SPIN Architecture Overview - Where are Strands Used ?

As is shown in the picture below:

 
Figure:   SPIN system decomposition. Arrows indicate strand related communication.

A shrink wrapped SPIN system has the following components:

• a trusted core that provides interfaces that manipulate bare hardware.
• a trusted event dispatcher.
• a trusted global scheduler that makes global CPU resource usage decisions.
• a trusted kernel-strand scheduler that is responsible for executing spindles that the dispatcher has identified as runnable.
• subsystems for memory management, IPC, networking, etc.
• possibly servers that use these subsystems to support a cohesive API (e.g. Unix).

All user level threads are strands that are scheduled by some strand scheduler. Further, there exists a kernel strand scheduler which runs the spindles marked by the dispatcher as ready to run. Hence, executing spindles conform to the strand interface and are scheduled just like every other executable entity in the system. That is, there is no inherent dichotomy between user-level strands and spindles except for the level of trust granted to them. This is intented to facilitate the migration of code from user-level into kernel space.