packet reservation protocol for personal mobile ...

6 downloads 0 Views 45KB Size Report
Newport, VIC 3015, Australia. Tel: +61 3 93600999, Fax: +613 93603999. Email: [email protected]. Richard Harris. School of Electrical and Computer.
PACKET RESERVATION PROTOCOL FOR PERSONAL MOBILE COMMUNICATIONS Abdullatif Glass Department of Planning and Training Technical Studies Institute P O Box 26833 Abu-Dhabi, UAE Tel: +971 2 5045937, Fax: +971 2 5854282 Email: [email protected]

Mark Amott Fujitsu Australia Pty, Ltd P O Box 268 Newport, VIC 3015, Australia Tel: +61 3 93600999, Fax: +613 93603999 Email: [email protected]

ABSTRACT A packet/cell reservation protocol is proposed for operation in a GPRS environment. The operation of the protocol exploits the advantages of contending random access protocols for reservation purpose while accommodating a hybrid form of GSM and ATM systems to enable the provision of services to mobile users. The protocol provides reservation priority for voice and interactive real-time services in a multichannel environment. The protocol also uses a voice activity detector to eliminate unused reserved time slots in any frame payload and then allocate them to other users that have a lower priority. A simulation model has been established to mimic the operation of the protocol for voice and data traffic. The results show better overall channel utilisation together with lower overall time delay for data messages. I. INTRODUCTION Recently, a considerable amount of research has been carried out in mobile communication systems to develop efficient protocols that are able to achieve better channel utilisation, reducing the end-to-end message delivery delay and generally to provide better quality of service (QoS). These developments and modifications have been motivated by the increasing importance and production of portable computing and personal telecommunications applications in both business and consumer markets. The deployment of these protocols is mainly centred around the use of random multiple access protocols for the purpose of the integration of voice, video, data, and multimedia services. The Packet Reservation Multiple Access (PRMA) protocol was first proposed for application in single channel applications [15]. In PRMA, distributed mobile users transmit to a controlling base station using a form of slotted ALOHA protocol with and without capture. Voice messages are given priority during the talk burst by reserving a particular time slot within a frame [1,2]. The capacity of such a protocol has been further investigated in cellular environment using self-adaptive channel allocation and Resource Auction Multiple Access (RAMA) for voice and data integration [6,7] and in the GSM system context [8]. A Centralised (C-PRMA) protocol scheme with the capability to integrate different classes of service in cellular and local wireless communication networks was then developed [9]. New variant modes of protocol operation for General Packet Radio Services (GPRS) services that allow users to compete in order to reserve a time slot in a GSM environment have also been investigated in [10,11]. In these protocols, accessing the radio channel is based on a modification to the well-known Carrier Sense Multiple Access with Collision Detection (CSMA/CD) protocol. Flexible Multiple Access (FMAC) and Inhibit Sense Multiple Access with Polling (ISMA/P) protocols were then suggested to improve deficiencies of the CSMA/CD protocol in the radio environment.

Richard Harris School of Electrical and Computer Engineering RMIT University, GPO Box 2476 Melbourne, VIC 3000, Australia Tel: +61 3 99252976, Fax: +61 3 99256010 Emails: [email protected]

In this paper, a reservation protocol is proposed which takes into consideration the advantages of its predecessor and eliminates their shortcomings. The operation of the new protocol makes use of the merits of contending protocols for reservation while implementing a hybrid form of GSM and ATM protocols for the provision of service to mobile users. While it also provides priority and conflict free access to authorised users, it uses voice activity detector to eliminates unused time slots and allocate them to lower priority users. A simplified simulation model has been built to investigate the operation of the protocol in a multichannel environment. The paper is divided into five sections. The next two sections give a description of how the overall channel is allocated for multichannel operation and the procedure by which a mobile user is granted access to service using the proposed protocol. A simulation model of the proposed protocol is then given together with a discussion and conclusion in the following sections. II. PROTOCOL CHANNEL ALLOCATION The allocated spectrum is subdivided into four channels. These are the reservation (access) channel (R-Ch), the control channel (C-Ch), the upward channel (U-Ch), and the downward channel (D-Ch). The bandwidth/rate of transmission of both the R-Ch and C-Ch constitutes only about 5-10 percent of the total system transmission bandwidth. The R-Ch is used by mobile users (MUs) to transmit mini-reservation packets to the controlling base station (BS) using contention protocols. The C-Ch is used to pass control messages from the BS to MUs. The U- and DChs’ constitute 90-95 percent of the whole allocated transmission bandwidth in duplex form with equal transmission rates. These two channels are then either divided into multiple equal transmission rate subchannels or variable rate channels. In an alternative scheme, this multiple channel system can be divided into two or more priority channels with movable boundaries depending on traffic requirements. The U- and D-Chs are used for voice, data, multimedia, and interactive real-time services to and from MUs. Figure (1a) shows the system layout. The whole transmission time is divided into ι frames (cycles) of equal duration and each frame is then subdivided into k time slots (packets). The value of ι and k can be selected based on the system’s application requirement. For example, in a GSM system, a 5ms frame is divided into 8 time slots. The frame and time slot allocations are used in both U- and D-Chs. The BS uses a framing principle for resource allocation. It determines the total number of arrivals on both the R-Ch (new arrival) and U-Ch (continuity reservation arrival) that can be served within the next frame. The continuity reservation arrival on the U-Ch is in the form of a piggybacked mini-slot attached to the MU transmitted packet.

