Analysis of IEEE in a Power Line System

Transcription

1 Analysis of IEEE in a Power Line System TORSTEN LANGGUTH, MARKUS ZELLER, HELMUT STECKENBILLER, RUDI KNORR Fraunhofer Institute for Communication Systems Hansastraße 32, München, Germany Abstract: - Power line Communication (PLC) is expected to be a major growth factor in outdoor and inhouse networks in the near future. Due to the different channel conditions, existing wireline MAC protocols cannot be used for power line communication. In this paper we analyze the design issues for a Power line MAC and present a performance evaluation of the wireless MAC protocol IEEE Key-Words: - Power line, Access control, MAC 1 Introduction A wide range of media and transmission technologies are used in home and access networks. Power line communication (PLC) is an alternative technique which is expected to be an important growth factor in outdoor and inhouse networks in the next years. PLC systems will be used in environments where cabling for wired communication systems does not exist, is very expensive, or as a result of competitive market situations. A lot of research has been done in the field of power line communication and especially at the physical layer during the last years. Despite from these developments there is also a need for medium access control (MAC) which meets the requirements of power line systems. This paper focuses on the evaluation of key features of power line MAC protocols. In section 2 we show the issues in designing MAC protocols for PLC. Section 3 presents an introduction to the used protocol. In section 4 we describe the simulation environment and results and section 5 concludes the major points. 2 Media access design MAC protocols cannot be considered independently from the requirements of the underlying physical medium. Due to its heterogeneity the power line channel has some differences to common wired channels. The distance between stations is not limited. Furthermore the PLC channel is a shared media. The power line channel is characterized by highly varying, heavy noise which is influenced by a lot of different noise sources with different characteristics [5]. In addition the power line channel is characterized by multi-path signal propagation [6]. Because of interference from other electrical devices the power line medium has a very high bit error rate. The high frequency signal power is limited by strong radiation. These effects are variable by time and location [7]. Furthermore the power line medium is not safe from interception. PLC signals will be transmitted to other houses in the neighborhood (outdoor system) or other apartments (inhouse systems). Due to cross talk effects and the broadcast structure of power lines, there is no privacy on the PLC system. To prevent from eavesdropping, there are authentication and encryption/decryption algorithms required. Furthermore the MAC protocol should support easy interworking with existing LAN protocols. To allow interoperability there is a need for standardization of protocols in the MAC layer. 2.1 Scenarios The common scenarios in PLC are outdoor and inhouse networks. The services, which have to be supported in both network types are internet access, phone, advanced energy services and home networking services. IP based Backbone Figure 1: Outdoor-System transformer station low voltage range 230V/400V The outdoor scenario represents the last mile from the backbone networks to the end user. The scenario is a star topology between the transformer station and the

2 households [Figure 1]. On the cables from the transformer station are from some tens up to few hundred households connected. All households on a cable share the capacity and have to use the network in a fair and controlled way. Estimations expect a maximum amount of up to 30 subscribers per transformer station [8]. The outdoor scenario is an infrastructure network where the access could be distributed or centralized. In the distributed scheme every station accesses the channel on its own decision. In the case of centralization the transformer station act as a prioritized station which controls the access of the households. The inhouse scenario is an ad-hoc network where normally no prioritized station exists [Figure 2]. In that case either a distributed access scheme can be used or a station gets control over the other stations by manual configuration. Due to the heterogeneity and the missing administration the inhouse systems must be extremely robust and easy to configure. stations varies with the location and scenario and leads to the hidden station scenario [Figure 3]. Let us assume the example from Figure 3, where station 1 and station 3 transmit to station 2. The distance between station 1 and station 3 is longer then the transmission range of both stations. They can not hear each other and are hidden. Station 2 can hear all the stations. Therefore the communication between station 1 and station 2 is perturbed in station 2 if station 3 transmits too. Including the fact that station 1 and station 3 are hidden, they can not co-ordinate their initiation of calls and the communication can be interrupted. Hidden stations exist both in outdoor and inhouse networks. STA 1 range of Station 1 STA 2 Figure 3: Hidden station scenario range of Station 3 STA 3 Intranet Printer Laptop with PCMCIA- Cart 230 V Modem Access control Figure 2: Inhouse-System PC with PLC-Modem Phone Modem PC with PLC- Cart Workstation Crosstalk effects between both scenarios could only be avoided by sophisticated filtering. The installation effort for the filters is too high for real implementation. Therefore both network types must coexist. Due to the interaction between outdoor and inhouse networks both scenarios should know the behavior of the other network type. 2.2 Hidden station problem Because of the heterogeneity of power lines the distance between stations is not bounded. Additionally the attenuation depends on the type and structure of cabling. Therefore the transmission distance of the 3 Access Scheme In the previous section we introduced the features of power line channels. Now we want to show, how these features can be utilized to design a MAC protocol. Media access schemes control the access to shared communication networks among spatial distributed stations. Aims of MAC protocols are high throughput and network utilization, support of different traffic classes and quality of service, fair capacity sharing, simplicity, robustness, and ease of implementation [4]. To achieve a good performance in both scenarios we use a distributed and random access scheme, where the stations compete for the access. This scheme is flexible, suits to internet burst traffic, and has a good utilization characteristic, if the access is coordinated. Both outdoor and inhouse networks interact due to crosstalk. To avoid performance reduction both scenarios should work together. For this reason we use a similar protocol in both scenarios. From the analysis of the channel characteristics can be seen that there are a lot of similarities between power line and wireless channels. That s why we suggest the use of methods, which have been successfully deployed in wireless MAC protocols and analyzed several protocols. Starting from the assumption of IP based packet transport we put our focus on the usability of the wireless LAN standard IEEE [1]. The standard is able to fulfil the following requirements: fair media sharing, advanced error recovery, Quality of Service

