an evolution of static priority scheduling in real time ... - IRNet Explore

An evolution of static priority scheduling in Real time systemAN EVOLUTION OF STATIC PRIORITY SCHEDULING INREAL TIME SYSTEMJAYA GUPTA 1 & SAROJ HIRANWAL 21,2 Sri Balaji College of Engineering & Technology, Affiliated by Rajasthan Technical University, Benad Road, Near Dadika Phatak, Jaipur-302012, IndiaAbstract- This paper reviews scheduling algorithm available in RTOS to schedule the tasks during execution in the RealTime System considering the static priority and evolution in the static priority scheduling in last decade. Real TimeOperating System (RTOS) success depends on completion of task in given deadline. Real Time Operating System (RTOS)used, where we require specific application that meet there deadline with keeping logically correct results as an importantconstraint. Scheduling can be done in static or dynamic manner. According to priority, scheduling can also be classified asstatic priority and dynamic priority in dynamic scheduling. This paper concentrates on the static priority schedulingalgorithm and the changes made or suggested by researchers and the proposed way to overcome from this problem.Keywords- RTOS, Scheduling Algorithm, Deadline, Dynamic, Static Priority.I. INTRODUCTIONTask scheduling refers to identifying the order inwhich tasks should execute in a system.Embedded system design involves how to assigntasks to hardware and software components. Theprocess of assigning a task, their execution orderis known as scheduling. Since most of the tasksin a real-time embedded system are periodic innature, the real-time task scheduling algorithmsmostly concentrate on periodictasks.[18]Scheduling algorithm can be classifiedin following categories:• Clock Driven Scheduling• Event Driven SchedulingA clock driven scheduling, the scheduling pointsare the interrupts received from a periodic clock.A basic problem with clock driven schedulingstrategies is their inability to handle a largenumber of tasks.In event driven scheduling,scheduler responds to different events in thesystem. The scheduling points are the events likearrival and completion of tasks. Scheduling alsocan be done on the basis of Priority. Priority isassigned to the tasks which may be static ordynamic, on the basis of priority we decide thesequence of task to be executed. In static prioritywe have two main algorithms: 1) RateMonotonic (RM) and 2) Deadline Monotonic.The most successful algorithm in static priorityis Rate Monotonic. In enduring paper we willdiscuss about the Rate Monotonic and thechanges suggested by the researchers in thestandard algorithm suggested by Liu andLayland. Section 2 of this paper describesoverview of scheduling algorithms available instatic priority with their respectivedisadvantages. Survey of related literature isgiven in Section 3. Proposed Algorithm is givenin Section 4, Conclusion and future work isgiven in Section 5.II.SCHEDULINGALGORITHM’S-OVERVIEW2.1 Rate Monotonic Algorithm Rate-monotonicscheduling is a scheduling algorithm in whichstatic-priority is used for scheduling.Fig 1: Classification of Scheduling algorithm.International Conference on Advanced Research in Mechanical, Electrical and Computer Engineering, 24- 26 th March 2013, Bangkok65

The static priorities are assigned according to theperiod of the job, the shortest period job get thehighest job priority and vice-versa.Theseoperating systems are generally preemptive andhave deterministic guarantees with regard toresponse times. [1, 3]A simple version of ratemonotonicanalysis assumes that threads havethe following properties:1. All processes run on a single CPU i.e. noparallel task execution.2. Deterministic deadlines are exactly equal toperiods3. Static priorities (the task with the higheststatic priority that is run able immediatelypreempts all other tasks)4. Tasks are independent from each other.5. Context switch times have no impact on theperformance.Liu & Layland (1973) proved that for a set of nperiodic tasks with unique periods, a feasibleschedule that will always meet deadlines exists ifthe CPU utilization is below a specific bound(depending on the number of tasks). [21] Theschedulability test for RMS is:An evolution of static priority scheduling in Real time systemversa.This priority policy is optimal for a set ofperiodic or sporadic tasks with followingconditions:1. All tasks deadlines should less than or equalto their minimum inter-arrival times (orperiods).2. All tasks have worst-case execution times(WCET) that are less than or equal to theirdeadlines.3. All tasks are independent and they do notaffect the working of other by blocking.4. No task voluntarily suspends itself.5. During execution there could be an instantoccur when all the task goes ready toexecute that instant is known as criticalinstant.6. Scheduling overheads (switching from onetask to another) are zero.7. All tasks have zero release jitter (the timefrom the task arriving to it becoming readyto execute).