OIT is working to improve connection reliability to the HPC cluster.
Starting in Fall 2020 all connections to the HPC cluster will be balanced through a cluster of load balancers.
The load balancers as the name imply balancer user requests for SSH and Remote Desktop connections while also automatically detecting failed login nodes improving reliability.
Impacts of load balancers:
- 6-hour connection timeout. Idle connections will timeout after 6 hours.
- The balancers should detect unresponsive login nodes and direct users to working nodes.
- The balancers should remember the last login node a user connected to and direct them to the same one.