ERIS provides a range of computational resources, platforms and scientific computing support for research and innovation at Mass General Brigham hospitals. Our high-performance analysis servers, compute clusters and storage are relied upon daily for data processing and analysis by research groups across the organization. Clinical computational workflows such as genome sequencing and radiation dosimetry are also supported.
Service Model - What is the cost?
Shared resources are available to all users at no cost. This includes a basic storage quota, shared computational resources and assistance from the ERIS Support teams. Additional storage and compute capacity can be acquired using research funds.
Selecting a Computational Resource
When choosing a computational resource from our list of services below, consider
- Your preferred computing platform (Microsoft Windows, Linux or Hadoop)
- What the software application requirements are - what platforms does the software run on?
- If you will work interactively with applications and data, or submit many jobs together for batch-processing
- How large is the data you will be working with, what are the storage requirements?
- How much memory will the application require?
‼ For details on the HPC Remediation plan see the section below
Modernized, more reliable version of the ERISTwo HPC cluster that rebuilds the existing infrastructure using current-generation tools
MGB Digital is currently in the process of improving the ERISTwo Linux Cluster as part of our HPC remediation plan. The remediation project leverages the existing ERISTwo cluster hardware while implementing new operational and system-level processes designed to enhance performance, stability, and usability, building a more modern and sustainable HPC platform. This upgraded version of ERISTwo is what we’re calling ERIS Nucleus.
Approach
We are taking the existing ERISTwo and ERISXdl infrastructure and rebuilding it using modern HPC cluster management tools and methodologies. The project follows a phased approach:
- Phase 1: Establish foundational infrastructure and services
- Phase 2: Deploy ERISTwo Nucleus (a subset of existing cluster nodes) as a modern HPC cluster
- Phase 3: Integrate the remaining cluster nodes into ERISTwo Nucleus
This is a major upgrade that includes a new RHEL 9 operating system, an updated software toolbox, and an upgraded 100 GbE network fabric. These changes will require many existing workflows and scientific applications to be thoroughly tested, so your participation in testing is essential.
Why are we doing this?
The ERISTwo HPC cluster has been experiencing performance, consistency, and stability issues. Additionally, the ERISXdl GPU cluster’s Slurm/Kubernetes setup has proven complex for many AI/ML and other workloads.
What are some of the underlying reasons for these issues?
- The existing 40GbE network is underperforming and poorly configured.
- The ERISTwo cluster still relies on legacy ERISOne services and processes, making day-to-day operations fragile.
- The cluster is manually managed and configured, increasing the risk of configuration drift and instability.
- Cluster services lack resilience, creating additional risks of instability.
High Performance Computing System with a job scheduler for batch jobs, storage and remote desktops with GPUs for graphical applications
The ERIS Linux Cluster is an ecosystem of scientific computing resources centered around a cluster of remote-desktop and compute nodes connected to very high speed storage. A large selection of popular scientific applications are installed and you can request additional software packages to be added. The cluster runs a Linux operating system and requires some familiarity with Linux for efficient use.
This platform is ideal for workflows that run many jobs in parallel, and for those that read and write many files or require very high speed access to data files. A job scheduling system queues jobs for dispatch to the compute nodes, allowing submission of many jobs at once. Linux remote-desktop nodes allow graphical applications for data visualization to interacting with data stored on the cluster, as well as software development and application testing. Some research groups also dispatch analysis pipeline jobs to the cluster through custom web-portals.
Typical Uses
- Medical image processing
- Genome sequencing
- Monte-carlo modeling
- MPI parallel workloads
- Very large memory jobs
- AI/ML and Deep Learning using NVIDIA GPUs
Supported methods of connecting to the cluster
- SSH command line terminal for job submission
- NoMachine Linux remote desktops for graphical applications
- Network file share (SMB/CIFS) for data transfer
- Web portals and applications (R Studio, Jupyter, etc)
Getting an account?
- Use the registration link
Application Availability on ERISTwo
Easily search the thousands of applications available on ERIS Scientific Computing resources today.
ERISTwo account required to access.
ERISXdl (ERIS Extreme Deep Learning) platform provides efficient, multi-GPU performance designed for Deep Learning applications
A highly secure, privacy-aware, data ecosystem equipped with self-service AI, machine learning, and research data tools.
Powerful servers with scientific applications installed and remote desktop capability, running on a Microsoft Windows OS