as IBM LSF, Altair PBS and SchedMD SLURM. In this case, using Amazon FSx for NetApp ONTAP is a more appropriate option with some trade offs in terms of maximum throughput available.

Both storage services mentioned above are integrated with AWS Backup, providing automated and centrally managed backup and restore capabilities.

For more details about the tradeoffs of the two options, review the documentation about choosing an Amazon FSx File System.

profile pictureAWS
publié il y a 8 mois2246 vues
2 Commentaires

Thank you for publishing this article Alberto. This showcases our maturity in hosting Petrel on AWS with various services, and how it can be hosted by our customers leveraging AppStream, Workspaces, and EC2.

a répondu il y a 8 mois

Thanks for this article Alberto! Did you by any chance perform cost estimates for each of these deployment options?

a répondu il y a 5 mois