PFPS: Priority-First Packet Scheduler for IEEE 802.15.4 Heterogeneous Wireless Sensor Networks

This paper presents priority-first packet scheduling approach for heterogeneous traffic flows in low data rate heterogeneous wireless sensor networks (HWSNs). A delay sensitive or emergency event occurrence demands the data delivery on the priority basis over regular monitoring sensing applications. In addition, handling sudden multi-event data and achieving their reliability requirements distinctly becomes the challenge and necessity in the critical situations. To address this problem, this paper presents distributed approach of managing data transmission for simultaneous traffic flows over multi-hop topology, which reduces the load of a sink node; and helps to make a life of the network prolong. For this reason, heterogeneous traffic flows algorithm (CHTF) algorithm classifies the each incoming packets either from source nodes or downstream hop node based on the packet priority and stores them into the respective queues. The PFPS-Earliest Deadline First (PFPS-EDF) and PFPSFirst Come First Serve (FCFS) algorithms present scheduling for each data packets using priority weight. Furthermore, reporting rate is timely updated based on the queue level considering their fairness index and processing rate. The reported work in this paper is validated in ns2 (ns2.32 allinone) simulator by putting the network into each distinct cases for validation of presented work and real time TestBed. The protocol evaluation presents that the distributed queue-based PFPS scheduling mechanism works efficiently using IEEE CSMA/CA. Keywords: wireless sensor networks, priority-based data delivery, buffer management, packet scheduling, heterogeneous traffic flows. 1. Introduction & Background The IEEE sensor network [1] is an emerging field and nowadays, it is being used widely for small scale applications due to its low resource needs and longer life features [2], and it approximately lasts up to 10 years using AA batteries. To name a few it covers process industry, healthcare [3], transportation, residential projects, tracking, monitoring [4-6] and much more. Thus it creates the scope for active researchers to design the flexible diversified communication network which can handle the multiple event data simultaneously dynamically. Generally, the occurrence of events in LR-WSNs is unpredictable; therefore, having the provision of data delivery mechanism with sensor MAC sublayer becomes an essential and necessary. But, developing a data carrier protocol to address such different aspects altogether is a truly challenging task in low data rate, low power, low processing capability, limited memory, and low transmission coverage network. The sensor network comprises the delay sensitive (for instance, body sensor networks, process assembly, etc.) and delays tolerant (for instance, monitoring, and tracking objects) applications. The existing contention-based protocols, namely S-MAC [7], T-MAC [8], B-MAC [9], XMAC [10], and Wise-MAC [11] are developed to address the problem of data delivery; still undergoes from the collision, unstable behavior due to sudden traffic load, and topology deviation. In [12], [13], and [14]; the slotted CSMA/CA of beacon-enabled sensor network protocols are presented to mitigate delay of emergency applications. To address these types of applications simultaneously, the complexity level of data processing is handled at various data collection points in the distributed sensor networks using CSMA/CA. Therefore, there is need for improvising the provision of data delivery decision at various intermediate levels instead of at a sink node. The scope of this paper covers following background cases for the PFPS using EDF (Earlier Deadline First) and FCFS (First-Come-First-Served) algorithm. Case#1 (Classification of heterogeneous information): In a mesh topology, many sensing devices are connected at various levels. The unique data is generated by each source node and delivers it via multiple hops. Here, the complexity of data transmission increases, therefore, the classification of data packets sent by networked source nodes to each hop becomes the necessity. Therefore, this case is designed as a part of the categorization of heterogeneous traffic flows algorithm (CHTF). For this reason, the dual queue is implemented to store priority traffic and regular traffic separately. Case#2 (Priority Assignment): The impact of static priority assignment and dynamic priority assignment over underlying MAC protocol shall be taken into consideration for differentiating the traffics. In FCFS, the priority assignment is kept static which does not change over multiple hops. However, in EDF approach, the priority metric is designed to update priority based on hop distance and delay. Case#3 (Queuing system for multi-traffic flows): A decentralization approach is the need of IEEE sensor networks which reduces the load on sink node. For this reason, considering middle-level processing nodes i.e. hop performs important job in the dense sensor environments. A packet level in each queue is the key parameter to prevent the buffer overflow. The reporting rate is updated time to time, based on the queue level. The prime objective is to prevent congestion so that a delay sensitive application in the multi-event environment does not suffer. The existing research focuses particularly on improving the packet delivery ratio; however, though it is important; still reducing the sensitive packet loss is also the necessity from the application context. This approach is incorporated at various actor points of the network to shrink weight of a sink. Case #4 (Reporting rate): Managing the data reporting rate of each contributing source node is a significant topic of research since the inception of LR-WSNs. The problems like packet loss, congestion, an network transmission time, and excess power usage arise due to improper design of reporting rate mechanism. Therefore, to handle heterogeneous traffic flows, the flexible reporting rate mechanism is designed using a dual queue and their 254 International Journal of Communication Networks and Information Security (IJCNIS) Vol. 9, No. 2, August 2017 processing rate. The key objectives of PFPS protocol are: 1) to reduce end to end delay of long distance priority flows 2) to prevent bottleneck problem, and 3) to increase network life. Figure 1 shows network communication model consisting of multiple hops in-between sensors and the sink node. Hops are considered the additional capability of processing the data packets in addition to receiving and transmitting the packets. The existing work covers mainly queue management, beacon-enabled slotted CSMA/CA, and beaconless CSMA/CA protocols for low rate IEEE 802.15.4 networks. In [15], congestion control and information prioritization approaches are presented for monitoring real-time data of the patient in wireless biosensor networks. The congestion control approach is applied at parent node whereas priority assignment of network bandwidth is applied over child nodes. The service differentiation is designed according to physical phenomena of a patient. The priority in node (PRIN) MAC protocol [16] is designed to prioritize the information using buffer management with one-hop network topology. The static priority is assigned to each node. RushNet [17] protocol presents traffic prioritization mechanisms for low priority (delay tolerant) and high priority (latency sensitive) types of applications. The throughput is improved using a token passing method to minimize the contention and avoid traffic jam, and multihop approach used to trim down the propagation delays. In [19], the author proposes the packet allocation rate viz., traffic jam discovery and intimation, and congestion control to avoid the buffer overflow problem. ECODA [20] uses dual buffer management approach to achieve the data transmission requirement of transient traffic and locally sensed traffic at each hop level. The congestion problem is resolved using the weighted buffer and flexible queue scheduler to prioritize the information. The PCCP congestion avoidance protocol [21] discusses the node priority index for increasing its significance in the network. It detects traffic jam degree according to packet arrival period and processing rate. In [22], EDF and FP algorithm are presented for urban traffic application with considering the various intersection points of the road. The performance of both algorithms is compared against each other on isolated traffic intersection points. Analysis model illustrates that EDF algorithm performs well against FP algorithm in terms of delay, a number of stops, and means the trip time of priority vehicles. The shortest-first CSMA/CA [23] approach presents the solution for “energy hole problem” near to the sink. The length detection and anti-starvation mechanism are