EDIC RESEARCH PROPOSAL 1 MAC Layer Performance Analysis of the IEEE 1901 Standard for Power Line Communications Christina Vlachou I&C, EPFL Abstract—Power line communications are one of the fastest out special wiring needed for the installation of the physical growing technologies in home networking. The IEEE 1901 network. The power line channel varies widely with frequency standard specifies the physical and MAC layers of high data and time. Connected appliances have a complex impedance rate power line communications. We study analytical models of the MAC layer of the IEEE 802.11 standard which uses a similar, causing reflections of the transmitted signal and, as a result, but much simpler CSMA/CA mechanism. The analysis is under multipath propagation. Switching the electrical appliances saturated assumptions. The analytical models usually assume changes the characteristic impedance of the transmission line independence between the stations (decoupling assumption) [8]. causing channel characteristics to change over large time- We discuss whether the decoupling assumption holds in the scales. The physical layer of the 1901 standard provides stationary regime of a system [3]. We also study a discrete time Markov chain model of the IEEE 1901 standard [7]. It is yet an modulation techniques with robust forward error correction open question whether the decoupling assumption holds in the which offer the possibility of high speed communications. stationary regime of the 1901 standard. Both wireless and power line transceivers cannot sense the As future research directions, we propose mean field methods medium while transmitting. Hence, both MAC layer protocols to analyze the performance of the IEEE 1901 standard. We also use Carrier Sense Multiple Access with Collision Avoidance discuss future research plans in home networking technologies. (CSMA/CA) mechanisms. The MAC layer of the 1901 stan- Index Terms—HomePlug, Medium Access Control (MAC), dard uses both Time Division Multiple Access (TDMA) and CSMA/CA, decoupling assumption, mean field analysis CSMA/CA methods. The network is synchronized with the AC line cycle by a central coordinator that transmits beacons I. INTRODUCTION in order to synchronize the two different access methods. The IEEE 1901 standard specifies the physical and MAC The central coordinator is the first device activated, but can layers of high data rate power line communications. The change depending on the physical channel conditions and on advantage of power line communications over wireless tech- the availability of other devices. We are interested in studying nologies is that they provide wider communication range with- the CSMA/CA method which is similar to the CSMA/CA method of the 802.11 standard. Proposal submitted to committee: July 5th, 2012; Candidacy In this proposal we study analytical models of the two exam date: July 12th, 2012; Candidacy exam committee: Jean- CSMA/CA methods. These CSMA/CA methods have some Yves Le Boudec, Patrick Thiran, Christina Fragouli. differences which are explained in the following paragraphs. This research plan has been approved: We study analytical models of the 802.11 standard in order to model similarly the CSMA/CA method of the 1901 standard. In order to resolve contention, the 802.11 MAC layer uses Date: ———————————— a time-slotted random backoff procedure [1] with a backoff counter and a contention window (CW ). At each transmission attempt the station senses the medium. If the medium is sensed Doctoral candidate: ———————————— idle, the station selects a backoff counter uniformly at random (name and signature) over the range [0;CWmin 1]; where CWmin denotes the minimum contention window− (equal to 32 for the 802.11b standard). The station decrements its backoff counter by 1 at Thesis director: ———————————— each time slot if the medium is sensed idle. If the medium is (name and signature) sensed busy, the station freezes its backoff procedure. When the medium becomes idle again, the station unfreezes its Thesis co-director: ———————————— backoff counter and starts decrementing it again. (if applicable) (name and signature) If the backoff counter is 0, the station transmits a frame. If the transmission is successful, the same procedure is followed for the next packet. If the transmission fails, the station Doct. prog. director:———————————— chooses a new backoff counter uniformly at random over (R. Urbanke) (signature) the range [0; 2CWmin 1] and retransmits when the backoff counter is 0. The contention− window is doubled whenever a EDIC-ru/05.05.2009 transmission attempt fails and does not go above a maximum EDIC RESEARCH PROPOSAL 2 PRS0 PRS1 Priority: CA0 and CA1 CA2 and CA3 CA3 3 3 BP C CW DC CW DC CA2 3 0 8 0 8 0 CA1 3 1 16 1 16 1 CA0 2 32 3 16 3 3 64 15 32 15 TABLE I: Busy