[2013] RTCSA.pdf

[2013] RTCSA.pdf
of 9
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  Scheduling Temporal Data for Real-time Requests inRoadside-to-Vehicle Communication Kai Liu ∗ Victor C.S. Lee ∗ Joseph K.Y. NG † Sang H. Son ‡ ∗ Dept of Computer Science, City University of Hong Kong, Kowloon, Hong KongEmail:,  † Dept of Computer Science, Hong Kong Baptist University, Kowloon, Hong  ‡ Dept of Information and Communication Engineering, DGIST, Daegu,  Abstract —Recent advances in wireless communication tech-nologies have spawned many new applications in vehicularnetworks. Data dissemination via roadside-to-vehicle commu-nication is a vital approach to enabling most of these appli-cations. In this work, we investigate in the scenario wheredata items are broadcasted from the road-side unit (RSU)in response to requests submitted by passing vehicles. Dataitems are associated with temporal constraints and updatedperiodically to reflect dynamic states of traffic information.Each request may ask for multiple temporal data items, andit is associated with a deadline, which may either be specifiedby the driver or imposed by the time when the vehicle drivesthrough the service region. In particular, we develop a real-time data dissemination model based on roadside-to-vehiclecommunication by formulating the time-constraint of requestsand the consistency requirement of retrieving temporal dataitems. On this basis, we propose an online scheduling algorithmto enhance the system performance in terms of maximizingrequest service and improving bandwidth utilization. Lastly, webuild a simulation model to evaluate the algorithm performancein a variety of situations. Experimental results demonstratethat the proposed algorithm outperforms existing algorithmssignificantly in both request serving and bandwidth utilization. I. I NTRODUCTION Recently, there has been much progress on developingcommunication technologies in vehicular networks. In par-ticular, the dedicated short range communication (DSRC)[1] protocol stack is being standardized by the IEEE 1609working group, which will be used exclusively for vehicularcommunications. It can be envisioned that many innovativeintelligent transportation systems (ITS), such as roadwayreservation systems [2], intelligent speed adaption systems[3] and autonomous intersection management systems [4],etc., will benefit from the efficient data dissemination invehicular networks. Typically, there are two communicationmodels in ITS: inter-vehicle and roadside-to-vehicle com-munications [5]. In the inter-vehicle communication model,vehicles share information with each other via the on-boardunit (OBU). One the other hand, in the roadside-to-vehiclecommunication model, information is provided from theroad-side unit (RSU), which is a fixed infrastructure installedalong the road. In practice, the two communication modelscan cooperate with each other. For example, vehicles canretrieve data items from the RSU within its service region,while vehicles can also share information among themselvesvia vehicle-to-vehicle communication. In this work, we focuson data dissemination in the roadside-to-vehicle communi-cation model.Although approaches for efficient data dissemination havebeen extensively studied in conventional mobile computingsystems ([6], [7], [8], [9]), unique characteristics in vehicularnetworks bring us new challenges. First, due to the limitedcommunication range covered by the RSU, passing vehiclescannot retrieve data items when driving out of the serviceregion. In addition, unlike many mobile applications, wherea user may be able to stay a relatively longer time in theservice region for data access (i.e. a user carries a smartphone to retrieve data items from a WIFI access point),the high mobility of vehicles imposes tight deadlines inroadside-to-vehicle data dissemination. Second, in vehicu-lar networks, a wealth of information changes frequentlyover time, such as traffic conditions, parking informationand vehicle locations, etc. Accordingly, the values of dataitems have to be updated to reflect the true status of theenvironment. Retrieving outdated data items cannot satisfythe request, and it is even potentially dangerous