HPCaaS from Rescale provides a fully managed turnkey HPC solution that significantly reduces the time to migrate your HPC jobs to a cloud environment.
HPC-as-a-Service with Rescale is an IBM Cloud solution designed to simplify and streamline the many steps required for HPC (high performance computing) jobs, including infrastructure deployment, data input, workflow management, and accessing results. Using this service, clients can rapidly execute their HPC jobs on IBM Cloud with their choice of application, compute type, and number of compute nodes.
The Rescale ScaleX® is a SaaS-based platform on IBM Cloud that solves the challenges of implementing, allocating, and managing public cloud infrastructure resources for HPC. This new HPCaaS from Rescale on IBM Cloud offering features a turnkey administrative portal and the ability to manage and monitor consumption within your organization.
Utilize HPC infrastructure on IBM Cloud
- Match jobs to the right architecture based on cost, capability (e.g., CPU, GPU type), or a mix of price-performance
- Access a wide variety of the latest compute architectures
- Automatically grow or shrink your cloud environment based on demand
- Keep your data secure with end-to-end enterprise-grade security
There are numerous scenarios that could be right when looking to move a job to the IBM Cloud, including jobs which do the following:
- Allow your teams to fail fast and quickly iterate until the right design is found
- Have ad-hoc requirements that cannot be met by your current on-premises environment
- Require the use of specific GPUs or CPUs that are not locally available
- Would benefit from large scale and flexibility of the IBM Cloud.
You can significantly reduce the need to constantly manage the application software licenses and upgrades to on-premises infrastructure. This decreases your overall cost and can accelerate your organization’s performance.
The following are some of the use cases that could fit well with the HPC-as-a-Service with Rescale solution:
- Manufacturing (e.g., CFD, CAE, thermal and electromagnetic analysis)
- Automotive (e.g., optimize vehicle drag co-efficient, CAE, CFD, crash simulation)
- Aerospace (e.g., CAE, wing design simulation, structural dynamics/reliability)
- Transportation (e.g., structural dynamics/reliability)
- Weather (e.g., weather forecasting)
Heathcare and life sciences
- Genetic Engineering
- Drug Discovery
- Digital Pathology
Integrated HPC packages
The Rescale ScaleX® platform seamlessly matches packaged HPC software applications with best-suited compute types available on IBM Cloud. The following are the software packages integrated into HPCaaS from Rescale solution: LS-DYNA 10+, STAR-CCM+, CONVERGE, ANSYS suite, Abaqus, VPS, CFD++, MSC Nastran, OpenFOAM, NX Nastran, 6Sigma, NAMD, Rocky DEM, EDEM (DEM), WRF, and Tensorflow.
Note: A new package can be added if requested, or customers can use a bring-your-own-HPC software model as well.
The following section walks through the typical steps when using the HPCaaS from Rescale solution to execute an HPC job on IBM Cloud. The following steps can be followed to run any HPC job on the Rescale ScaleX® platform.
Step 1: Create a job
Once you are on the Rescale platform dashboard, please note the top navigation bar, which has all the controls you need to manage your HPC jobs. The important one is Jobs, where you can manage all the jobs (current or past). To start working with a job, click on New Job, specify a name and pick the kind of job it is.
There are three kinds of jobs to choose from the Job Type drop-down menu on the right-hand side: Basic, DOE (design of experiment), and Optimization.
Step 2: Import the data
Once you create a job, the next step is to provide the data and HPC application code that needs to be executed. You can either upload the data from your computer or choose an already existing set of files from cloud object storage instance previously uploaded. All the uploaded data is managed in your own private visibility access region within a cloud object storage instance. Your data and files cannot be accessed by other users unless you explicitly grant them permissions.
Once the files are uploaded, you can see them as the job artifacts view shown below.
Step 3: Select the software package
Once the data has been uploaded, you need to select the HPC package you need to run your job. This is the same list of Integrated Software Packages provided in the previous section.
Once you pick the software package, you need to specify the version of the package you would like to use, as shown below.
Next, update the rest of the settings and the prepopulated command line of how to execute your specific workload. For example, below is configuration for an Ansys Fluent job.
The platform supports multiple licensing methods, you can use on-demand licensing tailored for cloud computing, or you can use your existing license.
Below is an example configuration for customer provided license of an Ansys Fluent job.
Step 4: Select the hardware configuration
The next step is to choose the hardware that you need for your HPC job. This is where the platform really shines—it gives you access to the best architectures for your job. The platform already has the intelligence to filter out hardware configurations that might not be a good match for your selected HPC software package. For example, if it’s an MPI job, a 1 Gbps-based VSI configuration might not work.
You can adjust the cores as per your job requirements. You can also define a maximum time limit after which the job should be terminated, in case it’s stuck at some step.
At this point, your job definition is complete, and you can click the Submit button to trigger the job or you can save the job to submit it later. You can execute, monitor, and manage your job from anywhere, anytime using the platform.
Step 5: Analyze the results
While the job is running, you can monitor the progress on its status page. The platform verifies the inputs and automatically provisions your HPC cluster with your choice of compute type and HPC software package.
You can also monitor your job status in real-time by using the Live Tailing option.
Here, all the output files are available during the run, such as the log files, convergence output files, and plots to verify the results of the simulation.
For an output that needs a virtual desktop to visualize the results, you can easily do this by downloading the files on your local computer and then loading them into the desktop.
Step 6: Job dashboard
Once the job is completed, it shows up on your job dashboard and you can go back and inspect all the properties of the job or submit the same job again.
Get started with HPCaaS from Rescale
HPCaaS from Rescale is the only SaaS-based, fully managed HPC solution for enterprise big compute and HPC workloads on IBM Cloud. With a few easy steps, as explained above, you can help your customers migrate their HPC-related workloads to IBM Cloud in no time.
Given the full end-to-end automation, the POC and setup is almost zero or very small. Once you get a paid account on IBM Cloud, you can access the HPCaaS from Rescale tile from IBM Cloud catalog page and follow the instructions above to demo the capabilities or help your customers execute a sample workload.
For some sample workloads with specific HPC software packages, you can review the Rescale published list here under “Tutorials” and “Video Tutorials” section.
Learn more about HPCaaS from Rescale on IBM Cloud.