A hierarchical multiprocessor bandwidth reservation scheme with timing guarantees

Abstract

A multiprocessor scheduling scheme is presented for supporting hierarchical containers that encapsulate sporadic soft and hard real-time tasks. In this scheme, each container is allocated a specified bandwidth, which it uses to schedule its children (some of which may also be containers). This scheme is novel in that, with only soft real-time tasks, no utilization loss is incurred when provisioning containers, even in arbitrarily deep hierarchies. Presented experiments show that the proposed scheme performs well compared to conventional real-time scheduling techniques that do not provide container isolation.

Globally scheduled real-time multiprocessor systems with GPUs

“…It was shown in [22] that bandwidth reservations, or containers, may be used to support soft real-time guarantees in multiprocessor systems. In a container-based system, a task set is organized into a hierarchical collection of containers.…”

“…Containers provide temporal isolation by hierarchically allocating execution time to contained tasks and containers. If each container schedules its contained tasks and containers using a window-constrained scheduling algorithm, 5 such as the G-EDF, then bounded tardiness can be ensured with no utilization loss [22]. The CM exploits both this and the ability to apply different schedulers to subsets of jobs.…”

Compositional multiprocessor scheduling: the GMPR interface

“…This section illustrates three of them (Leontyev and Anderson 2008;Shin et al 2008;Bini et al 2009). The interfaces are ordered by their increasing complexity and, consequently, by increasing accuracy of the guarantee test for applications running over the interface.…”

“…The interfaces are ordered by their increasing complexity and, consequently, by increasing accuracy of the guarantee test for applications running over the interface. Leontyev and Anderson (2008) proposed to use only the overall bandwidth requirement w (using their original notation) as an interface for soft real-time tasks. Being Π Fig.…”

“…We refer the interface model of Leontyev and Anderson (2008) as the multiprocessor bandwidth interface (MBI) and denote it as where w is the interface bandwidth and Π is the server period. Initially designed for soft real-time tasks, the MBI model can easily be extended for hard real-time systems.…”