Analysis of the scalability of hierarchical IEEE /Zigbee networks 1 Departamento de Tecnología Electrónica. University of Málaga ETSI de Telecomunicación, Campus de Teatinos, – Málaga- Spain Third International ICST Conference on Scalable Information Systems (Infoscale 2008) E. Casilari, A. Flórez-Lara, J.M. Cano-García UNIVERSIDAD DE MÁLAGA, SPAIN Vico Equense (Italy), 4 th June 2008
Analysis of the scalability of hierarchical IEEE /Zigbee networks 2 Index 1.Introduction: WPANs and /Zigbee 2.Overview of IEEE Strategies to avoid beacon collision 4.Results 5.Conclusions
Analysis of the scalability of hierarchical IEEE /Zigbee networks 3 Introduction: /Zigbee Standards IEEE (PHY and MAC) and Zigbee jointly describe a protocol stack for the definition of Wireless Personal Area Networks (WPAN). Aimed at providing solutions for low-cost wireless embedded devices (transceivers under 1$) with consumption and bandwidth limitations Low rate (up to 250 Kbps), short range (up to 10 m) communications In immature state but appealing candidate to support a wide set of services, particularly for low consume domotic sensor networks (although real time services are also contemplated for services such as voice or biosignals) Main challenge of /Zigbee: potentiality to set up self-organizing (ad hoc) networks capable of adapting to diverse topologies, node connectivity and traffic conditions. Advantages of mainly depend on the configuration of MAC sublayer
Analysis of the scalability of hierarchical IEEE /Zigbee networks 4 Operation modes of The MAC layer of IEEE enables two alternative operational modes: 1. Non beacon-enabled (point-to-point) mode: Access control is governed by non-slotted CSMA/CA Higher scalability but nodes must be active all time (elevated power consumption) Real time constraints cannot be guaranteed 2. Beacon-enabled mode, A coordinator node periodically sends beacons to define and synchronize a WPAN formed by several nodes Nodes can wake up just in time to receive the beacon from their coordinator and to keep synchronized (power efficiency) Synchronization permits to guarantee time slots (resources) to delay sensitive services Main problem: scalability → Time must be divided between clusters
Analysis of the scalability of hierarchical IEEE /Zigbee networks 5 Configuration of beacon enabled networks Two classes of nodes: the so-called Full-Function Devices (FFD) and the Reduced-Function Devices (RFD). Star topology: A FFD performs as the network ‘coordinator’, in charge of the communications of a set (or ‘cluster’) of RFD nodes (the ‘children’ nodes). The coordinator periodically emits a beacon to announce the network and to keep children synchronized Beacon Interval (BI), divided in an active part and an inactive part. Active part consists of a ‘Superframe’ of 16 equally-spaced time slots. Contention Free Period (CFP): guaranteed slots for certain nodes Contention Access Period (CAP): nodes compete for the medium access All the transmissions take place during the Superframe Duration (SD) In the inactive period all nodes (including the coordinator) may enter a power saving mode to extend the lifetime of their batteries
Analysis of the scalability of hierarchical IEEE /Zigbee networks 6 Structure of a superframe Where a = 15.36, 24 or 48 ms when a rate of 250, 40 or 20 kbps is employed Configuration of BO and SO: trade-off BO >> SO: almost all BI corresponds to the inactivity period, high power saving, low rate can be achieved Other case: lower power saving but higher rate
Analysis of the scalability of hierarchical IEEE /Zigbee networks 7 Zigbee Cluster-trees Apart from the tree networks with a single coordinator, the Zigbee standard permits the association of cluster coordinators to form cluster-trees. One of the coordinator nodes assumes the central role: PAN or Zigbee Coordinator (ZC). The rest of the coordinators are Zigbee Routers (ZRs) ZRs responsible for retransmitting the data from any ‘child’ node (leaf) within their clusters Zigbee specification does not impose any protocol nor algorithm to create this type of networks Existing commercial compliants modules do not support the formation of cluster-tree topologies Coexistence of more than one coordinator → possibility that beacons (simultaneously emitted by two adjacent coordinators) get lost due to collisions. Beacon collision provokes children to desynchronize from the router
Analysis of the scalability of hierarchical IEEE /Zigbee networks 8 Strategies to avoid beacon collision (I) IEEE Task Group 15.4.b has proposed two generic strategies to cope with beacon collision 1. Beacon-only period: a time window that is specifically reserved for the transmission of all the beacons in the network. Advantages: superframe duration of each cluster can be designed with independence of the rest Problems: -It modifies the superframe structure of the standard - The coexistence of active periods of different clusters augments the possibility of packet collision while it prevents the implementation of Guaranteed Time Slots
