#### DMCA

## On multiprocessor utility accrual real-time scheduling with statistical timing assurances (2006)

### Cached

### Download Links

- [www.real-time.org]
- [www.real-time.ece.vt.edu]
- DBLP

### Other Repositories/Bibliography

Venue: | In IFIP Embedded and Ubiquitous Computing (EUC |

Citations: | 8 - 1 self |

### Citations

318 | Proportionate progress: A notion of fairness in resource allocation
- Baruah, Cohen, et al.
- 1996
(Show Context)
Citation Context ...igration is allowed—e.g., at job boundaries. Carpentar et al. [6] have catalogued multiprocessor real-time scheduling algorithms considering the degree of job migration. The Pfair class of algorithms =-=[4]-=- that allow full migration have been shown to achieve a schedulable utilization bound (below which all tasks meet their deadlines) that equals the total capacity of all processors — thus, they are the... |

212 |
On a real-time scheduling problem
- Dhall, Liu
- 1978
(Show Context)
Citation Context ... algorithms other than Pfair (e.g., global EDF) have also been studied though their schedulable utilization bounds are lower. Global EDF scheduling on multiprocessors is subject to the “Dhall effect” =-=[11]-=-, where a task set with total utilization arbitrarily close to 1.0 cannot be scheduled satisfying all deadlines. To overcome this, researchers have studied global EDF’s behavior under restricted indiv... |

203 | Dynamic and Aggressive Scheduling Techniques for Power-Aware Real-Time Systems”,
- Aydin, Melhem, et al.
- 2001
(Show Context)
Citation Context ... statistical estimation of the demand is much more stable and hence more predictable than the actual workload; (2) worst-case workload is usually a very conservative prediction of the actual workload =-=[2]-=-, resulting in resource over-supply; and(3) allocating execution times based on the statistical estimation of tasks’ demands can provide statistical performance assurances, which is sufficient for our... |

126 | Multiprocessor EDF and Deadline Monotonic Schedulability Analysis
- Baker
(Show Context)
Citation Context ... 1) [22]. In [12], Goossens et. al show that EDF’s utilization bound is M − (M − 1) umax. This work was later extended by Baker for the more general case of deadlines less than or equal to periods in =-=[3]-=-. In [5], Bertogna et al. show that Baker’s utilization bound does not dominate the bound of Goossens et. al, and vice versa. While most of these past works focus on the hard real-time objective of al... |

120 | Priority-driven scheduling of periodic task systems on multiprocessors
- Goossens, Funk, et al.
- 2003
(Show Context)
Citation Context ...ations. For example, on M processors, Srinivasan and Baruah show that when the maximum individual task utilization, umax, is bounded by M/ (2M − 1), EDF’s utilization bound is M 2 / (2M − 1) [22]. In =-=[12]-=-, Goossens et. al show that EDF’s utilization bound is M − (M − 1) umax. This work was later extended by Baker for the more general case of deadlines less than or equal to periods in [3]. In [5], Bert... |

107 | Scheduling Dependent Real-Time Activities
- Clark
- 1990
(Show Context)
Citation Context ...verloads), our objective is to maximize the system-level total utility. This problem is N P-hard because it subsumes the N P-hard problem of scheduling dependent tasks with step TUFs on one processor =-=[7]-=-. 3 The gMUA Algorithm 3.1 Bounding Accrued Utility Let si,j be the sojourn time of the j th job of task Ti, where the sojourn time is defined as the period from the job’s release to its completion. N... |

85 | Improved schedulability analysis of EDF on multiprocessor platforms
- Bertogna, Cirinei, et al.
- 2005
(Show Context)
Citation Context .... In [12], Goossens et. al show that EDF’s utilization bound is M − (M − 1) umax. This work was later extended by Baker for the more general case of deadlines less than or equal to periods in [3]. In =-=[5]-=-, Bertogna et al. show that Baker’s utilization bound does not dominate the bound of Goossens et. al, and vice versa. While most of these past works focus on the hard real-time objective of always mee... |

