S. Ramamurthy. A lock-free approach to object sharing in real-time systems. PhD thesis, University of North Carolina, Chapel Hill, North Carolina, 1997.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....one of two possible forms: interference within a quantum and interference across multiple quanta. Both forms are illustrated in Figure 9. Bounding the number of interferences across multiple quanta is trivial in most cases due to practical considerations. In experiments conducted by Ramamurthy [19] on a 66 MHz processor, operation durations for a variety of common objects (e.g. queues, linked lists, etc. were found to be in the range of tens of microseconds. On modern processors, these operations will likely require no more than a few microseconds. Since quantum sizes typically range from ....

S. Ramamurthy. A lock-free approach to object sharing in real-time systems. Dissertation, University of North Carolina at Chapel Hill. 1997.

....both U and T have issued a request for . 3 Short Critical Section Protocols In this section, we present a simple method for implementing locks when all critical sections guarded by a common lock are short relative to the length of the scheduling quantum. In experiments conducted by Ramamurthy [21] on a 66 MHz processor, critical section durations for a variety of common objects (e.g. queues, linked lists, etc. were found to be in the range of tens of microseconds. On modern processors, these operations will likely require no more than a few microseconds. Since quantum sizes typically ....

.... to generate random task sets: jaej 2 [5; 100] j j 2 [25; 175 Delta log 2 M ] T:n 2 [2; 12] and T:p 2 [200; 5000] In addition, critical section lengths were chosen from the range [0:001; UB ] where UB is an experiment dependent value (as explained later) Based on Ramamurthy s observations [21], we assumed that the simplest critical sections (e.g. enqueuing to a shared queue) require at most a few microseconds on modern processors. Since a one millisecond quantum is typical, 0:001 quanta was chosen as a lower bound on critical section lengths. To reflect the assumption that short ....

S. Ramamurthy. A lock-free approach to object sharing in real-time systems. Dissertation, University of North Carolina at Chapel Hill. 1997.

....scheduling overhead. 5.4 Fault Tolerance and Overload The presence of faults may cause some processors to become overloaded . In particular, if K out of M processors fail, then all tasks now need to be executed on the remaining M K processors. The In experiments conducted by Ramamurthy [35] on a 66 MHz processor, critical section durations for a variety of common objects (e.g. queues, linked lists, etc. were found to be in the range of tens of microseconds. On modern processors, these operations will likely require no more than a few microseconds. Note that, in a Pfair scheduled ....

S. Ramamurthy. A lock-free approach to object sharing in real-time systems, 1997. PhD thesis, University of North Carolina at Chapel Hill.

....while quantum sizes are not changing. Even with the technology of several years ago, one could make the case that object calls are typically short compared to a quantum. As evidence of this, we cite results from experiments conducted by Ramamurthy to compute access times for several common objects [18]. These experiments were performed on a 25 MHz 68030 machine and involved objects ranging from queues to linked lists to medium sized balanced trees. Both lock based and lock free (see below) object implementations were evaluated. Ramamurthy found that, even on a slow 25 MHz machine, all object ....

S. Ramamurthy. A Lock-Free Approach to Object Sharing in Real-Time Systems. PhD thesis, University of North Carolina, Chapel Hill, North Carolina, 1997.

....while quantum sizes are not changing. Even with the technology of several years ago, one could make the case that object calls are typically short compared to a quantum. As evidence of this, we cite results from experiments conducted by Ramamurthy to compute access times for several common objects [21]. These experiments were performed on a 25 MHz 68030 machine and involved objects ranging from queues to linked lists to medium sized balanced trees. Both lock based and lock free (see below) object implementations were evaluated. Ramamurthy found that, even on a slow 25 MHz machine, all object ....

....access costs for each of the ten shared objects. Three objects were defined to have access times of 7 8 time units, five to have access times of 57 96 time units, and two to have access times of 134 180 time units. These values were chosen based upon the object access times reported by Ramamurthy [21]. Each task was assigned a computation cost of between one and three quanta, with a quantum being 1000 time units. The number of quanta required by a task was selected at random, with 40 of the generated tasks taking one quantum, 40 taking two, and 20 taking three. Each quantum in a task s ....

S. Ramamurthy. A Lock-Free Approach to Object Sharing in Real-Time Systems. PhD thesis, University of North Carolina, Chapel Hill, North Carolina, 1997.

No context found.

S. Ramamurthy. A lock-free approach to object sharing in real-time systems. PhD thesis, University of North Carolina, Chapel Hill, North Carolina, 1997.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC