The Partition Mapping and Multiplexing Layer (or Mapping Layer), translates the resource specifications of the
Policy Service (expressed in the STRG) into an ordered
sequence of spatial partitions for the underlying Partition
Mechanism Layer (mentioned in Section 2). The Mapping Layer makes no policy decisions, but rather implements the policy decisions given by the Policy Service.
The Mapping Layer comprises two main components:
the Planner and the Plan Executor. When the Planner
receives a new STRG from the Policy Service, it first
validates that this STRG does not violate basic security
or QoS requirements
4
, then generates a future plan for
distributing resources to Cells. The Planner invokes an
operation similar to bin-packing to assign Cells and resources to future partition time-slices.
The Plan Executor implements the resulting resource
plan. It can modify the plan being executed in predefined
ways to accommodate more dynamic resource-allocation
and time-multiplexing actions (e.g., activation of a Cell
upon the arrival of an event or redistribution of excess
resources among Cells).
In implementing the STRG, the Mapping Layer implements a variety of Cell activation policies. Examples
include the Pinned Policy (Cell given dedicated access to
cores), the Time-Triggered Policy (Cell active during predetermined time-windows for real-time predictability),
and the Time-fraction Policy (Cell active for a specified
fraction of the time). Most Cell activation policies are
non-preemptive: once a Cell is activated it is not suspended until its time-slice expires. The one exception is
that Cells can be given best-effort resources that may be
preempted by Cells with higher priority.
The Partition Mapping and Multiplexing Layer (or Mapping Layer), translates the resource specifications of the
Policy Service (expressed in the STRG) into an ordered
sequence of spatial partitions for the underlying Partition
Mechanism Layer (mentioned in Section 2). The Mapping Layer makes no policy decisions, but rather implements the policy decisions given by the Policy Service.
The Mapping Layer comprises two main components:
the Planner and the Plan Executor. When the Planner
receives a new STRG from the Policy Service, it first
validates that this STRG does not violate basic security
or QoS requirements
4
, then generates a future plan for
distributing resources to Cells. The Planner invokes an
operation similar to bin-packing to assign Cells and resources to future partition time-slices.
The Plan Executor implements the resulting resource
plan. It can modify the plan being executed in predefined
ways to accommodate more dynamic resource-allocation
and time-multiplexing actions (e.g., activation of a Cell
upon the arrival of an event or redistribution of excess
resources among Cells).
In implementing the STRG, the Mapping Layer implements a variety of Cell activation policies. Examples
include the Pinned Policy (Cell given dedicated access to
cores), the Time-Triggered Policy (Cell active during predetermined time-windows for real-time predictability),
and the Time-fraction Policy (Cell active for a specified
fraction of the time). Most Cell activation policies are
non-preemptive: once a Cell is activated it is not suspended until its time-slice expires. The one exception is
that Cells can be given best-effort resources that may be
preempted by Cells with higher priority.
การแปล กรุณารอสักครู่..