3 support, security support, encryption/decryption, service in ad hoc and infrastructure networks. The standard defines physical layer for OFDM and CDMA [2][3]. These are some of the requirements for power line systems. Therefore we want to introduce now the features of the IEEE The fundamental access method of the IEEE MAC is a distributed coordination function (DCF), known as carrier sense multiple access with collision avoidance (CSMA/CA). An aim of the DCF is the reduction of the collision probability. The transmission is subdivided in cycles of variable length. Start and end of a cycle are defined by a specific idle duration (DIFS) of the channel [Figure 4]. A station, which wants to transmit senses the medium to determine if another station is transmitting. If it is determined that the medium is idle, the station starts transmitting immediately. A station that senses the medium busy waits until the ongoing transmission is finished [1]. During a access cycle all stations which want to start transmitting have to wait until the end of the cycle. After the cycle all these stations want to send there data. Therefore the collision probability reaches it highest value at the end of the latest transmission cycle. To avoid the collision at the start of a new cycle the stations choose a random backoff. Then the stations decrement the backoff in every slot. If no other station started to transmit before the backoff is decremented to zero, a station starts its own transmission. Collisions can only occur if two stations selected the same backoff value [Figure 4]. These method reduces the collision probability where it is mostly needed. If another station selected a lower backoff value and started transmitting earlier the station freezes its backoff value and waits for the start of the next transmission cycle. In the next transmission cycle it uses the remaining backoff value from the previous competition. The backoff algorithm reduces the probability of simultaneous transmission of several stations at the beginning of a new transmission cycle. busy Medium DIFS PIFS Defer Access Select Slot and Decrement Backoff as long as medium is idle Transmission Cycle RTS CTS Data ACK Winner of Backoff transmits Data DIFS idle Medium Figure 4: DCF access cycle The DCF uses positive acknowledgements. A station generally responds to a data frame with an acknowledgement. The lack of reception of an expected ACK indicates to the source station that an error has occurred. To avoid the hidden station problem the DCF uses a RTS/CTS mechanism. Transmitter and Receiver exchange short control messages (request-to-send RTS and clear-to-send CTS) before the transmission of data [Figure 5]. This control messages include the network allocation vector, which the duration of the transmission estimates. All other stations stand back from transmission during that time. This mechanism is used to guarantee undisturbed transmission in case of hidden terminal scenarios. The service offered by the DCF is a contention service. This service does neither guarantee any bounded access delay nor bandwidth availability. src dest others DIFS RTS Data CTS ACK duration from RTS duration from CTS Figure 5: Network allocation vector new cycle To offer a service which supports bounded access delay the IEEE MAC may also incorporate an optional access method called point coordinated function (PCF), which is only usable on infrastructure networks. This access method uses a point coordinator (PC) to determine which station currently has the right to transmit. The operation is essentially that of polling with the PC operating as the polling master [1]. In point coordinated mode both services (DCF and PCF) are available. During the contention free period the PC polls all stations contained in its polling list and allows them to transmit time constrained traffic. During the contention period all stations are competing for the medium access by using the DCF. Beacon PIFS D1+ poll Contention Free Repitition Interval Contention Free Period Contention Period U1+ ack D2+poll + ack U1+ ack D3+poll + ack Figure 6: PCF access cycle PIFS D4+ poll U4+ ack Contention Window Backoff- Window Slottime CF- End send by Point Coordinator send by polled station 3.1 Changes of the standard In our investigations we found some points which should be changed in a MAC protocol for power lines. On one side these include features which are not necessary in a PLC system (power saving etc..). On the other hand these features are used to improve the performance of the MAC protocol. As a 1 st step we evaluated the frame format of

