null
Enabling discovery

OpenStack private cloud enable HPC facilities to rapidly and flexibly meet varying user demand. However, the boundary-stretching and stringent requirements of HPC present challenges in operating a private cloud.

FishOS overcomes these challenges and has been proven with major operators in HPC, enabling higher efficiency and lower OpEx.

Challenges

  • Traditional HPC environments restrict users’ operating environment, and present a high adoption barrier for new users
  • Hardware faults impact long running jobs, forcing users to either restart the computations, or implement laborious and error-prone check-pointing process
  • Long running jobs have to be interrupted to allow urgent jobs to run, or require complex error-prone process for job preemption
  • With resources being managed statically based on reservation, HPC facility operators face wasted resources and users face longer delay for computation results
  • HPC facility operators can only react to reservation abuse and misuse through long consultative process

Solution

  • FishOS enable HPC facilities to meet wide breadth of user operating requirements, without impacting other users — users have their private environment
  • FishOS enable dynamic fault-handling, preserving states of long running computation jobs, without laborious and error-prone checkpointing process
  • FishOS enables flexible, reliable job preemption, by dynamically managing computational resources
  • FishOS enable higher throughput, by best matching dynamic computational resource requirements to available resources
  • FishOS eliminates reservation abuse and misuse by managing resources based on live utilization

Finding out what we can do for you.