proposed. The nodes that are having the small data size packets are considered as high priority nodes whereas long size data packet holding nodes are defined with low priority nodes. To control periodic flows and network management control flows [24] management using guaranteed time slot approach. In [25], various MAC protocols have been surveyed to understand collision preventing techniques and achieving the greater reliability. A different synchronous, asynchronous, and hybrid protocols are thoroughly studied. This survey helps to design parameters of MAC protocols and existing state of the art work. At the end, authors have also put forth critical issues for open discussions. In [26], MMEDD presented for efficiently delivering the essential information in the wireless sensor networks for optimizing the power consumption. A hybrid operating system Contiki is used due to the low layer called Rime for reducing the energy consumption to the great extent. A pre-emptive multithreading module is applied for managing the multiple tasks parallels. Results are validated comparing with traditional approaches i.e. without multithreading mechanisms. Protothreads are used instead of threads. Authors have made claimed approximately around 9% higher than the classical approach. The proposed architecture is based on EDF and shows considerable improvements over traditional CSMA/CA MAC protocol. The term sink instead of the base station and actor node instead of hop node are alternatively used in the further discussions. The contributed strategies are summarized as follows. The CHTF algorithm presents the classification of heterogeneous traffic flows. It classifies the packets based on their priority level and stores either in the regular queue or in the priority queue. The PFPS proposes the priority first approach using dual queue management with preemptive strategy. The queue levels are defined according to their traffic type. The FCFS and EDF approaches are used for packet transmission to the next level. The priority metric is used to update the priority of a packet incrementally towards the base station to service long distance data-first over newly sensed flows at each actor node. The mathematical model is designed and developed to make it functional with respect to proposed operational steps of CHTF and PFPS algorithms in the network. Finally, the results are validated by performing various simulation cases in the ns2 (ns2.32allinone) discrete event simulator using CSMA/CA. Finally, PFPS is implemented over real time TestBed (6+1) with 24 sensors in total. The performance is examined and compared with the FCFS approach. The residual sections of this paper are organized as follows. Chapter-2 describes about the proposed PFPS protocol description. Explanation of performance analysis is given in chapter 3. Furthermore, outcomes are concluded and present the future scope. 2. PFPS Protocol Description 2.1 Network Model and Assumptions A communication model comprises source nodes, hop, and the base station. A source node is responsible for only sensing, delivering and receiving the information to and from its upstream node. A hop node and sink node are capable of sensing, processing, receiving, and transmitting the information. There are two main types of flows to be considered for protocol design, namely priority traffic flow (for e.g. saturation level) and regular traffic (for e.g. temperature). The size of data packets for both traffics is same. The nodes are placed randomly and AODV routing protocol is used for network formation. Each level i.e. at hop node, two queues is designed to store the regular and priority-based traffics. The incoming traffics are put in respective queues and are managed using First Come First Served (FCFS) mechanism or Earliest Deadline First (EDF) 255 International Journal of Communication Networks and Information Security (IJCNIS) Vol. 9, No. 2, August 2017 mechanism. A variable fairness index is achieved using PFPS algorithm. A classification of heterogeneous traffic flow (CHTF) algorithm performs filtration of incoming packets of different traffic flows and stores either in priority traffic queue or regular traffic queue. In overflow condition, data packets are not stored in the queue. In PFPS algorithm, transmission priority is given more to the priority-based traffic of queue-2 at each hop. Considering the sensitivity or criticality of a particular traffic flow, it is necessary to deliver at the earliest. For this reason, the decision of priority packet transmission is taken based on queue condition. However, if delay sensitive packets are not delivered within a deadline, then the data becomes useless or network usage is turn out and considered as wastage of resources. Therefore, PFPS algorithm is proposed to address the delay and deliver packets using effective use of queuing operations to the great extent. Descriptions of mathematical terms are explained in Table 1. 