for somesafety-critical applications. Therefore, scheduling temporaldata items is another non-trivial issue in roadside-to-vehiclecommunication. Lastly, it is common for a request askingfor multiple dependent data items in many ITS applications.For instance, in traffic information systems, a driver mayenquire the traffic condition of different roads simultaneouslyto decide which route to take. In these applications, the querycannot be fully processed until all the requested data itemsare retrieved. With the above motivations, there is a pressingneed to design an efficient scheduling algorithm in roadside-to-vehicle communication systems to meet the dramaticallyincreasing demand for real-time data services and enhancethe overall system performance.The main contributions of this work are outlined asfollows. First, we develop a real-time roadside-to-vehicledata dissemination model by formulating the time-constraintof requests and the consistency requirement of retrievingtemporal data items. Second, we propose an on-line schedul-ing algorithm, which aims at satisfying as many requests aspossible by exploiting the broadcast effect and utilizing the 2013 IEEE International Conference on Embedded and Real-Time Computing Systems and Applications 297 978-1-4799-0850-9/13/$31.00 ©2013 IEEE  bandwidth efficiently. Third, we build a simulation modelfor performance evaluation. Two representative real-timescheduling algorithms, EDF (Earliest Deadline First) [10]and SIN (Slack time Inverse Number of pending requests)[9], are implemented for performance comparison. A com-prehensive simulation study demonstrates that the proposedalgorithm can outperform existing solutions significantly.The rest of this paper is organized as follows. Section IIreviews the related work. In Section III, we develop a real-time data dissemination model and formulate the problem.In Section IV, we propose a new scheduling algorithm.In Section V, we build the simulation model and give anextensive performance evaluation. Last, we conclude thiswork and discuss future research directions in Section VI.II. R ELATED  W ORK Current research on data dissemination in vehicular net-works largely focused on solving communication qualityrequirements at the MAC layer ([11], [12], [13], [14]).Fujimura and Hasegawa [11] proposed a Vehicle and Road-side Collaborative MAC protocol (VRCP) to support bothroadside-to-vehicle and vehicle-to-vehicle communications.Rather than considering data scheduling, the protocol de-signed two channel access modes, and investigated on modeswitching mechanisms. In particular, one is ad-hoc mode(Mode-A) with non-persistent CSMA scheme for decentral-ized vehicle-to-vehicle communication. The other is infras-tructure mode (Mode-I) with TDMA scheme for centralizedroadside-to-vehicle communication. Maeshima et al. [12]designed a MAC protocol for supporting emergency messagedelivery in roadside-to-vehicle communication systems. Thegeneral idea is that, whenever an emergency notificationoccurred on the control channel, the transmission of generalinformation on the service channel would be suspended toensure timely delivery of emergency messages. Jhang andLiao [13] proposed a Proxy-based Vehicle to RSU (PVR)communication protocol based on IEEE 802.11 DistributedCoordination Function (DCF). It is designed to mitigatequery upload contentions by electing proxy vehicles to helpthe data upload for other vehicles. Non-proxy vehicles whichattempt to communicate with the RSU must forward theirdata items to a proxy vehicle, so that it relieves the contentionfor the uplink channel and improves the throughput of thesystem. Mak et al. [14] proposed a coordinated MAC modein the presence of an RSU to compliment with ad hocapproaches when no RSU is available. It aims to improveperformance for both safety and non-safety applications bydesigning a multi-channel coordination mechanism, whichis used to minimize collisions between vehicle-to-vehicleand roadside-to-vehicle communications. The above effortsprovided solid basis with respect to improving communica-tion qualities at the MAC layer. Nevertheless, none of themaddressed data scheduling problems from the applicationpoint of view.The data