New publication: “Adaptive request scheduling for the I/O forwarding layer using reinforcement learning”
In this paper, we propose an approach to adapt the I/O forwarding layer of HPC systems to applications’ access patterns. We propose to do it dynamically at runtime, using a reinforcement learning technique to observe the I/O workload of the system and determine the best parameter value for each observed access pattern during its execution.
GekkoFS 0.7.0 is now available, with enhancements to the filesystem configuration options, more detailed log output, and several bugfixes.
The GekkoFS file system jointly developed by BSC’s SSEC and JGU’s ECS teams takes the number 4 spot in the IO500’s ‘10-Node Challenge’.
Version 0.6.2 of GekkoFS has been released with new configuration options, improvements to the log subsystem, bug fixes and code quality enhancements.