2.2 PFPS Protocol Implementation In this section, two algorithms are discussed, namely CHTF and PFPS. In order to implement the priority based data transmission approach, the design of two queues is presented. The queues are priority queue and regular queue implemented on the hop. The regular queue is designed for regular traffic flow (for example, temperature) and priority queue is designed for priority traffic flow i.e. time constraint packets (for example, O2 saturation level monitoring) as shown in algorithm-1. Based on different input flows; the fairness index is examined. If the applications are having hard timing constraints then queue would be developed from priority perspective which will be a better option to achieve the deadlines. Each queue is comprised of different traffic(s) at a given time at hop is described in equation (1). The total traffic load (T , ) at hop k, either q or q T , = ∑ T , q , , (1) Where let q be a queue either type 1 (regular traffic) or 2 (priority traffic), i = 1, 2 . The total incoming traffic load on any particular hop node (T !",#) using both queues is as described in equation (2). T $,# = λ# = ∑ T $ r%; p% ( ( );% ( ($;% = ∑ *∑ (T $ , , 1) ) , ∑ (T $ . , 2) $ / % , (2) The quantity of packets delivered by a hop is expressed in equation (3). T 0 ,# = μ# = ∑ T 0 r%; p% ( ();% ( ($;% = ∑ ∑ (T 0 , , 1) ) , ∑ (T 0 , 2) $ % , (3) The remaining number packets in both queues are defined as T ) at a particular instance as measured in equation (4). T ) = λ# − μ# (4) The overall probability of packet processing of any particular hop is mentioned in equation (5). P# = μ λ ≤ 1 (5) The probability of each hop is computed individually to examine the affecting attributes of both queues separately. The probability of q and q for hops (k = 1,2,3, ... , n); (P#) are as explained in equation (6). The probability of processing packets at each stage of the hop is defined in decreasing order according to success rate. The values move from 1 towards 0, when the hop count goes higher because of traffic load for each flow. However, the decreasing factor is less in the case of priority traffic flows. The distance of one hop indicates that connected node to hop one is a one hop farthest from a sink node. P# ≥ P#; , P#; , ... , ≥ P ≥ P ; P#(hops) = >μ? λ?@A λA ,...@μ B? λ B?@ λ C DE (6) The overall average probability of all contributing hops is stated in equation (7). PD F = ∑ μGH.I λGH.I GH.IJ? # ≤ 1 (7) The objective function is enhancing priority based traffic ratio simultaneously maintaining a fairness index at a reasonable level. This approach is designed in the viewpoint of heterogeneous nature of LR-WSNs. In order to handle them simultaneously, the queuing system plays a non-trivial role to attain a target of each specific event requirement. For Table 1. Summary of Mathematical Notations Term Definition Term Definition T flow type T ) remaining packets in queues T , regular traffic flow p# priority of k hop T . priority traffic flow μ#,% average incoming packet rate in a queue i q queue includes 1 and 2 λ#,% average outgoing packet rate in a queue i k number of hops q " lower limit of queue i T , traffic flow at hop k in a type queue i q 0 upper limit of queue i T $,# incoming traffic at k hop q )!K maximum limit of queue i T 0 ,# outgoing traffic at k hop q F L size of queue i T $ , incoming regular flow R reliability of regular traffic T $ . incoming priority flow R reliability of priority traffic T 0 , outgoing regular flow α, β, γ, queue processing tuning parameters T 0 outgoing priority flow δ! additive & multiplicative increase tuning parameters m number of regular traffic nodes δN additive & multiplicative decrease tuning parameters n number of priority traffic nodes η, η , η packet count variables r% regular traffic in queue1 P range variable p% priority traffic in queue1 f" reporting frequency level R, S Priority weight parameters (0.5,0.05) TUVW Energy consumption for all control packets XYZVU