scheduling problem has been extensively ex-amined in mobile computing environments. For non-real-time systems, there are a number of classical schedulingalgorithms. FCFS (First Come First Served) [15] broadcastsdata items sequentially according to their arrival order. SRPT(Shortest Remaining Processing Time) [16] chooses the dataitem with the shortest service time to broadcast. Wong [17]proposed two well-known strategies for scheduling requestsin on-demand broadcast systems: MRF (Most RequestedFirst) and LWF (Longest Wait First). MRF broadcasts thedata item which has the largest number of pending requests toaccount for the productivity of broadcast. LWF calculates thetotal time that all pending requests for a data item have beenwaiting. The data item with the longest total waiting timeis chosen for broadcast. Aksoy and Franklin [8] proposed alow overhead and scalable scheduling algorithm called RXW(Number of pending Requests Multiply Waiting time). Itcalculates the number of pending requests for a data itemmultiplied by the amount of time that the oldest outstandingrequest for that data item has been waiting. At each broadcasttick, the request with the maximum RXW value will bechosen. RXW combines the benefits of MRF and FCFS toenhance scheduling performance. It has been demonstratedthat it is the most effective algorithm in minimizing theaverage response time for non-real-time requests [18].In real-time scheduling, EDF (Earliest Deadline First)[10] is one of the foremost classical scheduling algorithms,and it has been adopted into on-demand broadcast systems.The algorithm selects and broadcasts the data item withthe shortest remaining lifetime to cater for the urgency of requests. Recently, an on-line scheduling algorithm for timecritical on-demand broadcast called SIN (Slack time InverseNumber of pending requests) [9] has been proposed. It ismotivated by two existing strategies: EDF, which considersthe urgency of requests, and MRF, which considers theproductivity of data broadcast. It has been demonstrated thatSIN can outperform other real-time algorithms significantlyin on-demand broadcast environments. In this study, wewill implement these two representative real-time schedulingalgorithms for performance comparison.III. P RELIMINARIES  A. System model The roadside-to-vehicle data dissemination model isshown in Figure 1. The system characteristics are summa-rized as follows. •  On-demand broadcast: The RSU is installed at a fixedposition (i.e. at a road intersection) in providing in-formation services to passing vehicles. The OBU ismounted on the vehicle, which enables the vehicleto submit requests and retrieve data items from theRSU. The circle represents the radio coverage of theRSU, which is called the  service region . When thevehicle enters the service region, it can submit requeststo the RSU via the uplink channel (i.e. the controlchannel as defined in IEEE 1609.4 [19]). Accordingto certain scheduling algorithms, the RSU selects anddisseminates data items from the local database to servepending requests via the downlink channel (i.e. the 298  DownlinkChannel      )     )     )     )     )     )     )     )     )     )     )     )     )     )     )     ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) ) Wired NetworksRoad-side Unit Server Updates Processing Unit Database Queries UplinkChannel             )            )            )            )            )            )            )            )            )            )            )            )            )            )            )       )       )       )       )       )       )       )       )       )       )       )       )       )       )       )       )       )       )       )       )  Fig. 1. System model of roadside-to-vehicle data dissemination service channel as defined in IEEE 1609.4 [19]). This isknown as the  on-demand broadcast   environment [20]. •  Temporal data item: In vehicular environments, a wealthof information changes frequently over time, such astraffic conditions, available parking slots and location-based services, etc. In order to maintain such highlydynamic information, the RSU is connected to the wirednetwork to form a  backbone network  , so that the