[16] Drawback:• It is less predictable as task dependsupon the deadlines.• It is less controllable.III. RELATED WORK DONEWhere Ci is the computation time, Ti is therelease period (with deadline one period later),and n is the number of processes to bescheduled.[15,18,20]Drawbacks:• It is difficult to handle aperiodic and sporadictasks.• It is not optimal if deadline and period of thetask are differing.2.2 Deadline MonotonicDeadline-monotonic scheduling is a schedulingalgorithm in which static-priority is used forscheduling. The static priorities are assignedaccording to the deadline of the job, the shortestdeadline job get the highest job priority and ice-Since 1973, when Liu and Layland proved thatfor a set of n periodic tasks with unique periods,a feasible schedule that will always meetdeadlines exists if the CPU utilization is below aspecific bound. In field of static schedulingalgorithm designing lot of research have beendone like some worked on finding the optimalscheduling algorithm, do the comparison ofscheduling algorithm, try to create algorithmbetter for all type of task or suggest the changesin existing algorithm for improvement.Maximum research done in static priorityscheduling is related to Rate Monotonic, as it isthe optimal algorithm available in thisscheduling technique. Few of them areconcluded in the given below table with theirrespective years:International Conference on Advanced Research in Mechanical, Electrical and Computer Engineering, 24- 26 th March 2013, Bangkok66

An evolution of static priority scheduling in Real time systemTable 1: Summary of different proposed algorithms for task scheduling in real time system3.1 Limitations:International Conference on Advanced Research in Mechanical, Electrical and Computer Engineering, 24- 26 th March 2013, Bangkok67

An evolution of static priority scheduling in Real time systemIn above created table we read about manyalgorithms and suggestions in existing algorithmpatterns. In “Modified Rate-MonotonicAlgorithm for Scheduling Periodic Jobs withDeferred Deadlines”, Author proposed semistaticpriority-driven algorithm; each job isassigned two fixed priorities, the higher priorityfor its previous job request and the lower priorityfor recent job request. When the amounts ofdeference are sufficiently large, jobs can befeasibly scheduled using this algorithm withcondition that their total utilization is equal to orless than one. Limitation of this algorithm is, in aspecial case modified rate-monotonic algorithmis not optimal for this case when the period ratiois larger than two.[2]In “A Reservation-BasedAlgorithm for Scheduling Both Periodic andAperiodic Real-Time Tasks” author concludethat however there approach is good but not bestas there some issues could arise like: if the taskswitching time is not negligible, this overhead inthe RB algorithm may affect the derivation. Agood scheduling algorithm (EDF, MUF) mayimprove the probability of guaranteeingaperiodic tasks. [5, 13] In “Modified Version ofRate-Monotonic Scheduling Algorithm and itsEfficiency Assessment” Author identifies specialcase of systems execution using RMA, for thisspecial case it was shown that least upper boundto load factor is improved. Author nameddelayed rate-monotonic algorithm. There are fewset of task that can run safely with DRM andbehave unsafe with RM. Due to importance ofsafety verification of real-time systems prior totheir being operational there is still need to makeit more efficient.[6]In “Space of Rate MonotonicSchedulability” Approach to analyze theschedulability of periodic tasks under the RateMonotonic priority approach, such an approachallowed us to accurately decide the feasibilityregion in the space of task computation timesand able to derive a tunable guarantee test.Tunability property is important in special caseswhere performance of a polynomial time test isnot adequate for achieving high processorutilization, and the overhead introduced by exacttests is too high for an on-line admissioncontrol. In the proposed algorithm we canintroduce different cases like when taskcomputation times are considered as randomvariables with known probability distribution.