III. PROTOCOL OPERATION With reference to figure (1), the protocol operation starts when MU initiates a request to transmit. Initially, the request starts as a new arrival on the R-Ch when a MU sends a mini-packet using contention protocols such as Aloha, CSMA, CSMA/CD or CSMA/CA with all their variants (the request could also be initiated from the U-Ch or from the BS itself via a wireconnected system). The mini-packet contains the addresses of the originating and intended MUs and the type of message (voice, data, or interactive message) that wishes to transmit. The latter allows the serving BS for better resource allocation and service provision. This is achieved by providing a priority for delay sensitive traffic and also optimisation of overall channel utilisation. The BS receives incoming voice, data and interactive service calls (requests) and provides the capability of reservation for voice calls during the talk burst period on both U- and D-Chs. The number of arrivals within frame m time-period will be served during frame m+1. The BS responds on the C-Ch by sending mini-control packets that contain the address of the authorised MU to transmit/receive within the next frame on U/D-Ch. The authorisation contains the subchannel number Cj, where j=1,2…n, and the time slot Si, where i=1, 2, …, k of that subchannel (in figure 1b, k=8). The mini-packet also contains information regarding priority and rate of transmission in case of variable transmission rate subchannels, where a variable number of time slots will be assigned in that case. This control information may contain information regarding system management such as hand-off and transmission power control, cell splitting…etc. On reception of the mini-packet authorisation, the MU tunes to the allocated subchannel to send/receive its packet within the specified time slot/s on the U-/D-Ch. The controlling BS function is then to redirect the received packet according to its address. If the packet is intended for a user within the same cell, then the BS will forward the packet on the D-Ch following the previous procedure. This means the intended MU will be notified, via the C-Ch, about the packet addressed to it and therefore tunes to the specified D-Ch and time slot/s within the frame time. The latter action assumes that there is sufficient intelligence at the BS to make such a decision without referring to the main Mobile Switching Centre (MSC). This assumption is required especially when most of the users are communicating with peers in the same cell in a form of a wireless (LAN). Voice calls generate fixed rate traffic and thus require fixed transmission rate. Data traffic however, is generated in burst form, and thus can be served either on a fixed or variable transmission rate basis. In the proposed protocol, time slots on the subchannels per frame can be reserved either with a fixed or a variable number. Since voice has priority in service, the unused time slots can be allocated for data calls based on a fixed or variable transmission rate. In the fixed transmission rate, the mobile user will be assigned a fixed number of slots in each frame on a periodical basis. In the variable transmission rate however, mobile users will be assigned either a fixed number of slots on variable rate subchannels or a variable number of slots on fixed rate subchannels. III.1. Example of Operation To explain the operation of the proposed system let us consider again the example shown in figure (1b). Here, MUs send their requests on R-Ch for a transmission. Both voice and data calls on R-Ch are requesting to establish new calls. These voice calls are given priority 2 whereas data calls are without priority. Thus, v21 means a second priority call which arrived first in sequence in frame m. The arrivals on R-Ch will be served on first-comefirst-served (FCFS) basis, and those requests involved in