64 |
A time-driven scheduling model for real-time systems
- Jensen, Locke, et al.
- 1985
(Show Context)
Citation Context ...duling algorithms such as EDF are optimal (on one processor). Deadlines by themselves cannot express both urgency and importance. Thus, we consider the abstraction of time/utility functions (or TUFs) =-=[14]-=- that express the utility of completing an application activity as a function of that activity’s completion time. We specify deadline as a binaryvalued, downward “step” shaped TUF; Figure 1(a) shows e... |

54 |
Deadline-based scheduling of periodic task systems on multiprocessors
- Srinivasan, Baruah
- 2002
(Show Context)
Citation Context ...sk utilizations. For example, on M processors, Srinivasan and Baruah show that when the maximum individual task utilization, umax, is bounded by M/ (2M − 1), EDF’s utilization bound is M 2 / (2M − 1) =-=[22]-=-. In [12], Goossens et. al show that EDF’s utilization bound is M − (M − 1) umax. This work was later extended by Baker for the more general case of deadlines less than or equal to periods in [3]. In ... |

49 |
System support for automated profiling and optimization
- Zhang, Wang, et al.
- 1997
(Show Context)
Citation Context ...es two steps: (1) profiling its execution time usage, and (2) deriving the probability distribution of that usage. A number of measurement-based, off-line and online profiling mechanisms exist (e.g., =-=[24]-=-). We assume that the mean and variance of Yi are finite and determined through either online or off-line profiling. We denote the expected execution time demand of a task Ti as E(Yi), and the varianc... |

45 | Efficient scheduling of soft real-time applications on multiprocessors
- Srinivasan, Anderson
(Show Context)
Citation Context ...ile most of these past works focus on the hard real-time objective of always meeting all deadlines, recently, there has been efforts that consider the objective of bounding the tardiness of tasks. In =-=[21]-=-, Srinivasan and Anderson derive a tardiness bound for a suboptimal Pfair scheduling algorithm. In [1], for a restricted migration model, where migration is allowed only at job boundaries, Andersen e... |

44 |
Mechanisms for Reliable Distributed Real-Time Operating Systems — The Alpha Kernel
- Northcutt
- 1987
(Show Context)
Citation Context ...olutions and tradeoffs exist. Examples include linear-time stochastic UA scheduling [15], and using special-purpose hardware accelerators for UA scheduling (analogous to floating-point co-processors) =-=[18]-=-.sSome important corollaries about gMUA’s timeliness behavior can be deduced from EDF’s behavior under conditions (1) and (2). Corollary 2. Under conditions (1) and (2), gMUA always completes the allo... |

42 | An edf-based scheduling algorithm for multiprocessor soft real-time systems
- Anderson, Bud, et al.
- 2005
(Show Context)
Citation Context ...ntly, there has been efforts that consider the objective of bounding the tardiness of tasks. In [21], Srinivasan and Anderson derive a tardiness bound for a suboptimal Pfair scheduling algorithm. In =-=[1]-=-, for a restricted migration model, where migration is allowed only at job boundaries, Andersen et. al present an EDF-based partitioning scheme and scheduling algorithm that ensures bounded tardiness.... |

40 |
Tardiness bounds for global EDF scheduling on a multiprocessor
- Devi, Anderson
(Show Context)
Citation Context ...deadlines) that equals the total capacity of all processors — thus, they are theoretically optimal. However, Pfair algorithms incur significant overhead due to their quantum-based scheduling approach =-=[10]-=-. Thus, scheduling algorithms other than Pfair (e.g., global EDF) have also been studied though their schedulable utilization bounds are lower. Global EDF scheduling on multiprocessors is subject to t... |

35 | P.: On recent advances in time/utility function real-time scheduling and resource management
- Ravindran, Jensen, et al.
- 2005
(Show Context)
Citation Context ...accrual (or UA) criteria, and scheduling algorithms that optimize them, as UA scheduling algorithms. On single processors, UA algorithms that maximize total utility under step TUFs (see algorithms in =-=[20]-=-) default to EDF during under-loads, since EDF satisfies all deadlines during under-loads. Consequently, they obtain the maximum total utility during under-loads. During overloads, they favor more imp... |

