Medium Access Control Protocols of the PRMA Type in non-geostationary Satellites

Transcription

1 Medium Access Control Protocols of the PRMA Type in non-geostationary Satellites Giovanni Giambene Dipartimento di Ingegneria dell Informazione Università degli Studi di Siena Via Roma, Siena, Italy Abstract The challenge of future mobile multimedia networks is to provide worldwide tetherless communication services. Low Earth Orbit-Mobile Satellite Systems (LEO-MSSs) will play a significant role by filling the coverage gaps of future generation terrestrial cellular networks. This lecture presents research results on demand-assignment Medium Access Control (MAC) schemes able to share efficiently LEO satellite resources among users and to support isochronous traffics and the ubiquitous access to the Internet. 1 Introduction Future generation mobile communication systems will achieve a global coverage by integrating a terrestrial cellular component and a satellite one [1],[2]. The satellite system will play a complementary role with respect to its terrestrial counterpart; typical operational environments for satellite systems are regions where the provision of the terrestrial coverage is either technically or economically unfeasible. The role of mobile satellite systems is: (i) to allow the global roaming of users; (ii) to provide Quality of Service (QoS) levels comparable with those of terrestrial systems; (iii) to permit the rapid deployment of mobile services in underdeveloped regions. The satellite component of future mobile communication systems will be based (partly or totally) on non-geostationary constellations. In particular, this study focuses on Low Earth Orbit Mobile Satellite Systems (LEO-MSSs), since they are close to the earth and allow the use of low-power lightweight mobile terminals [3]. In what follows, an earth-fixed cell system [4] will be assumed where antenna beams are steered so as to point towards a given cell on the earth during the satellite visibility time. Satellites are power and bandwidth limited. Therefore, it is essential that satellite resources are efficiently utilized. Hence, suitable Medium Access Control (MAC) protocols must be identified for the management of resources in a cell. MAC schemes contain a set of rules according to which the power-bandwidth resource is assigned to the different communications. In particular, real-time traffic (i.e., isochronous voice traffic) and Available Bit Rate (ABR) data traffic are considered. Several types of Medium Access Control (MAC) schemes have been proposed for satellite systems [5], but the identification of an efficient MAC protocol, able to guarantee suitable QoS for different traffics, is still an open research issue. A MAC protocol taxonomy can be envisaged as described below. 1. Fixed access protocols that grant permission to transmit only to one terminal at once, avoiding collisions of messages on the shared medium. Access rights are statically defined for the terminals. 1

2 2. Contention-based protocols that may give transmission rights to several terminals at the same time. These policies may cause two or more terminals to transmit simultaneously and their messages to collide on the shared medium. This class encompasses pure Aloha, Slotted-Aloha and Reservation-Aloha [6]. 3. Demand-assignment protocols that grant the access to the network on the basis of requests made by the terminals. The reason for the presence of many different MAC protocols is that they are suitable for some applications, but often do not meet the requirements for other applications. For instance, fixed access schemes are not efficient with bursty traffics, because they can not adapt to varying traffic conditions. This lecture presents research results concerning new demandassignment MAC schemes that are evolutions of the classical Packet Reservation Multiple Access (PRMA) protocol. These novel protocols are able to integrate the management of isochronous and data bursty traffics and, therefore, can be useful in the Satellite- Asynchronous Transfer Mode (S-ATM) scenario. 2 The classical PRMA protocol in LEO-MSSs The PRMA protocol was originally proposed for terrestrial microcellular systems [6]: it is based on Time Division Multiple Access (TDMA) and combines random access with slot reservation. The efficiency of PRMA relies on managing voice sources with Speech Activity Detection (SAD): only during a talkspurt, a voice source has reserved one slot per frame to transmit its packets. A feedback channel broadcast by the cell controller informs the terminals about the state of each slot (i.e., idle or reserved) in a frame. As soon as a new talkspurt is revealed, the terminal tries to transmit a packet in the first idle slot (contending state), according to a permission probability scheme [6]. When the transmission attempt of a terminal is successful on a slot, the terminal obtains the reservation of this slot. The main limiting factor for the use of PRMA in LEO-MSSs is the high Round Trip propagation Delay (RTD) value that prevents the mobile terminals on the earth to know immediately the outcome of their transmission attempts. In LEO-MSSs, RTD values vary from 5 ms to 30 ms, depending on the satellite constellation altitude and the minimum elevation angle from mobile terminals to the satellite. The satellite recognizes the request made by a terminal by decoding the header of the received packet on an unreserved slot. For a conservative study, RTD is assumed always equal to its maximum value, RTD max, for a given LEO-MSS. For the correct protocol behavior, T f must be greater than or equal to RTD max + ε, where ε is the packet header transmission time. Consequently, when a terminal attempts to transmit on a given slot, it receives the outcome of its attempt before the beginning of the same slot in the next frame. Of course, the selection of the T f value must also account for both the requirements on the voice end-to-end delay, the voice codec and the packetization process. It is considered here T f RTD max (ε is negligible). If more terminals attempt to send their packets on the same slot, there is a collision (unless capture phenomena occur [6]): these terminals know that they must reschedule new transmissions only after RTD. This delay is particularly significant for the real-time voice service. In order to relax these problems, the following Section presents several solutions that are also suitable for other scenarios where RTD is greater than the packet transmission time (e.g., high altitude aeronautical platforms recently proposed to provide high bit-rate transmissions in heavy traffic urban areas [7]). 3 Novel schemes based on PRMA This Section surveys three novel MAC schemes that derive from the modification of PRMA in order to make it more suitable for the LEO satellite scenario. Let us consider M v Voice Terminals (VTs) and M w Data Terminals (DTs) per carrier per cell. 2