dataitems stored in the local database can be kept up-to-date by the sensors and information providers from thebackbone network. In this model, we consider that thesystem updates the values of data items periodically. Insuch a manner, a data item has multiple versions andeach version only valid for a period of time, which isknown as the  temporal data item  [21]. In other words,once an update is installed for a data item, a new versionis created and its old version becomes invalid. •  Real-time multi-item request: In typical ITS applica-tions, it is common for the driver to submit requestsasking for multiple dependent data items. For instance,drivers may enquire the traffic condition of differentroads simultaneously to decide which route to take. Inthese applications, broadcasting part of the requesteddata items cannot satisfy the request and the query can-not be fully processed until all the requested data itemsare received. In addition, the vehicle cannot retrieve dataitems once it leaves the service region. This imposesa deadline to each request [22]. Specifically, a requestcan be satisfied only if all its requested data items areretrieved before its deadline expires. We call this  real-time multi-item request  .  B. Problem Formulation The set of temporal data items in the database is denotedby  D  =  { d  1 , d  2 , ···  , d  |  D | }  , where  |  D |  is the total number of data items. Each temporal data item  d  i  (1  ≤  i  ≤ |  D | ) hasmultiple versions over time due to the update, denoted by d  i (  j )  (  j  = 1 , 2 , 3 ··· ), and each version of   d  i (  j )  is characterizedby a 3-tuple:  <  V  , U  ,  E   > , where  V  ( d  i (  j ) )  is the value of  d  i (  j ) ;  U  ( d  i (  j ) )  is the update time of the current version; and  E  ( d  i (  j ) )  is the expiry time of the current version. At time t  ,  V  ( d  i (  j ) )  is valid only if   U  ( d  i (  j ) )  ≤  t   <  E  ( d  i (  j ) ) . After thetime of   E  ( d  i (  j ) ) , a new version  d  i (  j + 1 )  will be created, andthe values  <  V  , U  ,  E   >  will be updated accordingly. Giventhe update interval  l ( d  i ) , the expiry time can be representedby  E  ( d  i ) = U  ( d  i )+ l ( d  i ) . For the sake of easier expressionand emphasizing the dynamic attribute of the values of   U  ( d  i ) and  E  ( d  i ) , we denote  U  ( d  i | t  )  as the update time of   d  i  in itsversion at time  t  . Similarly, we denote  E  ( d  i | t  )  as the expirytime of   d  i  in its version at time  t  . Last, the time taken tobroadcast a data item is denoted by  τ  , which is referred toas the  service time .The request  Q  is characterized by a 3-tuple:  <  RD , ST  ,  DL  > . For the  m  th request  Q m  ( m  =  1 , 2 , 3 ··· ),  RD ( Q m )  is the set of data items requested by  Q m (  RD ( Q m )  ⊆  D ), which is represented by  RD ( Q m ) = { d  1 m , d  2 m , ··· d  |  RD ( Q m ) | m  } , where  |  RD ( Q m ) |  is the number of data items requested by  Q m  (1  ≤ |  RD ( Q m ) | ≤ |  D | ), and d  nm  ∈  D  (1  ≤  n  ≤ |  RD ( Q m ) | ).  ST  ( Q m )  is the time when  Q m is submitted.  DL ( Q m )  is the deadline of   Q m .In order to satisfy a request asking for multiple temporaldata items, the request must read the versions of the requireddata items with respect to the same time instance, which isdetermined by the time when the first data item is broadcastfor this request. Accordingly, the three conditions to serve arequest are stated as follows: •  The broadcast time of each data  d  nm  for  Q m , denotedby  BT  Q m ( d  nm ) , should be later than the time when  Q m is submitted, because the vehicle only monitors andretrieves data items after submitting the request. Denote FBT  Q m  as the time to broadcast the first data item for Q m , the first condition is represented by: FBT  Q m  >  ST  ( Q m )  (1)where  FBT  Q m  =  min (  BT  Q m ( d  nm ))  for  ∀ d  nm  (1  ≤  n  ≤|  RD ( Q m ) | ). •  Each requested data  d  nm  has to be retrieved before therequest deadline  DL ( Q m ) . The retrieval time of   d  nm  iscalculated by  BT  Q m ( d  nm )+ τ  , where  τ   is the service time.Denote  LBT  Q m  as the time to broadcast the last data itemfor  Q m , the second condition is represented by:  LBT  Q m  + τ   ≤  DL ( Q m )  (2)where  LBT  Q m  =  max (  BT  Q m ( d  nm ))  for  ∀ d  nm  (1  ≤  n  ≤|  RD ( Q m ) | ). •  All the retrieved data items have to be valid by the timewhen the request is processed (i.e. the time when thelast data item is retrieved, which is  LBT  Q m  + τ  ). At time FBT  Q m , each requested data item  d  nm  is associated witha corresponding expiry time, denoted by  E  ( d  nm | FBT  Q m ) .Given the version of each data item at time  FBT  Q m ,denote  EE  Q m ( FBT  Q m )  as the earliest expiry time of the requested data items by  Q m , the third condition isrepresented by:  LBT  Q m  + τ   ≤  EE  Q m ( FBT  Q m )  (3) 299                                                                                           Fig. 2. Timing example to serve a request where  EE  Q m ( FBT  Q m ) =  min (  E  ( d  nm | FBT  Q m ))  for  ∀ d  nm (1 ≤  n  ≤ |  RD ( Q m ) | ) in their version at time  FBT  Q m .To have better understand, Figure 2 illustrates how arequest can be satisfied by showing the relationship of thetimings.Suppose  Q m  is submitted at  t  3 , and two data items arerequested. That is,  RD ( Q m ) =  { d  1 m , d  2 m }  and  |  RD ( Q m ) |  =  2.The deadline of   Q m  is at  t  8 . The  j  th version of   d  1 m  isgenerated at  t  1 . Given the update interval  l ( d  1 m )  , the (  j + 1)th and (  j + 2) th versions of   d  1 m  will be generated at  t  4  and  t  7 respectively. Similarly, the  k   th version of   d  2 m  is generated at t  2 , and given its update interval of   l ( d  2 m ) , the ( k  + 1) th versionof   d  2 m  will be generated at  t  9 . In order to serve  Q m , accordingto the first condition, the broadcast time of   d  1 m  and  d  2 m  shouldbe later than  t  3 . In this example,  d  1 m  is the first data itembroadcasting for  Q m , which gives  FBT  Q m  = t  5 , and it satisfiesthat  t  5  > t  3 . According to the second condition, both  d  1 m  and d  2 m  have to be retrieved before the request deadline  t  8 . In thisexample,  d  2 m  is the last data item broadcasting for  Q m , whichgives  LBT  Q m  = t  6 , and it satisfies that  t  6  + τ   ≤ t  8 . Accordingto the third condition, at  t  5 ,  d  1 m  is in its (  j + 1) th version,and  d  2 m  is in its  k   th version. So, we have  E  ( d  1 m (  j + 1 ) | t  5 ) = t  7 and  E  ( d  2 m ( k  ) | t  5 ) =  t  9 . Accordingly, the earliest expiry time  EE  Q m ( FBT  Q m ) =  min ( t  7 , t  9 )  , which is  t  7 . In this example, t  6  +  τ   ≤  t  7 , which satisfies the third condition. With thesatisfaction of the three conditions,  Q m  can be satisfied bysuch a scheduling. Note that the broadcast sequence of therequested data items will not influence the service to  Q m  (i.e.broadcasting  d  2 m  before  d  1 m  can also satisfy  Q m ).However, note that  Q m  cannot be satisfied if any of itsrequested data item is broadcast during  [ t  3 , t  4 ]  (suppose thetime duration between  t  3  and  t  4  is less than the time tobroadcast two data items, namely,  [ t  3 , t  4 ]  <  2 · τ  ). This isbecause as long as  t  3  <  FBT  Q m  <  t  4 ,  d  1 m  will be in its  j th version, and hence  E  ( d  1 m (  j ) | FBT  Q m ) = t  4  . In such a case,the earliest expiry time  EE  Q m ( FBT  Q m ) =  min ( t  4 , t  9 ) , whichis  t  4 . However, it can never satisfy that  LBT  Q m  +  τ   ≤  t  4 because  [ t  3 , t  4 ]  <  2 · τ  . Therefore, broadcasting either  d  1 m  or d  2 m  during  [ t  3 , t  4 ]  will violate the third condition, causing theunsatisfactory of   Q m .With the above knowledge, this work is dedicated todesigning an on-line algorithm for scheduling real-timerequests with temporal data items in roadside-to-vehicledata dissemination systems. The objective is to enhance theoverall system performance in terms of efficient utilizationof the broadcast bandwidth, and satisfying as many requestsas possible. The primary notations used in this paper aresummarized in Table I.IV. PSU S CHEDULING  A LGORITHM In this section, we propose an on-line scheduling algo-rithm, which considers the request characteristics of   Pro-ductivity ,  Status  and  Urgency  (PSU) in scheduling. To beelaborated below, by incorporatingthese three characteristics,PSU is able to exploit broadcast effect, improve bandwidthutilization, and enhance the service chance. Before goinginto the detailed procedures of PSU, several preliminarydefinitions are stated as follows.  