4 and found that some fields are not used in the PLC system. These include the fields for the distribution system address and the power management field. Furthermore we limit the maximum transmit unit to 1500 Byte. That value coexists very well with Ethernet and PPP, which is used in most access systems for authentication and accounting. As a side effect this leads to shorter access cycles and decreases the probability of PCF start variations. Additionally the backoff window should be adapted too. The window size depends on the number of active stations. If only few stations are used the backoff window could be limited to twice the number of active stations. As a last point we examined the inter frame spaces (IFS). Proper chosen IFS values have a strong influence on the performance of the MAC protocol. These values are derived from the processing time and the RxTx turnaround delay and depend on the hardware. Therefore we do not propose any new values but we note that they should be adapted on the hardware in a real implementation. 4 Simulation 4.1 Goals and Environment With respect to the above discussed issues we found several key points in evaluation the performance of the protocol. We evaluated the general performance of the protocol with respect to the offered load, the number of active stations and the bit error rate (BER). Additionally we put our focus on the capability to support quality of service. We simulated a broadcast LAN with a variable number of active stations. In DCF mode all stations have the same functionality and priority. In PCF mode a network master controls the media access. Our PCF scheduler uses a round-robin polling list to determine the order and bandwidth of the individual connections. A slave can have multiple entries in the polling list to increase the link bandwidth. The further task of the master is to check the correct reception of the packets and to initiate ARQ retransmissions in case of defects. The available bandwidth of the physical layer is 1 Mb/s. Bit errors are not expected to be equal or poisson distributed but rather have bursty character. This influences the packet error rate even if there is no error correction scheme implemented [11]. Our bit error generation model allows the adjustment of the total bit error rate and the burstiness. The distribution of the bit errors can be viewed as a combined distribution consisting of a poisson distribution which describes the occurrence of individual bursts and a second poisson distribution which describes the number of errors inside a burst. The combination of both distributions is called Neyman-Contagious Distribution of type A []. The access control models contain the relevant parts of the standard including the Quality of service functionality. Parts of the standards which are not necessary in power line channels like the power saving functionality are out of the scope of the paper. In the simulation model we have integrated two different traffic sources. The 1 st source generates burst data traffic whose distribution is taken from a trace file that contains packet size and interarrival time of internet traffic and has been recorded over 24 hours at Bellcore Morristown Research and Engineering facility [9]. However the inter arrival times of packets are exponentially distributed in order to be able to generate variable load conditions. The 2 nd source simulates a deterministic voice data source with variable bit rate and packet size. This source is used in all PCF aware stations. In our simulation we have chosen a data rate of 64kb/s. We simulated LANs with 2, 4, 8 and 32 active stations. 4.2 Simulation Results In our 1 st simulation we evaluated the performance of CSMA/CA-functionality including the RTS/CTS mechanism and the backoff scheme. Figure 7 shows the throughput of the mechanism with respect to the bit error rate. We simulated bit error rates from -7 to -3. In the range of -7 to -6 almost no data are lost. For this reason the throughput is close to the maximum reachable value in both cases. The throughput is decreasing dramatically if the error rates reach values lower than -4. The minimum overhead is 17.4%. Under light load conditions the throughput can reach values which are close to the offered load Figure 7: Throughput of DCF 8 stations mode Figure 8 presents the results of throughput simulations with both DCF and PCF. The PCF is used to transmit constant bitrate voice streams of 64kbps data rate. That means we have a constant PCF load of 256 kbps. If the BER is low the throughput is in all simulated cases