[10]In “Rate Monotonic SchedulabilityConditions Using Relative Period Ratios”, Thispaper presents a special case of RMschedulability bound as a function of z1 and z2,the ratios of the smallest and the largest virtualperiods to the largest period. We can reduce thedifference between the smallest and largest jobperiod and a periodic task system can beachieve a higher Rate-Monotonicschedulability. Limitation of this approach is thatthis provide good result in uniprocessor but inmultiprocessor environment it still need to beexecute.[11] In ”Response Time Analysis ofAsynchronous Periodic and Sporadic TasksScheduled by a Fixed-Priority PreemptiveAlgorithm” [7]. This algorithm can be used tostudy the schedulability of asynchronousperiodic and sporadic tasks scheduled by a fixedprioritypreemptive algorithm. It can be used to:• Diminish the pessimism induced by thecritical instant• Allow offset relationships• Determine the worst response time of eachjob– analyze response jitter– account for missed deadlines• Calculate the asynchronous critical instant– integrate sporadic tasks– increase analysis speedThe calculation of the response times of all jobsunder a particular condition is necessary.Therefore, the proposed methods do not scalewell against the number of periodic tasks,especially task sets using co-prime periods.In“Modified Rate-Monotonic Algorithm forScheduling Periodic Tasks with DifferentImportance in Embedded System”, Authorproposes a new scheduling algorithm based onRM. It improves the scheduling of long periodtasks which are important by adding twoparameters, importance and laxity. Authorsuggests if importance is higher when task laxityis zero could preempt the running task. Authorsalso suggest that the algorithm could be decreasethe deadline-missing ratio of the tasks and theCPU resource could be used more efficiently.Limitation of this algorithm is it can use only inlimited area where priority does not depends ondeadline or on period.[8, 15]In “Global Rate-Monotonic Scheduling with Priority Promotion”In this paper, author presents the Rate-Monotonic until Zero Laxity (RMZL) algorithm,which applies the laxity-driven prioritypromotion strategy to the global Rate-Monotonicalgorithm. Authors execute the schedulabilitytest and the tardiness bound in the algorithm.According to author RMZL algorithm is able toInternational Conference on Advanced Research in Mechanical, Electrical and Computer Engineering, 24- 26 th March 2013, Bangkok68

execute more tasks than the Rate-Monotonic andthe RMUS. As per paper the RMZL algorithm iscompetitive with and is even better for a largernumber of CPU cores than the EDZL algorithm.Limitation of this paper is that they presentschedulability test is pessimistic, while itoutperforms the existing fixed-priorityalgorithms. There is one more issue is there howto detect the zero-laxity condition. [12, 14]In lastalgorithm “A New Scheduling Algorithm forReal Time System” author purpose a algorithmwhich eliminates the drawbacks of considering asingle component which could be period ordeadline to assign the priority. This algorithmcan be implemented in real time system withcondition where period and deadline doesn’tvary with time. In rate monotonic algorithm allavailable task consider as equal important andpriority decide by the period of the task, due towhich some unimportant task could bescheduled before an important task. Deadlinemonotonic algorithm where all the tasksconsider as equal important and they scheduledon the basis of deadlines same problem couldoccur in this algorithm also. Authors proposearchitecture to overcome with this drawback bycombining the advantages of period anddeadline. Percentage of weightage is provided toall the parameter and based on their weight, apriority component calculated on the basis ofwhich we schedule the task. Limitation of thisproposed algorithm is that, it can beimplemented in real time system with conditionwhere period and deadline doesn’t vary withtime. [17,19]IV.PROPOSED ALGORITHM-ARCHITETUREAn evolution of static priority scheduling in Real time systemthe sorting of task will be done on the basis ofthis selective component.V. CONCLUSION & FUTURE WORKThis paper presents a categorized summary ofwork done in the area of static priority basedtask schedulingin RTOS, types of schedulingalgorithm like clock driven & event driven andthere subclass algorithms for e.g. cyclic, tabledriven, rate monotonic, deadline monotonic andtheir advantages and disadvantages . My reviewconcludes that there is no single algorithmavailable which optimal for any type of tasks,every algorithm has some finite limitation.Themajor challenge is, no algorithm has beensuggested such as to use one or more than one,important parameters to make an equation suchthat the created algorithm could select a set oftask and do there scheduling as no task