3 3.1 PRMA with hindering states In the classical PRMA scheme, assuming a typical LEO system where RTD max is about equal to 16 ms, a terminal could perform at most two access attempts before dropping the first packet (as shown in Section 4.1, the voice packet deadline is 32 ms). To remove this constraint, a modified PRMA protocol is presented here, where a terminal is allowed to attempt transmissions (according to the permission probability scheme) also while it is waiting for the outcome of a previous attempt. If the previous attempt has been unsuccessful, this modification permits a faster access scheme. Otherwise, these further attempts are useless and may hinder the accesses to other terminals. Accordingly, this scheme has been called PRMA with Hindering States (PRMA-HS). It has been shown in [8] that the advantages of this fast retransmission scheme overcome the problems due to useless attempts. In order to integrate VT and DT traffics, different permission probabilities values have been considered; p v and p d, respectively, where p v > p d to prioritize the voice real-time traffic. 3.2 Modified PRMA scheme In the Modified PRMA (MPRMA) protocol, a given field in the packet header is devoted to notify the satellite if a transmission request (i.e., the first packet of a terminal that must acquire transmission rights) comes from a DT or from a VT; accordingly, different algorithms are used. In particular, the management of VTs is as in the classical PRMA scheme. Whereas, DT requests are served as described below. If a message is generated by a DT when its buffer is idle, its first packet (= request packet) is transmitted on an available slot, according to a permission probability, p d. Collisions may occur with the attempts of other terminals. If this first packet does not experience a collision, its header contains a DT transmission request that is stored into a buffer on the satellite in order to form a queue of DTs that need to transmit. The controller on board of the satellite manages a queue of uplink transmission requests for each MPRMA carrier of a cell and decides allocations of slots (not used by VTs) to DTs [9]. In order to guarantee a certain number of idle slots for new accesses of DTs and VTs, the controller assigns an idle slot to a given DT (with its request at the head of the satellite queue) in the next frame, according to an access probability p a. Let us consider a DT that is allowed to transmit the last packet of its presently served message and that has in its buffer other messages arrived in the meanwhile. A suitable flag is set in the header of this last packet so that the DT requests the transmission of another message to the satellite (piggybacked request). Both the random access packet and the packet used for the piggybacked request use a suitable field in the header to notify the message length to the satellite. 3.3 DRAMA protocol In the Dynamic Resource Assignment Multiple Access (DRAMA) scheme, the frame contains first access slots and, then, information slots. Access slots are minislotted: a terminal sends an access burst on a minislot to request transmission resources. The number of minislots per slot is about equal to the number of packet headers that can be transmitted in a slot time. The number of minisloted slots per frame is always greater than or equal to one. This value is dynamically updated trying to have a total number of minislots equal to the number of terminals that have set-up a connection (satellite-atm) and that are not transmitting. This permits to maximize the throughput of successful requests per minislot. Colliding terminals will reattempt in the next contention phase. A feedback channel informs the mobile terminals at the beginning of each frame about the number of minislotted slots for the next access phase. With the DRAMA protocol the satellite becomes an ATM scheduler: slot allocations are dynamically updated; the feedback channel is used to send assignment commands for each 3

4 slot. Slots are first assigned to fulfill the VT requests. Remaining slots are used to serve DT requests. On the basis of currently active transmission requests, the satellite knows how many slots will be destined to the VTs in the next frame and decides slot assignments to data traffics by using a cyclic policy among all DTs. Each data traffic transmission request conveys the number of packet of the related message. Hence, the satellite may decide also multiple slot allocations to a given DT in a frame (if room). The piggybacking scheme is used by data sources to communicate new access requests when they are already transmitting to the satellite. 4 Traffic source models The traffic models for generating isochronous and data bursty traffics are detailed below. 4.1 Voice sources Each VT uses SAD to distinguish between talkspurts and silent pauses in a conversation. Talking and silent phases lengths are exponentially distributed with mean values t 1 = 1 s and t 2 = 1.35 s, respectively. The voice activity factor is ψ v = t 1 / (t 1 + t 2 ) An active VT generates one packet per frame (L pkt = R s T f information bits and H p header bits). The number of slots per frame and the slot duration are: N = R c T f / (L pkt + H p ) [slots/frame] and T s = T f /N where. is the floor function, R c is the channel bit-rate and R s is the voice source bit-rate. A VT discards the first packet from its buffer, when it experiences a transmission delay greater than a maximum value, D max ; typically, D max = 32 ms [6]. When a packet is discarded, the VT tries to obtain a reservation with the next packet. Let P drop denote the packet dropping probability for a VT (P drop 1% for an acceptable speech quality). The voice throughput is η v = M v ψ v (1 - P drop )/N [packets/slot]. 4.2 Internet-like traffic sources A simplified WWW traffic model is used [10], where a DT produces packet calls separated by a reading time (traffic at session level). The number of datagrams generated per packet call is geometrically distributed with expected value N p. During a packet call, the datagram interarrival time is exponentially distributed with mean rate µ pkt. The reading time is exponentially distributed with mean rate µ rd. For numerical evaluations it is considered: N p = 25 datagrams/packet call, µ rd = 0.25 s -1 and µ pkt = 2q datagrams/s, where q {1, 2, 3, 4, 5, 6, 7}. The datagram length in bytes has the same truncated and discretized Pareto distribution shown in [10]. The activity factor for this type of DT is given by ψ w = (N pc /µ pkt )/(N pc /µ pkt +1/µ rd ). The mean message arrival rate is λ = µ pkt ψ w datagrams/s. The total traffic load produced by all DTs is: t d = λt s L w M w [packets/slot] where L w denotes the mean length of a datagram in packets. The burstiness degree of a traffic source is defined as the peak-to-mean traffic ratio. The variable value for µ pkt (depending on q) allows modulating the DT burstiness degree that is equal to 1/ψ w. The service for DTs is characterized by the average datagram delay, T dtg, i.e., the time from the datagram arrival to the DT buffer until the complete datagram transmission. 4