32 | Formally verified application-level framework for real-time scheduling on POSIX real-time operating systems
- Peng, Ravindran, et al.
- 2004
(Show Context)
Citation Context ... similar to that of many past UA algorithms [20]. Our prior implementation experience with UA scheduling at the middleware-level have shown that the overheads are in the magnitude of sub-milliseconds =-=[16]-=- (sub-microsecond overheads may be possible at the kernel-level). We anticipate a similar overhead magnitude for gMUA. Though this cost is higher than that of many traditional algorithms, the cost is ... |

30 |
et al. An example real-time command, control, and battle management application for alpha
- Maynard, Shipman
- 1988
(Show Context)
Citation Context ...the Y-axis. Utility ✻ 0 (a) ✲ Time Utility ✻ ❜ ❜ 0 (b) ✲ Time Utility ✻ 0 (c) ❙ ❍❙ ❍❙ ✲ Time Figure 1: Example TUF Time Constraints: (a) Step TUFs; (b) AWACS TUF [8]; and (c) Coastal Air defense TUFs =-=[17]-=-sMany embedded real-time systems also have activities that are subject to non-deadline time constraints, such as those where the utility attained for activity completion varies (e.g., decreases, incre... |

26 |
et al. An adaptive, distributed airborne tracking system
- Clark
- 1999
(Show Context)
Citation Context ...ed real-time systems that are emerging in many domains such as robotic systems in the space domain (e.g., NASA/JPL’s Mars Rover [9]) and control systems in the defense domain (e.g., airborne trackers =-=[8]-=-) are fundamentally distinguished by the fact that they operate in environments with dynamically uncertain properties. These uncertainties include transient and sustained resource overloads due to con... |

17 | Software Organization to Facilitate Dynamic Processor Scheduling - CLARK, JENSEN, et al. - 2004 |

9 |
best-eort real-time scheduling algorithms
- Fast
- 2004
(Show Context)
Citation Context ... executed. Thus, gMUA produces the same schedule as global EDF. 1 When UA scheduling is desired with low overhead, solutions and tradeoffs exist. Examples include linear-time stochastic UA scheduling =-=[15]-=-, and using special-purpose hardware accelerators for UA scheduling (analogous to floating-point co-processors) [18].sSome important corollaries about gMUA’s timeliness behavior can be deduced from ED... |

8 |
et al. A categorization of real-time multiprocessor scheduling problems and algorithms
- Carpenter
- 2004
(Show Context)
Citation Context ...processors by processors’ local scheduling algorithm, like single processor scheduling; and (3) restricted migration, where some form of migration is allowed—e.g., at job boundaries. Carpentar et al. =-=[6]-=- have catalogued multiprocessor real-time scheduling algorithms considering the degree of job migration. The Pfair class of algorithms [4] that allow full migration have been shown to achieve a schedu... |

7 |
Edf and rm multiprocessor scheduling algorithms: Survey and performance evaluation
- Zapata, Alvarez
- 2005
(Show Context)
Citation Context ...henomena, in general, can also affect UA scheduling algorithms’ performance, and counter such algorithms’ ability to maximize the total attained utility. We discuss this with an example inspired from =-=[23]-=-. We consider the case when the execution time demands of all tasks are constant with no variance, and gMUAi estimates them accurately. Example A. Consider M + 1 periodic tasks that are scheduled on M... |

5 |
Adapting Pfair Scheduilng for Symmetric Multiprocessors
- Holman, Anderson
(Show Context)
Citation Context ...lobal UA scheduling on an SMP system with M number of identical processors. We consider global scheduling (as opposed to partitioned scheduling) because of its improved schedulability and flexibility =-=[13]-=-. Further, in many embedded architectures (e.g., those with no cache), its migration overhead has a lower impact on performance [5]. Moreover, applications of interest to us [8, 9] are often subject t... |

1 |
Symmetric multiprocessing.http://www.qnx.com/products/rtos/smp.html. Last accessed
- QNX
- 2005
(Show Context)
Citation Context ...processing capacity is often needed. Responding to this trend, RTOS vendors are increasingly providing multiprocessor platform support — e.g., QNX Neutrino is now available for a variety of SMP chips =-=[19]-=-. But this exposes the critical need for real-time scheduling for multiprocessors — a comparatively undeveloped area of real-time scheduling which has recently received significant research attention,... |