Callout Subsystem Scalability

The callout subsystem is redesigned to include the following features:

• Performance and scalability improvements:

  • Per-CPU data structures to minimize mutex contention

  • Per-CPU callout processing to improve scalability

  • Event-based implementation that avoids polling overhead

• High-resolution timers for improved functionality. Many API calls use high-resolution timers and do not experience latency because the system rounds off the specified intervals. These timers include commonly used calls such as poll() and nanosleep().

• Observability improvements:

  • Comprehensive set of options for the MDB dcmd callout

  • New MDB dcmd calloutid

  • New callout kstats