5 equal to the offered load. This hold also when the background load exceeds the capacity. In that case only the throughput of the DCF traffic is decreased. Obviously the throughput of the PCF is independent of the DCF. In the case of higher bit error rate the throughput is also constant but due to the losses lower then the offered load. This shows the stability and robustness of the scheme. The decrease of throughput is lower then in the DCF. This is caused by the lower packet error probability because of shorter packets in the PCF. On account of the extended access method the minimum overhead is increased Figure 8:Throughput of 4 stations PCF and 4 stations DCF mode Figure 9 illustrates the delay of the DCF. The delay increases with the number of stations. If more stations are competing for access every station has to wait longer. Therefore the mean delay increases. The simulations showed that in that scenario all stations have similar delay. The gradient of rise of delay increases in the range of saturation. If the load increases the data are queued longer before transmission. That is significant in the range of system saturation. In that case no more data can be transmitted and the new data must be queued longer Figure 9: Delay per station DCF mode with 2, 4, 8 or 32 active stations The flattening of the graphs under high load comes from the limited queue length in the simulations. Because of the automatic repeat of lost frames (ARQ) the delay increases if the bit error rate increases. All lost frames must be send at least twice. This leads to longer queuing delay in the sending stations. Especially on higher load conditions the delay does not meet real time requirements. Due to the independence of PCF and DCF the delay [Figure ] of stations in PCF mode does not vary with increased background load. The delay is getting worse with increased errors. Nevertheless the latency is in a range, where real time requirements can be satisfied. The stations in DCF mode have the same performance like in DCF-only mode with reduced channel capacity. Under light load the delay can be low but under heavy load the delay increases rapidly and is bounded by the queue size of sender. Although the delay is low it is not fixed. Owing to the start variations of PCF mode which are caused by access cycle extension of DCF the delay has a small variation. But this variation does not lead to a violation of the real time requirements of deterministic traffic sources Figure : Delay of 4 stations PCF and 4 stations DCF mode Figure 11 illustrates the behavior of the DCF with respect to the number of active stations. Due to the fair capacity sharing there is less impact of the number of stations on the overall throughput. The maximum available overall throughput (823 kb/s) is equal for 2, 4, 8 or 32 active stations. The available bandwidth is equally shared among all active stations. The low influence of the number of stations can be justified by the backoff algorithm. In our simulations the stations choose backoff values in the range up to 127 slots. It is easy to understand that the probability of equally chosen backoff does not dramatically increase in the range from 2 to 32 active stations.

6 maximum throughput in % of load BER e-07 BER e-05 BER e-04 BER e number of active stations Figure 11: Utilization of DCF mode 5 Conclusion In this paper we presented an analysis of the wireless MAC standard IEEE in a PLC system. We did a performance analyze of the mechanism evaluating the ability to support high throughput and low delay in power line systems. Our results show, that the protocol offers a fair and stable service with good utilization even under high load and with many stations. The random access method in the DCF is well suited to burst traffic, which is often generated by the most internet applications. In the internet access scenario with many stations, bursty traffic and low delay limits the IEEE DCF is a good choice as a MAC protocol. To deal with real-time-requirements an optional coordinated access method is needed. In that case one station acts as the coordinator and polls the concerned stations. Although the PCF is not able do provide hard delay bounds, the service offers the ability to support real-time applications. [4] H. R. van As, "Media access techniques: The evolution towards terabit/s LANs and MANs, Computer Networks and ISDN Systems 26 (1994), pp [5] H. Philipps, "Modelling of Powerline Communication Channels", Proc. 3rd Int'l Symposium on Power-Line Communications and its Applications (ISPLC '99), Lancaster (UK), 1999, pp [6] M. Zimmermann, K. Dostert, "A Multi-Path Signal Propagation Model for the Power Line Channel in the High Frequency Range", Proc. 3rd Int'l Symposium on Power-Line Communications and its Applications (ISPLC '99), Lancaster (UK), 1999, pp [7] A. Voglgsang, T. Langguth, G. Körner, H. Steckenbiller, R. Knorr, "Measurement, Characterization and Simulation of Noise on Powerline Channels", Proc. 4th Int'l Symposium on Power-Line Communications and its Applications, Limerick (IR), 2000 [8] Powerline Forum, [9] H. J. Fowler and W. E. Leland, "Local Area Network Traffic Characteristics, with Implications for Broadband Network Congestion Management", IEEE JSAC, 9(7), September 1991, pp , ftp://ita.ee.lbl.gov/traces/bc-paug89.tl.z [] J. Neymann, "On a new class of contagious distributions, applicable in entomology and bacteriology", Ann. Math. Statist.,, No. 35 (1939) [11] G.G. Pullum, "Nachbildung von Burst-Fehlern bei digitaler Übertragung", in "Rauschen und Stochastik in der Nachrichtentechnik", hrsg. K.W.Cattermoele, J.J. O'Reilly, VCH Verlagsgesellschaft, 1988 Acknowledgements The authors are especially indebted to Corina Scheiter, Markus Augel, Mike Heidrich and Walter Zimmer for their contributions. References [1] Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specification, IEEE /D, Draft International Standard ISO/IEC , Jan. 14, 1999 [2] Draft Supplement to Standard: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specification: High Speed Physical Layer in the 5 GHz Band, IEEE P802.11a/D7.0, [3] Draft Supplement to Standard: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specification: High Speed Physical Layer extension in the 2.4 GHz Band, IEEE P802.11b/D7.0,