Node processing time TU[ Energy consumed for data & control packets TYZVU Energy for processing 256 International Journal of Communication Networks and Information Security (IJCNIS) Vol. 9, No. 2, August 2017 that reason, dual queue model is designed which is based on number of packets in regular and priority queues, defined as η , η respectively. The frequency rate is updated based on the buffer level and processing rate, as expressed in equation (8). \ ≤ P ≤ 1; ]^1 ≤ _`,a? b`,a? ≤ ⍺ d || ^1 ≤ _`,aA b`,aA ≤ β d || ^1 ≤ _`,a?,A b`,a?,A ≤ \ d ˄ ((f g ≤ h < f jkl) || (f g ≤ h < f jkl))m ; (n1, o1) ⍺ ≤ P ≤ \ ; ]^1 ≤ _`,a? b`,a? ≤ ⍺ d || ^1 ≤ _`,aA b`,aA ≤ β C || ^1 ≤ _`,a?,A b`,a?,A ≤ \ d ˄((f p ≤ h < f g)||(f p ≤ h < f g))m ; (n2, o2) ⍺q ≤ P ≤ \q; ]^1 ≤ _`,a? b`,a? ≤ ⍺qd || ^1 ≤ _`,aA b`,aA ≤ βqC || ^1 ≤ _`,a?,A b`,a?,A ≤ \qd ˄((0 ≤ h < f p )||(0 ≤ h < f p ))m ; (n3, o3) uvhxyz{|x; (o4) (8) The various conditions of queue level and processing rate are a1 to a3 as described below. The additive increase (from case a1 to a3) is applied in the following cases. a1: f"@ ← f" ∗ δ!? , γ ≤ P ≤ 1 ; a2: f"@ ← f" ∗ δ!A , ⍺ ≤ P ≤ γ ; a3: f"@ ← f" ∗ δ!‚ , ⍺q ≤ P ≤ γq The multiplicative decrease is applied in the following cases from b to b„. b1: f"@ ← f" δN? … , P < γ ; b2: f"@ ← f" δNA … , P < γ ; b3: f"@ ← f" δN‚ … , P < γ ; b4: f"@ ← f" δN† … Let α, β, and γ be the tuning parameters for achieving the desired reliability of traffic flows. The values of ‡ are 0.93, 0.66, and 0.33 of ‡ , ‡ , and ‡q respectively. The values of β are 0.96, 0.766, and 0.5 for β1, β2, and β3; respectively. Let γ be processing tuning attribute with different values such as 0.945, 0.716, and 0.415 for \ , \ , and \q respectively. Finally, the values of ˆk are 1.05, 1.10, 1.15 for ˆk?, ˆkA , and ˆk‚; respectively and the values of ˆ‰ are 1.15, 1.10, 1.01 for ˆk? , ˆkA , and ˆk‚; respectively. However, the value of ˆ‰†is 1.20. The priority metric (Š‹Œ) is used to compute the priority level of each outgoing packet at actor node. The static priority or previous priority is considered and its delay () from originating node to current node is taken into consideration. The hop count (hU) indicates the number of hops away from the base station. It is expressed in equation 9. Š‹Œ = Š‹Œ + ‘ ’[ (9) For example, Š‹Œ(5) = 1 + ”.•∗• ”.”•∗”.” q = 3.55 Š‹Œ(4) = 3.55 + ”.•∗„ ”.”•∗”.” – = 5.81 And so on. Algorithm-1 (CHTF): It classifies all incoming packets according to their priority level and stores into respective queues. There are mainly two queues designed, the queue-1 is for regular traffics, for example, temperature and queue-2 are for priority traffic, for example, O2 saturation level detection. Algorithm-1 (CHTF): Classification of Heterogeneous Traffic Flows Input: ˜™, ˜š <: ˜›œš Output: žŸ, ← ¡¢£¤¥ Prerequisites: Priority Assignment to each packet of different traffic flows during packet formation at source node i.e. RFD Begin 1. h ← 0; h ← 0; h ← 0; 2. u 3. h = h + 1; 4. ¦§ (XZ ∈ XŒ©Ya)˄(h < f jkl) u 5. f ← f + XZ; h = h + 1; 6. x«|x{§ (XY ∈ XŒ©Ya)˄(h < f jkl) u 7. f ← f + XY; h = h + 1;// 8. x­ {§ 9. zh{«x(XŒ©Ya! = ̄°±±) 10. x­ u − zh{«x «uuŠ End They are categorized into two types of flows, namely priority traffic flow and regular traffic flow. The overall incoming rate η is maintained for various cases mentioned in PFPS algorithm as well as making the decision of delivery rate with respect to load on both queues. The individual packet level (η , η ) for each queue computed separately in order to fasten the data transmission based on lower, upper, and maximum threshold levels. The packet delivery preference is given first to priority-based traffic i.e. O saturation level data as it is considered to be an important and critical of queue-2. However, while transmission of packets the fairness index is achieved at great extent by managing the queuing decision within threshold limits. Before storing into the respective queue, each packet and the queue limit is checked. This process is continued until packets are coming at each hop. Algorithm-2 (PFPS): The various levels of regular and priority queues are expressed in point a1 and a2, respectively.The levels are the part of PFPS design in order to maintain every hop from congestion free with the priority-first approach. For this reason, the queue level differentiation is taken into account. The purpose is to serve the emergency traffic first or delay sensitive traffic-flow first instead of delay