5 4.3 Stability considerations Under the assumption of stability, the total data throughput must be equal to the total input data traffic, i.e., t d, and the probability that a slot is used to send a packet successfully must be lower than 1 (the maximum possible utilization): t d + η v < 1 [packets/slot] λt s L w M w + M v ψ v (1 - P drop )/N < 1 [packets/slot] This condition may be used for a proper sizing of the connection admission control protocol. 5 Results and comparisons Extensive simulation runs have permitted to compare the proposed MAC schemes in a typical LEO satellite constellation at 780 km altitude with RTD max = 15 ms. Simulation results have been obtained assuming: a channel bit-rate R c = 765 kbit/s, a source bit-rate R s = 32 kbit/s, a packet header of H p = 64 bits, a frame duration T f = RTD max = 15 ms. Moreover, for PRMA and PRMA-HS p v = 0.6 and p d = 0.2 [8]; for MPRMA p v = 0.6, p d = 0.4 and p a = 0.8 [9]; for DRAMA, the number of minislots per slot N m = 8. Assuming a configuration with 12 DTs and 21 VTs, simulation results have been shown in Figs. 1 and 2 for P drop and T dtg, respectively. The data traffic increase in abscissa has been obtained by progressively increasing q (i.e., q = 1, 2, 3, ) in the mean datagram interarrival rate. On the basis of these results, it is evident that MPRMA outperforms PRMA-HS in terms of T dtg and maintains P drop below the maximum acceptable value. However, the MPRMA scheme does not allow a significant performance improvement. Hence, the novel DRAMA scheme has to be considered since it considerably reduces both T dtg and P drop so achieving a better utilization of satellite resources with respect to PRMA, PRMA-HS and MPRMA. 6 Conclusions This lecture has shown research results on MAC protocols for LEO-MSSs. Three different evolutions of the PRMA scheme have been compared; in particular: PRMA-HS, MPRMA and DRAMA. It has been shown that the DRAMA scheme permits to improve the management of isochronous and data bursty traffics. The study presented in this lecture provides useful considerations for the design of MAC protocols able to support isochronous traffics and the ubiquitous access to the Internet in S-ATM systems. 1.0E-01 Packet dropping probability, Pdrop 1.0E E E-04 PRMA MPRMA PRMA-HS DRAMA 1.0E Total data traffic, td Figure 1. Performance comparison among the considered schemes in terms of P drop. 5

6 Mean datagram delay, Tdtg [slots] 5.7E+03 PRMA PRMA-HS 4.9E+03 MPRMA DRAMA 4.1E E E E E E Total data traffic, t d Figure 2. Performance comparison among the considered schemes in terms of T dtg. References [1] A. Guntsch, M. Ibnkahla, G. Losquadro, M. Mazella, D. Roviras, A. Timm, EU s R&D Activities on Third-Generation Mobile Satellite Systems (S-UMTS), IEEE Comm. Mag., Vol. 36, No. 2, Feb [2] P. Taaghol, B. G. Evans, E. Buracchini, R. De Gaudenzi, G. Gallinaro, J. Ho Lee, C. Gu Kang, Satellite UMTS/IMT2000 W-CDMA Air Interfaces, IEEE Communications Magazine, Vol. 37, No. 9, pp , Sept [3] Official Web sites with addresses: [4] J. Restrepo and G. Maral, Coverage Concepts for Satellite Constellations Providing Communications Services to Fixed and Mobile Users, Space Communications, Vol. 13, No. 2, pp , [5] H. Peyravi, Medium Access Control Protocols Performance in Satellite Communications, IEEE Comm. Mag., Vol. 37, No. 3, March [6] S. Nanda, D. J. Goodman, U. Timor, Performance of PRMA: a Packet Voice Protocol for Cellular Systems, IEEE Trans. on Veh. Tech., Vol. 40, No. 3, Aug [7] G. M. Djunknic, J. Freidenfelds, Y. Okunev, Establishing Wireless Communications Services via High-Altitude Aeronautical Platforms: A Concept Whose Time Has Come?, IEEE Comm. Mag., Vol. 35, No. 9, Sept [8] E. Del Re, R. Fantacci, G. Giambene, and S. Walter, Performance Analysis of an Improved PRMA Protocol for Low Earth Orbit Mobile Satellite Systems, IEEE Trans. on Veh. Tech., Vol. 48, No. 3, May [9] R. Fantacci, G. Giambene, R. Angioloni, A Modified PRMA Protocol for Voice and Data Transmissions in Low Earth Orbit Mobile Satellite Systems, to appear on IEEE Trans. on Veh. Tech. [10] ETSI, Selection Procedures for the Choice of Radio Transmission Technologies of the UMTS (UMTS Version 3.1.0); ETSI, Sophia-Antipolis, Cedex, France, Nov