miss thedeadline or make the throughput time delayminimum. In last section we proposed analgorithm architecture using which we canreduce the deadline missing time.REFERENCES:[1] John Lehoczky , Lui Sha and Ye Ding “The RateMonotonic[2] Scheduling Algorithm: Exact Characterization AndAverage Case Behavior” CH2803-5/89/0000/0166,1989 IEEE. Liu and C. L. Liu“Modified Rate- Monotonic Algorithm for SchedulingPeriodic Jobs with Deferred Deadlines” IEEEtransactions on sofiware engineering, vol. 19, no. 12,december 1993[3] Krithi Ramamritham and John A. Stankovic ”Scheduling Algorithms and Operating Systems Supportfor Real-Time Systems” proceedings of the IEEE, vol.82. no. I , january 1994.After doing the literature survey, we are now incondition to purpose an architecture using whichwe can calculate a selective component.Selective component will take 30% weightagefrom period, 20% from deadline and 50%weightage will be taken from user definedpriority. On the basis of this selective componentnow task scheduling we be performed, in caseany two or more task calculate the sameselective component then period will be checkand short period task will execute first. For eg.Let’s consider a task whose period is 45,deadline is 55 and the user defined priority is 4.Now the selective component will be calculated30% of 45 + 20% of 55 + 50% of 4= 26. Now[4] John A. Stankovic, Marco Spuri hlarco Di Natale andGiorgio C Buttazzo” Implications of ClassicalScheduling Results for Real-Time Systems”. 0018-9162/95,1995 IEEE.[5] Kang G. Shin and Yi-Chieh Chang ” A Reservation-Based Algorithm for Scheduling Both Periodic andAperiodic Real-Time Tasks”. IEEE Transactions onComputers, vol. 44, no. 12, december 1995.[6] Mahmoud Naghibzadeh ” A Modified Version of Rate-Monotonic Scheduling Algorithm and its EfficiencyAssessment” 1530-1443/02, 2002 IEEE.[7] Manuel Coutinho, Jos´e Rufino and Carlos Almeida”Response Time Analysis of Asynchronous Periodic andSporadic Tasks Scheduled by a Fixed-PriorityPreemptive Algorithm” 1068-3070/08, 2008 IEEE.[8] Jiwen Dong ,Yang Zhang “A Modified Rate-MonotonicAlgorithm for Scheduling Periodic Tasks withInternational Conference on Advanced Research in Mechanical, Electrical and Computer Engineering, 24- 26 th March 2013, Bangkok69

An evolution of static priority scheduling in Real time systemDifferent Importance in Embedded System”ICEMI’2009.International Journal of Advances in Science andTechnology, Vol. 1, No. 4, 2010.[9] Li Jie, Guo Ruifengand Shao Zhixiang “The Researchof Scheduling Algorithms in Real-time System” 2010International Conference on Computer andCommunication Technologies in AgricultureEngineering.[10] Enrico Bini, Giorgio C. Buttazzo “The Space of RateMonotonic Schedulability” Proceedings of the 23rdIEEE Real-Time Systems Symposium (RTSS’02),1052-8725/02 2002 IEEE[11] Wan-Chen Lu, Hsin-Wen Wei, Kwei-Jay Lin “RateMonotonic Schedulability Conditions Using RelativePeriod Ratios” Proceedings of the 12th IEEEInternational Conference on Embedded and Real- TimeComputing Systems and Applications (RTCSA'06)2006[12] Shinpei Kato, Akira Takeda and Nobuyuki Yamasaki”Global Rate-Monotonic Scheduling with PriorityPromotion” IPSJ Transactions on AdvancedComputing System (ACS), Vol. 2, No. 1, pp. 64–74,March 2008.[13] Nasro Min-Allah · Samee Ullah Khan · Nasir Ghani ·Juan Li · LizheWang ·”A comparative study of ratemonotonic schedulability tests” SpringerScience+Business Media, LLC 2011.[14] M.Kaladevi and Dr.S.Sathiyabama,” A ComparativeStudy of Scheduling Algorithms for Real Time Task”[15] Liu, C. L.; Layland, J. (1973), "Scheduling algorithmsfor multiprogramming in a hard real-timeenvironment", Journal of the ACM 20 (1): 46–61.[16] Alan Burns and Andy Wellings (2001). Real-TimeSystems and Programming Languages (3rd Edition).Addison Wesley.[17] Yaashuwanth .C and Dr.R. Ramesh” A New SchedulingAlgorithm for Real Time System” International Journalof Computer and Electrical Engineering, Vol.2, No.6,December, 2010 1793-8163[18] Santanu Chattopadhyay , Embedded System DesignPHI, ch-5, pg- 53-79.[19] Enrico Bini and Giorgio C. Buttazzo, “SchedulabilityAnalysis of Periodic Fixed Priority Systems” in IEEETransactions on Computers,Vol. 53, No. 11, pp. 1462-1473, Nov. 2004.[20] Lui Sha, Ragunathan Rajkumar and Shirish s. sathaye“Generalized Rate-Monotonic Scheduling Theory: AFramework for Developing Real-Time System” ,Proceeding of the IEEE, Vol 82, No.1, January 1994.[21] Alan A. Bertossi , Andrea Fusiello “Rate-Monotonicscheduling for hard-real-time systems” 1997 ElsevierScience.International Conference on Advanced Research in Mechanical, Electrical and Computer Engineering, 24- 26 th March 2013, Bangkok70