A Flexible Scheduling Framework Supporting Multiple Programming Models with Arbitrary Semantics in Linux∗

We present a hierarchic scheduling framework for Linux called Group Scheduling that facilitates the creation of arbitrary thread schedulers. Traditional approaches to developing new scheduling semantics require semantic mappings onto existing schedulers, such as static-priority. Group Scheduling allows for a direct implementation of semantics, allowing clear mappings at any level a developer desires. In order to effectively support scheduling semantics, integration with concurrency control is necessary (e.g. priority inheritance). However, when considering arbitrary scheduling semantics hard-wired solutions such as PI can’t adapt. We present Proxy Execution as a general mechanism to resolving policy conflicts that arise as tasks from different scheduling domains interact through the RT-Mutex primitive.