Mixed-Criticality Scheduling with I/O

Abstract

Abstract-This paper addresses the problem of scheduling tasks with different criticality levels in the presence of I/O requests. In mixed-criticality scheduling, higher criticality tasks are given precedence over those of lower criticality when it is impossible to guarantee the schedulability of all tasks. While mixed-criticality scheduling has gained attention in recent years, most approaches typically assume a periodic task model. This assumption does not always hold in practice, especially for realtime and embedded systems that perform I/O. In prior work, we developed a scheduling technique in the Quest real-time operating system, which integrates the time-budgeted management of I/O operations with Sporadic Server scheduling of tasks. This paper extends our previous scheduling approach with support for mixed-criticality tasks and I/O requests on the same processing core. Results show that in a real implementation the mixedcriticality scheduling method introduced in this paper outperforms a scheduling approach consisting of only Sporadic Servers.

Tuned Pipes: End-to-End Throughput and Delay Guarantees for USB Devices

“…PIBS uses a single replenishment to avoid fragmentation of replenishment list budgets caused by short-lived interrupt bottom half service routines. By using PIBS for interrupt threads, the scheduling overheads from context switching and timer reprogramming are reduced [7]. In prior work [2], we showed that for n Main VCPUs and m I/O VCPUs running on a single physical CPU, temporal isolation is guaranteed amongst all VCPUs if:…”

An I/O Virtualization Framework With I/O-Related Memory Contention Control for Real-Time Systems

“…The ones that are for the exclusive use of a VM can be accessed in pass-through and treated as simpler cases of the shared ones (details are provided in the following). Similarly to other proposals (e.g., hypervisors such as Xen [13], or other mechanisms proposed in scientific papers [16]- [20]), in our architecture (shown in Figure 1) all the shared physical devices are exclusively assigned to a single VM by the hypervisor, called I/O-management VM (I/O VM in short), following a pass-through approach. The I/O VM implements a softwaredefined sharing mechanism using shared-memory buffers and an I/O scheduler to dispatch the accesses to each I/O device.…”

MCS-IOV: Real-Time I/O Virtualization for Mixed-Criticality Systems

“…VMs). MultiPARTES [69] and Airbus' MP-IOV [55] used para-virtualisation [64] to establish an I/O virtualization system for a MCS; A readybuilt separation kernel (i.e., Quest-V [51]) was extended by Missimer et al [53] to support I/O virtualisation. In these methodologies, different system modes are assigned to the VMs, and a secondary scheduling between the VMs is also built to guarantee the more critical I/O requests can be served earlier (i.e., in [37], [38], partitioned RM scheduling of VMs based on periodic servers with fixed period are implemented; in [53], sporadic servers and priority inheritance bandwidthpreserving servers are built to schedule I/O requests among the VMs).…”

“…However, strong isolation usually leads to huge resource waste and poor performance [44]. In recent years, a significant trend in MCS design is to provide certain form and level of resource sharing (e.g., virtualization technology [64]) while keeping sufficient temporal, spatial and fault isolation [53], [55].…”