Definition 1:  Unserved set of a request: At time  t  , theset of unserved data items of a request  Q m  is represented by US  Q m ( t  ) =  { d  1 m  , d  2 m  , ··· d  | US  Qm ( t  ) | m  } , where  | US  Q m ( t  ) |  is thenumber of unserved data items (0 ≤| US  Q m ( t  ) |≤|  RD ( Q m ) | ),and  US  Q m ( t  ) ⊆  RD ( Q m ) .Given a pending request  Q m , if none of its requested dataitems have been retrieved ( US  Q m ( t  ) =  RD ( Q m ) ,  Q m  is calledas the  unserved request  . In contrast, if parts of the requesteddata items have been retrieved ( US  Q m ( t  )  ⊂  RD ( Q m )  and US  Q m ( t  )   =  φ  ),  Q m  is called as the  partially-served request  .Recall that each multi-item request is associated with adeadline. Meanwhile, each requested data item is associatedwith an expiry time. Therefore, there is a practical timebound to serve a request when considering both the requestdeadline and the data expiry time. Moreover, for unservedrequests and partially-served requests, they have differentattributes in terms of the time bound. In the following, weanalyze the time bound for these two categories of requests.First, we define the  tentative time bound   for the unservedrequest. This is the time bound used to indicate that if anunserved request would start its service at time  t  , then whatis the latest time to complete the service.  Definition 2:  Tentative time bound: At time  t  , if   Q m is an unserved request, the tentative time bound for  Q m ( TTB Q m ( t  ) ) is either its request deadline (  DL ( Q m ) ), or theearliest expiry time of its requested data items at time t   (  EE  Q m ( t  ) ), whichever is earlier. That is,  TTB Q m ( t  ) = min (  DL ( Q m ) ,  EE  Q m ( t  )) , where  EE  Q m ( t  ) =  min (  E  ( d  nm | t  ))  for ∀ d  nm  (1  ≤  n  ≤ |  RD ( Q m ) | ) in their versions at time  t  . TTB Q m ( t  )  is not the finalized time bound of   Q m , and thevalue of   TTB Q m ( t  )  may change with  t  . The reason is that, ithas not yet determined to serve  Q m  at  t  . Although the requestdeadline  DL ( Q m )  is a constant, the value of   EE  Q m ( t  )  mayvary with time. Therefore, at different scheduling points, thedynamic value of   EE  Q m ( t  )  may result in different  TTB Q m ( t  ) .Considering the example shown in Figure 2, if   t   is in [ t  3 , t  4 ] , then  TTB Q m ( t  ) =  min (  E  ( d  1 m (  j ) | t  ) ,  DL ( Q m )) , which is t  4 . In contrast, if   t   is in  [ t  7 , t  8 ] , then  TTB Q m ( t  )  changes tomin (  E  ( d  2 m ( k  ) | t  ) ,  DL ( Q m ))  , which is  t  8 .Note that the tentative time bound is only the attributeof unserved requests. As soon as the service starts for  Q m ,the versions of its requested data items, as well as thecorresponding expiry time of these data items have beendetermined. In other words, for any requests which have been 300  TABLE IS UMMARY OF NOTATIONS Notations Descriptions Notes  D  set of temporal data items  D  =  { d  1 , d  2 , ···  , d  |  D | } d  nm  the  n  th data requested by  Q m  d  nm  ∈  DU  ( d  i | t  )  update time of   d  i  in its version at  t U  ( d  i | t  ) ≤ t  E  ( d  i | t  )  expiry time of   d  i  in its version at  t E  ( d  i | t  )  >  t l ( d  i )  update interval of   d  i  U  ( d  i | t  )+ l ( d  i ) =  E  ( d  i | t  ) τ   service time of a data item  RD ( Q m )  set of data items requested by  Q m  RD ( Q m ) = { d  1 m , d  2 m , ··· d  |  RD ( Q m ) | m  }  DL ( Q m )  deadline of   Q m  BT  Q m ( d  nm )  broadcast time of   d  nm  for  Q m FBT  Q m  time to broadcast the first data for  Q m  FBT  Q m  =  min (  BT  Q m ( d  nm ))  EE  Q m ( t  )  earliest expiry time of the requested data  EE  Q m ( t  ) =  min (  E  ( d  nm | t  )) US  Q m ( t  )  unserved set of   Q m  US  Q m ( t  )  ⊆  RD ( Q m ) TTB Q m ( t  )  tentative time bound for an unserved  Q m  TTB Q m ( t  ) =  min (  DL ( Q m ) ,  EE  Q m ( t  ))  DTB Q m  determined time bound for a partially-served  Q m  DTB Q m  =  min (  DL ( Q m ) ,  EE  Q m ( FBT  Q m )) SQ ( t  )  set of schedulable requests  EDP d  i ( t  )  effective data productivity of   d  i  ERP Q m ( t  )  effective request productivity of   Q m scheduled to serve, their time bounds have been determined.In view of this, we define the  determined time bound   for thepartially-served request as follows.  