Cubeman Features

An overview of Cubeman features are listed below.

Features Benefits
Comprised of 100% open source components
  • Avoid restrictive license agreements and fees
  • Run on premises or in the cloud
  • Preserve flexibility
Compatible with mainstream workload scheduler
  • Leverage hundreds of existing integrations
  • Tap a large community of expertise
  • Preserve investments in existing scripts and expertise
Fair-share scheduling
  • Allocate resources fairly according to configurable policies
  • Ensure that resource requirements are met for critical projects
Job slot or resource pre-emption
  • Ensure critical jobs, users or groups get resources when they need them
  • Ensure service levels are met
  • Respect asset ownership
MPI support
  • Integration through blaunch improves the reliability of MPI jobs
  • Prevents runaway MPI jobs/tasks
  • Provides up to date information on MPI job/task status
  • Maintains compability with future MPI versions
General purpose GPU support
  • Ensures that GPU jobs get the resources that they need
  • Leverages Cubeman powerful features for flexibility and future compatibility
Docker support
  • Provides resource guarantees and application isolation
  • Fast management operations
  • Application encapsulation and cloud mobility
Cloud & VM friendly auto-scaling
  • Easily add or remove nodes on the fly
  • Minimize administrator effort when changing cluster configurations
  • Quickly tap idle resources
RPM packages
  • Pre-built binaries to accelerate and simplify the deployment of Cubeman
Commercial support available from Skycloud Software
  • World-class cluster management expertise
  • Support from the source - leaders in the development of open source workload management
  • Clusters deployed in accordance with best practices

In addition to the features above, Cubeman provides a wealth of technical capabilities including:
  • Job arrays - for DOEs or parametric sweeps
  • Queue dispatch and run windows
  • Application license scheduling with pre-emption support
  • Exclusive scheduling
  • Interactive job support
  • Job limits per user, per queue and per host
  • Job dependencies for creating multi-step workflows
  • Mix serial and parallel jobs with workload tiling options
  • Scheduling based on dynamic machine load
  • Dynamic master selection, n-way failover
  • Plug-in authentication to enterprise data sources
  • Flexible resource requirement syntax
  • Pre-execution, post-execution and job wrapper scripts
  • Customizable job submission criteria
  • Support for popular x86 64 bit Linux distributions including RHEL, CentOS, SUSE, Debian and Ubuntu