Andromeda–> Unity HPC Transition Plan

Over the last few years, we have built an entirely new HPC/AI environment, support system and infrastructure in partnership with UMass Amherst called UNITY: https://unity.uri.edu. We have received very positive feedback on Unity from all types of HPC/AI users at URI, and just performed a major upgrade to the system increasing capacity for URI users by ~50%. Unity is located at the MGHPCC, which is a cutting-edge near-zero Carbon facility specially designed for research computing by major research universities in our region. 

Andromeda is a much older computational resource sited at URI. It has an aging storage system, obsolete security infrastructure and consumes ~18 kW of energy that results in ~40 metric tons of CO2 in the atmosphere per year. Moreover, our small support team at URI has difficulty with supporting 2 different major computational systems at multiple locations, especially given that we are in the process of building 2 additional systems (Sanctuary and Harmony) at MGHPCC.

For these reasons, we are considering transitioning URI users away from Andromeda to Unity over the next few months. Once the transition is complete, we will relocate viable Andromeda hardware and merge it into Unity so that there is minimal loss of compute capacity for URI. 

Here are the details of this transition plan: 

• Our team will offer focussed support to Andromeda users to aid their transition to Unity over the next few months. 
• Research data transfer to Unity would be the top priority right from the beginning. Since some data transfers may take a long time, we will not set a deadline for moving data off the NAS storage systems attached to Andromeda.  
• After May 30th, we will shut down the Andromeda compute nodes thus halting the execution of any future jobs. And we will change the filesystem to read-only mode. Users would still be able to migrate their files over to Unity.
• We will schedule a full shutdown of Andromeda at URI on July 1st. We will package and ship the viable Andromeda hardware to the MGHPCC and have it integrated into Unity and make it available by September 1st. 

Please don’t hesitate to reach out with questions or concerns.

Received a wonderful note from Dr. Ginis: Pranav Sai received the award for the best abstract at PEARC24 held in Providence last summer. He presented RICHAMP and its implementation on URI nodes at MGHPCC for real-time storm forecasting.