tolerant traffic. Moreover, the queue level thresholds are defined according to the type of flow. The more weight is given 257 International Journal of Communication Networks and Information Security (IJCNIS) Vol. 9, No. 2, August 2017 to high priority flow instead of regular traffic. The threshold limit is increased little over the limits of regular traffic queue just to get more time for data packet delivery. In addition, the one level up facility is also preferred to high priority traffic queue. Both queues are defined with three levels, namely lower limit, upper limit, and maximum limit. Algorithm-2(PFPS): Priority First Packet Scheduler Input: (˜™ ∈ 3Ÿ)||(˜š ∈ 3 ) Output: traffic flow based reliability i.e. ́™, ́š <: ́›œš Prerequisites: Priority Assignment to each packet of different traffic flows during packet formation at source node i.e. RFD Begin 1. u 2. {§ (((h ≤ f p ) ˄ ((h ≤ f p )||(f p ≤ h ≤ f g))) u 3. transmit XY ← f ; dequeue(XY); h = h − 1; 4. x«|x{§(((f p < h ≤ f g)|| (f g < h ≤ f jkl)) ˄ (h ≤ f p ))u 5. vyn­| ̧{v(XZ) ← f ; xf1x1x(XZ); h = h − 1; 6. x«|x{§((f g < h ≤ f jkl)˄ ((f p ≤ h ≤ f g)||(f g < h < f2 ̧no)) 7. yuŠ(XZ) ← fZ; h = h − 1; // 1­v{« {v »ux| ox«uz f g «x1⁄4x« 8. vyn­| ̧{v XY ← f ; xf1x1x XY ; h = h − 1; 9. x«|x{§(f == x ̧Šv1⁄2 ˄ f ! = x ̧Šv1⁄2) u 10. vyn­| ̧{v XY ← f ; xf1x1x(XY); h = h − 1; 11. x«|x{§(f !! = x ̧Šv1⁄2 ˄ f == x ̧Šv1⁄2) 12. vyn­| ̧{v(XZ) ← f ; xf1x1x(XZ); h = h − 1; 13. x­ {§ 14. zh{«x((f ||f )! = x ̧Šv1⁄2) End The decision is depended on the number of packets in each queue. However, the processing rate of the queue is also computed to ensure the transmission of data packets using CSMA/. The data transmission bandwidth limit is set to 250kbps according to standard in protocol configuration. The levels of the queue are defined as follows. f jkl = ¿ .À q f ÁÂÃaÄ ; f g = ¿ q f ÁÂÃaÄ ; f p = ¿ q f ÁÂÃaÄ (10) f jkl = Å2.9 3 f ÁÂÃaÇ ; f g = Å2.3 3 f ÁÂÃaÇ ; f p = Å1.5 3 f ÁÂÃaÇ (11) These various levels of the queue are used for transmitting the appropriate data packets. The overall energy consumption (T) is the sum of energy consumed in various steps involved during in-network processing, as described in equation (12). The energy consumption is directly proportional to energy consumed for transmission of data and control packets (TU[); and processing overheads (TYZVU) at a node. TŒVŒkp = TU[ + TYZVU (12) The energy consumed in waiting time (ȋŒ) is the time required to gain the channel access by performing the number of times backoff (X‰U`VÉÉ) i.e. symbols used, as mentioned in equation (13). T‹Œ = X‰U`VÉÉ ∗ ȋŒ (13) The total power transmit of a node is the sum of total number of packet transmitted (ÊY`Œ) multiplied by the power consumed for each outgoing data packet (Ȍl), as expressed in equation (14). TŒl = ÊY`Œ ∗ Ȍl (14) The total energy consumed for data packets at a node is computed using two parameters, namely power consumed (ÈZl) for a single receiving packet and total number of data packets received (hY`Œ), mentioned in equation (15). TZl = ÊY`Œ ∗ ÈZl (15) The total energy consumed for network control packets (TUŒZp) of a node is the sum of number of packets (ËÈUŒZp) handled and power taken for receiving and transmitting each packets, as expressed in equation (16). TUŒZp = ËÈUŒZp ∗ (ÈZl + Ȍl) (16) The overall energy consumed (T) is the addition of energy consumed of data packets and control packets, as described in equation (17). T = TŒl + TZl+TUŒZp (17) The energy consumption for packet processing at a node depends on how much time it takes to process per unit time, as stated in equation (18). TYZVU = TUVW + XYZVU (18) 3. Performance Analysis 3.1 Simulation Setup & Result Discussions The simulation experimentations were conducted in the ns2 simulator (2.32 ns-allinone version). The network parameters ion and setting are summarized in Table 2. The simulations are performed by varying number of nodes, interval time, and simulation time. The each network performance metric is validated in different distinct cases. Figure 1-3 describes energy consumption over varying node densities, interval time, and simulation time period. It is observed that 18% less average energy is consumed compared with the FCFS approach, as shown in figure 1 , 12% less average energy is consumed over different time intervals as described in figure 2, and 9% less average power is utilized over different simulation time setup as depicted in figure 3. Table 2. Network Attribute Summary