Citation Context

... presents overheads of our time-sensitive Linux system. Finally, in Section 5 we state our conclusions. 1.1 Related Work The scheduling problem has been extensively studied by the real-time community =-=[11, 14, 20]-=-. However, most of the scheduling analysis is based on an abstract mathematical model that ignores practical systems issues such as kernel non-preemptibility and interrupt processing overhead. Recentl...

Citation Context

... presents overheads of our time-sensitive Linux system. Finally, in Section 5 we state our conclusions. 1.1 Related Work The scheduling problem has been extensively studied by the real-time community =-=[11, 14, 20]-=-. However, most of the scheduling analysis is based on an abstract mathematical model that ignores practical systems issues such as kernel non-preemptibility and interrupt processing overhead. Recentl...

Citation Context

...umber of interrupts generated by the timing mechanism and can be mitigated by the soft-timer mechanism [5]. Thus, our firm-timer implementation uses soft timers. Finally, the Nemesis operating system =-=[10]-=- is designed for multimedia and other time-sensitive applications. However, its structure and API is very different from the standard programming environment provided by general-purpose operating syst...

Citation Context

... a timesensitive Linux system was to accurately implement a proportion-period scheduler. This scheduler is used to provide a reservation mechanism for a higher-level feedback-based real-rate scheduler=-=[22]-=-. The real-rate scheduler uses an application-specific progress rate metric in time-sensitive tasks to automatically assign correct allocations to such tasks. For example, the progress of a producer o...

Citation Context

... presents overheads of our time-sensitive Linux system. Finally, in Section 5 we state our conclusions. 1.1 Related Work The scheduling problem has been extensively studied by the real-time community =-=[11, 14, 20]-=-. However, most of the scheduling analysis is based on an abstract mathematical model that ignores practical systems issues such as kernel non-preemptibility and interrupt processing overhead. Recentl...

Citation Context

...parts of interrupt processing after realtime processes. We are currently investigating this solution. An alternative method for evaluating the scheduler behavior is to use a kernel tracer, such as LTT=-=[23]-=-, that can register the occurrence of certain key events in the kernel. These events can be analyzed later after program execution. Kernel tracers are often used in real-time systems for verifying the...

Citation Context

... First, we present firm timers, an efficient high-resolution timer mechanism. Firm timers incorporate the benefits of periodic timers, one-shot timers available on modern hardware [8] and soft timers =-=[5]-=- and provide accurate timing with low overhead. Second, we use a preemptible kernel design for a responsive kernel. Finally, we use both priority and reservation-based CPU scheduling mechanisms for su...

Citation Context

...their performance overhead on throughput-oriented applications is not clear. A different approach for providing real-time performance is used by other systems, such as RTLinux [6], RTAI [13], and KURT=-=[21]-=-, which decrease the unpredictability of the system by running Linux as a background process over a small real-time executive. In this case, real-time tasks are not Linux processes, but run on the low...

Citation Context

...eds of time-sensitive applications which require more precise allocation. Thus, recently several approaches have been proposed to improve the timing response of a general-purpose system such as Linux =-=[17, 24]-=-. These approaches include improved kernel preemptibility and a more generic scheduling framework. Since their focus is hard real-time, they do not evaluate the performance of non-real time applicatio...

Citation Context

...reased interrupts with fine-grained timing. The data structures for one-shot timers are less efficient than for periodic timers. For instance, periodic timers can be implemented using calendar queues =-=[7]-=- which operate in ¢¤£¦¥¨§ time, while one-shot timers require priority heaps which require ¢©£����¨��£���§�§ time, where � is the number of active timers. This difference exists because periodic timer...

Citation Context

... this value in an array. The scheduler behavior is inferred at the userlevel by simply measuring the time difference between successive elements of the array. A similar technique is used by Hourglass =-=[18]-=-. Table 1 shows the maximum deviation in the proportion allocated and the period boundary for each of the two 3 The current proportion-period scheduler allows task periods that are multiples of 512 us...

Citation Context

...applications, their performance overhead on throughput-oriented applications is not clear. A different approach for providing real-time performance is used by other systems, such as RTLinux [6], RTAI =-=[13]-=-, and KURT[21], which decrease the unpredictability of the system by running Linux as a background process over a small real-time executive. In this case, real-time tasks are not Linux processes, but ...

Citation Context

...scheduled Time Scheduler Application scheduled (activation) Figure 1: Execution sequence after wall-clock time expiration event. the most common source of latency in an operating system such as Linux =-=[4]-=-. Such a mechanism can be implemented efficiently by using two techniques, one-shot timers available on most common modern hardware and soft timers [5]. These techniques are complementary and can be c...

Citation Context

...eds of time-sensitive applications which require more precise allocation. Thus, recently several approaches have been proposed to improve the timing response of a general-purpose system such as Linux =-=[17, 24]-=-. These approaches include improved kernel preemptibility and a more generic scheduling framework. Since their focus is hard real-time, they do not evaluate the performance of non-real time applicatio...

Citation Context

...ng time-sensitive applications. Thus most of the existing real-time kernels or real-time extensions to Linux provide high resolution timers. The high resolution timers concept was proposed by RT-Mach =-=[19]-=- and has subsequently been used by several other systems such as Rialto [9]. In a general-purpose operating system, the overhead of such timers can affect the performance of throughput-oriented applic...

Citation Context

...ion when an execution flow enters the kernel. To support this level of kernel preemptibility, kernel data must be explicitly protected using mutexes or spinlocks. The Linux preemptible kernel project =-=[12]-=- uses this approach and disables kernel preemption only when a spinlock is held. In a preemptible kernel, the non-preemptible section latency is determined by the maximum amount of time for which a sp...

Citation Context

...rt their functionality to the user level through the standard POSIX API. Conversely, a responsive kernel without accurate timing has only a few applications. For example, the low latency Linux kernel =-=[16]-=- provides low latency only when an external interrupt source such as an audio card is used. Similarly, a scheduler that provides good theoretical guarantees is not effective when the kernel is not res...

Citation Context

...is the amount of CPU allocation required every period for correct task execution. The proportion-period model can be effectively implemented using well known results from real-time scheduling research=-=[15]-=-. In this implementation, classical real-time scheduling techniques (EDF or RM priority assignment) are used and a task is allowed to execute as a real-time task for a time � equal to the product of i...

Citation Context

...or real-time extensions to Linux provide high resolution timers. The high resolution timers concept was proposed by RT-Mach [19] and has subsequently been used by several other systems such as Rialto =-=[9]-=-. In a general-purpose operating system, the overhead of such timers can affect the performance of throughput-oriented applications. This overhead is caused by the increased number of interrupts gener...