Definition 3:  Determined time bound: If   Q m  is apartially-served request, and the time to broadcast thefirst data item for  Q m  is  FBT  Q m , then the determinedtime bound for  Q m  (  DTB Q m ) is either its request deadline  DL ( Q m ) , or the earliest expiry time of its requested dataitems at time  FBT  Q m  (  EE  Q m ( FBT  Q m ) ), whichever is ear-lier. That is,  DTB Q m  =  min (  DL ( Q m ) ,  EE  Q m ( FBT  Q m )) , where  EE  Q m ( FBT  Q m ) =  min (  E  ( d  nm | FBT  Q m ))  for  ∀ d  nm  (1  ≤  n  ≤|  RD ( Q m ) | ) in their versions at time  FBT  Q m .Note that  DTB Q m  is a determined value, which will notchange with time. At each scheduling point, the system mayhave both unserved and partially-served pending requests.In order to check whether a request has the chance to besatisfied at time  t  , we define the  schedulable request   ( SQ ( t  ) )as follows.  Definition 4:  Schedulable request: At time  t  , denote SQ ( t  )  as the set of requests which are schedulable. It isconstructed by the following two subsets of requests ( SQ  1 ( t  ) and  SQ  2 ( t  ) ): •  Case a):  Q m  is an unserved request. It is schedulableif   Q m  can retrieve all of its requested data items be-fore the tentative time bound  TTB Q m ( t  ) . The requestsubset is represented by  SQ  1 ( t  ) =  { Q m | US  Q m ( t  ) =  RD ( Q m )  and t   +  |  RD ( Q m ) | ·  τ   ≤  TTB Q m ( t  ) } , where |  RD ( Q m ) |  is the total number of requested data items,and  τ   is the service time of a data item. •  Case b):  Q m  is a partially-served request. It is schedula-ble if   Q m  can retrieve the remaining requested data itemsbefore the determined time bound  DTB Q m . The requestsubset is represented by:  SQ  2 ( t  ) =  { Q m | US  Q m ( t  )  ⊂  RD ( Q m )  and US  Q m ( t  ) !  =  φ   and t   +  | US  Q m ( t  ) | · τ   ≤  DTB Q m } , where  | US  Q m ( t  ) |  is the number of unserveddata items.To sum up,  SQ ( t  ) =  SQ  1 ( t  ) ∪ SQ  2 ( t  ) , and  SQ ( t  )  ⊆  Q ( t  ) ,where  Q ( t  )  is the set of all pending requests at time  t  .In the following, we introduce how the PSU captures therequest productivity, status and urgency in scheduling to ex-ploit the broadcast effect, improve the bandwidth utilizationand enhance the request service chance, respectively. •  A. Exploit broadcast effect In broadcast environments, disseminating one data itemcan to serve all the requests which are pending for it.Intuitively, scheduling the data item with more pendingrequests has the potential to achieve higher  data productivity ([8], [9]). PSU considers the data productivity in makingscheduling decisions for the sake of exploiting the benefit of broadcast effect. Nevertheless, as illustrated in Def. 4, notall the pending requests are schedulable at each schedulingpoint. Therefore, different from the previous study, whichconsiders all pending requests when computing the dataproductivity, we define the  effective data productivity  asfollows.  Definition 5:  Effective data productivity: At time  t  , theeffective data productivity of   d  i  (1  ≤  i  ≤ |  D | ), denoted by  EDP d  i ( t  ) , is the number of requests in  Q d  i ( t  ) , where the set Q d  i ( t  )  is constructed by any  Q m  which satisfies: a)  d  i  is in theunserved set of   Q m ; b)  Q m  is schedulable at time  t  . That is, Q m ∈ Q d  i ( t  )  if and only if   d  i ∈ US  Q m ( t  )  and  Q m ∈ SQ ( t  ) . Withthe constructed set of   Q d  i ( t  ) , we have  EDP d  i ( t  ) =  Q d  i ( t  )  .Accordingly, for a multi-item request, the  effective request  productivity  is defined as follows.  Definition 6:  Effective request productivity: At time 301

[2015] ITSC.pdf

Aug 6, 2018


Aug 6, 2018
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks