Björn B. Brandenburg

A Flexible Real-Time Locking Protocol for Multiprocessors

Real-time scheduling algorithms for multiprocessor systems have been the subject of considerable recent interest. For such an algorithm to be truly useful in practice, support for semaphore-based locking must be provided. However, for many global scheduling algorithms, no such mechanisms have been proposed. Furthermore, in the partitioned case, most prior semaphore schemes are either inefficient or restrict critical sections considerably. In this paper, a new flexible multiprocessor locking scheme is presented that can be applied under both partitioning and global scheduling. This scheme allows unrestricted critical-section nesting, but has been designed to deal with the common case of short non-nested accesses efficiently.

On the Scalability of Real-Time Scheduling Algorithms on Multicore Platforms: A Case Study

Multicore platforms are predicted to become significantly larger in the coming years. Given that real-time workloads will inevitably be deployed on such platforms, the scalability of the scheduling algorithms used to support such workloads warrants investigation. In this paper, this issue is consideredand an empirical evaluation of several global and partitioned scheduling algorithms is presented. This evaluation was conducted using a Sun Niagara multicore platformwith 32 logical CPUs (eight cores, four hardware threads per core). In this study, each tested algorithm proved to be a viable choice for some subset of the workload categories considered.

An Empirical Comparison of Global, Partitioned, and Clustered Multiprocessor EDF Schedulers

As multicore platforms become ever larger, overhead-related factors play a greater role in determining which real-time scheduling algorithms are preferable. In this paper, such factors are investigated through an empirical comparison of global, partitioned, and clustered EDF scheduling algorithms on a 24-core Intel system. On this platform, global EDF proved to be a non-viable choice for hard real time systems, while clusters of size six practically approximated global approaches. For soft real-time systems, clustered EDF scheduling algorithms proved to be particularly effective. This study suggests that future global scheduling research should focus on small-to-medium multicore platforms rather than large platforms.

Real-Time Synchronization on Multiprocessors: To Block or Not to Block, to Suspend or Spin?

In the domain of multiprocessor real-time systems, there has been a wealth of recent work on scheduling, but relatively little work on the equally-important topic of synchronization. When synchronizing accesses to shared resources, four basic options exist: lock-free execution, wait-free execution, spin- based locking, and suspension-based locking. To our knowledge, no empirical multiprocessor-based evaluation of these basic techniques that focuses on real-time systems has ever been conducted before. In this paper, we present such an evaluation and report on our efforts to incorporate synchronization support in the testbed used in this effort.

Optimality Results for Multiprocessor Real-Time Locking

When locking protocols are used in real-time systems, bounds on blocking times are required when ensuring timing constraints. While the term “blocking” is well-understood in the context of uniprocessor real-time systems, the same is not true in the multiprocessor case. In this paper, two definitions of blocking are presented that are applicable to suspension-based multiprocessor locking protocols. The need for two definitions arises because of differences in how suspensions are handled in existing schedulability analysis. For each definition, locking protocols are presented that have asymptotically optimal blocking behavior. In particular, protocols are presented for any job-level static-priority global or partitioned scheduling algorithm.

Is Semi-Partitioned Scheduling Practical?

Semi-partitioned schedulers are -- in theory -- a particularly promising category of multiprocessor real-time scheduling algorithms. Unfortunately, issues pertaining to their implementation have not been investigated in detail, so their practical viability remains unclear. In this paper, the practical merit of three EDF-based semi-partitioned algorithms is assessed via an experimental comparison based on real-time schedulability under consideration of real, measured overheads. The presented results indicate that semi-partitioning is indeed a sound and practical idea. However, several problematic design choices are identified as well. These shortcomings and other implementation concerns are discussed in detail.

On the Implementation of Global Real-Time Schedulers

An empirical study of implementation tradeoffs (choice of ready queue implementation, quantum-driven vs. event driven scheduling, and interrupt handling strategy) affecting global real-time schedulers, and in particular global EDF, is presented. This study, conducted using UNC’s Linux-based LITMUSRT on Sun’s Niagara platform, suggests that implementation tradeoffs can impact schedulability as profoundly as scheduling-theoretic tradeoffs. For most of the considered workloads, implementation scalability proved to not be a key limitation of global EDF on the considered platform. Further, a combination of a parallel heap, event-driven scheduling, and dedicated interrupt handling performed best for most workloads.

An Implementation of the PCP, SRP, D-PCP, M-PCP, and FMLP Real-Time Synchronization Protocols in LITMUS^RT

We extend the FMLP to partitioned static-priority scheduling and derive corresponding worst-case blocking bounds. Further, we present the first implementation of the PCP, SRP, D-PCP, M-PCP, and FMLP synchronization protocols in a unified framework in a general-purpose OS and discuss design issues that were beyond the scope of prior algorithmic-oriented work on real-time synchronization.

Integrating Hard/Soft Real-Time Tasks and Best-Effort Jobs on Multiprocessors

We present a multiprocessor scheduling framework for integrating hard and soft real-time tasks and best-effort jobs. This framework allows for full system utilization, and ensures that hard real-time deadlines are met and that deadline tardiness is bounded for soft real-time tasks. Dynamic slack reclamation is employed to reduce tardiness and to improve the response time of best-effort jobs. The approach is validated using an implementation within the Linux kernel.

A Comparison of the M-PCP, D-PCP, and FMLP on LITMUSRT

This paper presents a performance comparison of three multiprocessor real-time locking protocols: the multiprocessor priority ceiling protocol (M-PCP), the distributed priority ceiling protocol (D-PCP), and the flexible multiprocessor locking protocol (FMLP). In the FMLP, blocking is implemented via either suspending or spinning, while in the M-PCP and D-PCP, all blocking is by suspending. The presented comparison was conducted using a UNC-produced Linux extension called LITMUSRT. In this comparison, schedulability experiments were conducted in which runtime overheads as measured on LITMUSRT were used. In these experiments, the spin-based FMLP variant always exhibited the best performance, and the M-PCP and D-PCP almost always exhibited poor performance. These results call into question the practical viability of the M-PCP and D-PCP, which have been the de-facto standard for real-time multiprocessor locking for the last 20 years.