Building a high performance computing cluster for research university purposes is no insignificant task. There is much to be considered when deciding how to build out the right university HPC cluster for your important work. But how do you weigh specific factors against each other and determine what to focus on?
5 Things to Consider when Building an HPC Cluster for University Research
- Determine the community that is going to utilize the resource.
Is the resource going to be used by an individual? A research group with a number of users? A specific department? A campus wide system? This is important to note from the get-go, because the larger the network of the system, the more complex the conversations need to be in designing it. If multiple departments are going to be relying on the system, there will be several inputs to consider in determining which capabilities the system will require.
- Determine how you’re going to get funding for the cluster.
Is your university HPC system going to be funded by your university, NSF grants, start up packages, or something different? How will the resources be financed? This is significant in ensuring that the financier receives adequate resources in return.
- Begin to design your cluster based upon the two first steps while maintaining a keen focus on potential growth in the future.
Keeping in mind potential future growth is critical to ensure that the purchased system will be useful 5 years down the road. For example, with a bandwidth limited network, other components added later may not provide anticipated performance improvements and will just bottleneck your work.
- Continue exploring new technologies.
It’s always a good idea to keep an eye on new advancements in components that are coming to market. Next generation GPUs are developed regularly and tend to be a significant focus for research universities when building an HPC cluster. Advancements in parallel file systems are also important to identify because they allow universities an expandable storage system.
- Determine who will administer the cluster and be responsible for operation and maintenance.
Will your university HPC cluster be managed by a dedicated resource? Will it be department personnel or even a student? You’ll want to make sure the person managing the system is well-versed in the technology and hardware. Student labor, while affordable and temporarily convenient, is not always the best option, as they may not have the necessary knowledge and need to be replaced when they leave the university.
As you can see, there is definitely a lot to consider when building out a cluster for research university purposes. That’s why it’s so important to pick the right vendor to partner with, one that will listen to what your needs are and create a system to match. PSSC Labs does just that, working closing with you throughout the entire process, ultimately delivering a production-ready system, regardless of time and budgetary constraints.
For more information on how we help research universities acquire the research instruments they need, visit us at: https://pssclabs.com/industries/higher-education/