tones transmitted during the two priority ≥ resolution slots (PRS0, PRS1) in the IEEE 1901 Standard. TABLE II: IEEE 1901 standard values for CW and DC. Different priority levels have different parameters. value CWmax (equal to 1024 for the 802.11b standard). The CSMA/CA random backoff procedure of the IEEE 1901 standard is more complex than the procedure described due to collisions. above [2]. First, there are four priority levels CA0, CA1, CA2, • Homogeneous network: All stations have the same mini- CA3 (priority from the lowest to the highest) and only the mum and maximum contention windows. stations with the highest priority contend for the medium in Each model may use further specific assumptions that we the contention period. The highest priority is determined by mention in the following sections. In Section II we present two priority resolution slots PRS0, PRS1. Stations transmit a different analytical models for the backoff procedure of the busy tone in these slots according to their priority as shown IEEE 802.11 standard. Section III presents a discrete time in Table I. If the stations sense a busy tone in one of these Markov chain model of the IEEE 1901 standard [7]. Finally, slots, they differ from transmission. concluding remarks and our research proposal are given in Second, the backoff procedure uses three counters: the Section IV. backoff procedure counter (BP C), the backoff counter and the deferral counter (DC). BP C is set to 0 at the first transmission attempt. The deferral counter and contention window values II. MODELING SATURATED 802.11 NETWORKS are set according to BP C as shown in Table II. After choosing these values, BP C is increased by 1. The backoff counter is A. Bianchi’s model chosen uniformly at random over the range [0;CWmin 1] and is decremented by 1 at every idle time slot, as in− the The backoff procedure of the IEEE 802.11 is modeled with 802.11 standard. If the backoff counter reaches 0, the station a discrete time Markov chain in [5]. In the 802.11 standard the i transmits. After a successful transmission BP C is set to 0 and contention window CWi at backoff stage i is 2 CWmin. The the same procedure is followed for the next packet. After a Bianchi model consists of the variables (s(t); b(t)), with s(t) transmission failure, the contention window is doubled (up to representing the backoff stage and b(t) denoting the backoff a maximum value as shown in Table II) and a new backoff counter of the station at time slot t N. The backoff counter 2 counter is chosen. The deferral counter is updated according is chosen uniformly at random over the range [0;CWi 1], − to BP C and then BP C is increased by 1. as in the real protocol. The specific assumptions of this model If a station senses the medium busy, it freezes its backoff are the following: counter. The difference with the 802.11 standard is that after • For each station the collision probability p conditioned sensing the medium idle again, the station may resume its on the event that the station transmits is time-invariant backoff counter depending on the deferral counter value. More and is independent of the backoff stage. precisely, if the deferral counter is 0, the station selects a • The backoff processes of the stations are independent. new contention window (CW ) and a new deferral counter according to the backoff procedure counter (Table II). Then After processing the equations of the stationary distribution a new backoff counter is chosen uniformly at random over of the irreducible Markov chain, the probability of transmis- the range [0;CW 1] and BP C is increased by 1. If the sion τ is given by − deferral counter is not 0, the station decrements the deferral 2(1 2p) τ(p) = ; (1) and backoff counters by 1. The station keeps decrementing − m (1 2p)(CWmin + 1) + pCWmin(1 (2p) ) the backoff counter by 1 during each idle time slot, as in the − − CW 802.11 case. where min is the minimum contention window of the m In this proposal we study analytical models of the station and is the backoff stage such that the contention win- CW = 2mCW CSMA/CA mechanisms described above. The general assump- dow is max min. After reaching backoff stage m tions that hold for all the models are the following: , the station changes backoff stage only after a successful transmission (the retry limit K is infinite). The conditional col- • Perfect sensing: There are N stations in a single con- lision probability and the transmission probability are related tention domain. by • Saturated conditions: All stations always have packets to p = 1 (1 τ(p))N−1; (2) transmit. − − • Perfect channel: There is no packet loss or error due to and τ, p can be computed from the fixed point equation (2), the physical layer. Thus, transmission failures are only which has a unique solution [5].