Analysis of the scalability of hierarchical IEEE /Zigbee networks 9 Strategies to avoid beacon collision (II) 2. Sequencing of the beacons and Superframes: in non-overlapped periods during the Beacon Interval Advantages: Standard is respected, GTS can be implemented Problems: scheduling of beacons within the different Beacon Interval and especially the duration of the superframes must be carefully designed. Otherwise: serious problem of scalability
Analysis of the scalability of hierarchical IEEE /Zigbee networks 10 Objective Assumptions: Pessimistic case: Any node can interfere the rest, no radio planning (all nodes transmit in the same channel)→ Superframes cannot overlap Hierarchical cluster-tree, all traffic flowing to the ZC (typical case of a sensor network) Problem to solve: to define the superframe durations ( SO i ) of the clusters Objective: to maximize the utilization of the BI Condition to be accomplished in any case (for a network of N C coordinators: routers+ZC):
Analysis of the scalability of hierarchical IEEE /Zigbee networks 11 Policies to distribute the Beacon Interval (I) 1. Equidistribution: All Superframe orders are set to the same value 2. Fixed Priorization of the superframe order of the coordinator: Superframe order of the coordinator is set to twice the value of the rest
Analysis of the scalability of hierarchical IEEE /Zigbee networks 12 Policies to distribute the Beacon Interval (II) 3. Topology based distribution: The order is particularized for each router depending on the number of the leaf nodes Proposal of an iterative algorithm: l i be the number of leaf nodes ‘depending’ of the i -th coordinator (or supported traffic) The SO of the coordinator with the highest l j is increased in one unit If the BI is not exceeded by the sum of the SDs, the increase of SO is admitted & l j is divided by two The process is repeated while no SO can be increased without exceeding the BI
Analysis of the scalability of hierarchical IEEE /Zigbee networks 13 Simulation parameters Ad hoc simulator in C++ Packet level results formatted so they can be analyzed with Chipcon CC2420 Packet Sniffer Three different network topologies: three-layer hierarchy in which leaf nodes (those generating traffic) do not have any children. Simulations for different traffic loads Network performance evaluated by means of the throughput: ratio between the number of bytes that are successfully transmitted per leaf node and per superframe and the Beacon Interval.
Analysis of the scalability of hierarchical IEEE /Zigbee networks 14 Evaluated scenarios (II) Scenario 1: routers support different traffic Scenario 2: Coordinators supports many routers Scenario 3: the coordinator support the same traffic than router
Analysis of the scalability of hierarchical IEEE /Zigbee networks 15 Results (I) Policy for determining the Superframe Orders Same SO for all nodes Prioritization of Zigbee coordinator (ZC) Topology Based Distribution SO=6SO 1 =6, SO 2,3 =3SO 1 =7, SO 2,3 =6 ρThroughputρ ρ 55%4210 bps56%810 bps56%6478 bps 47%3562 bps45%648 bps46%5371 bps 40%3077 bps34%486 bps41%4720 bps Policy for determining the Superframe Orders Same SO for all nodes Prioritization of Zigbee coordinator (ZC) Topology Based Distribution SO=5SO 1 =6, SO 2,3,4,5 =3SO 1 =7, SO 2,3,4,5 =5 ρThroughputρ ρ 56%814 bps56%810 bps56%3255 bps 45%651 bps45%648 bps45%2604 bps 34%488 bps34%486 bps39%2278 bps Scenario 1 Scenario 2
Analysis of the scalability of hierarchical IEEE /Zigbee networks 16 Results (II) Scenario 3 Policy for determining the Superframe Orders Same SO for all nodes Prioritization of Zigbee coordinator (ZC) Topology Based Distribution SO=7SO 1 =6, SO 2 =3SO 1 =7, SO 2 =7 ρThroughputρ ρ 39%2604 bps 45%163 bps* 39%2604 bps 36%2115 bps36%2115 bps 34%1953 bps34%1953 bps Results of the topologies in which the Zigbee coordinator concentrates the traffic (e.g.: the scenario 2) evidence that resources cannot be equally distributed among the clusters. Scenario 3; limit case in which a router has to transport the same traffic of the Zigbee Coordinator. SO order of both clusters must be equal
Analysis of the scalability of hierarchical IEEE /Zigbee networks 17 Conclusions & Future Work Problem of configuring SD is a key aspect for hierarchical /Zigbee cluster-trees Even in small networks with less than twenty nodes a proper design of the duration of the superframes is crucial to achieve a reasonable network performance. An iterative strategy to design the SD of the nodes of a Zigbee network has been proposed. SD is defined as a function of the topology (traffic) Simple policies to distribute the beacon interval without taking into account the topology and traffic condition in the PAN leads to an inefficient network design Future work should investigate the adaptation of this type of algorithms to more complex situations: node mobility, not all the routers interfere, etc.