collision (eg. v22 and v23) should repeat their attempts when they are not receiving confirmation (i.e. service allocation) for transmission after the lapse of a predefined time period. Arrivals at the BS received on U-Chs and via the wire-connected network are however given a higher priority 1 (v1) and their arrival sequence is distinguished by letters a, b, c, … etc. Data reservations on the U-Chs are also given higher priority compared to the data requests on R-Ch but always lower than any voice calls. The BS authorisation of transmission/reception to the MUs is sent through the C-Ch using the mini-control packets. With reference to figure (1b) reservation on both R- and U-Chs, the sequence of transmission, channel and slot allocations are shown on C-Ch. va11 means a voice continuation call authorisation with high priority 1 and its sequence is a on U-Ch should be served first. The next two digits mean this call has been designated to channel 1 of the multichannel U-Chs and should transmit on slot number 1 of that subchannel in the next frame (i.e. frame m+1). All other authorisations for MUs transmission/reception in frame m+1 are shown on C-Ch and how priority is given for a subchannel and slot allocation. IV. PERFORMANCE EVALUATION Operational performance of the proposed protocol can be investigated using three alternative modelling techniques. These are analytical, simulation and prototype modelling. The latter is the most difficult to realise practically and also the most cost ineffective and therefore it is not considered. In this paper, a simulation model has been built to realise the operation of the protocol while the analytical model is still under development. The simulation model is one which abstracts the operation of the proposed protocol to accommodate a multiple class of traffic mainly voice and data. There are many assumptions that were made within the operational scenario of the model in order to make it simple in realisation. The channel (server) is assumed ideal with no loss of packets/cells due to fading or corruption by noise which usually leads to high error rates and hence packets are discarded at destination. The latter is feasible since with the use of robust error correction coding the effect of channel impairments can be minimised or relatively eliminated. The model also considers both the R-Ch and the C-Ch are inherently available with the traffic generation source and queuing system for voice and data traffic. Two Poisson-distributed sources for traffic arrival are used to generate voice and data calls with a negative exponential distribution of call duration. Calls duration is a random number of time slots, which also represents the service time of the channel. Voice calls are served directly and with reservation priority whenever resources are available otherwise calls will be dropped/blocked. Data calls are however, served with no reservation priority and when resources are not available they will be queued until served or dropped when the buffering queue is full. Figure (2) shows a simplified simulation model. The queuing system provides service to data calls in three different disciplines. These are the head of line (HOL) or first come first served, smallest call/message first (SF), and largest call first (LF). Call duration/size represents the number of slots required on the channel per frame. In the model, single and multiple slotted subchannels are used as the main resource of service. For these subchannels, four different insertion strategies for data call are used in a total of ten time slots. Single Channel Contiguous (SCC), Multiple Channel Contiguous (MCC), Single Channel Non Contiguous (SCNC), and Multiple Channel Non Contiguous (MCNC). In single channel insertion strategy, each subchannel is dealt with separately, and calls are assigned to a single or a number of contiguous slots on a single subchannel. Calls cannot be assigned over channel boundaries. Thus, if a call cannot be assigned in its entirety to a single subchannel, then

insertion is attempted in the next subchannel, and so forth until the last subchannel in the channel set. An MCC insertion strategy does not take notice of channel boundaries. It views the entire channel set as one long slotted channel, with all single channels placed end to end. Thus, calls can be assigned across single channel boundaries. A contiguous insertion strategy can only assign a given call to contiguous free slots. For example, if a call requires three slots, then it can only be assigned to three sequential free slots in the channel/s. A non contiguous strategy can assign a call across a number of contiguous slots. V. RESULTS AND DISCUSSION A number of simulation tests were carried out for a single and two-channel system. Voice and data traffic were generated using Poisson sources. After an equilibrium warm up period of 100 frames for each run/test, 10 000 frames per run was found enough to give results with 95% confidence interval. The effect of various insertion strategies and different queuing systems for data traffic were tested under a range of voice and data traffic conditions and increasing voice reservation ratio. The results collected from these tests include the overall channel utilisation for both voice and data traffic, data traffic time delay, and voice and data blocking ratios. The results show a compromise between the different insertion strategies and queue systems with respect to all above parameters depending on the range of voice traffic conditions. The SCC insertion strategy is the simplest, and perhaps the most inefficient strategy, but it requires the least overhead to function, compared to all other strategies. The MCC strategy requires more overhead to function but is superior to SCC case regarding the queue length, blocking and delay. The MCNC strategy is the most complicated of all. It must contain a mechanism to be able to track a call across multiple channels. The three queuing systems have also influenced the results obtained. The head of line system is effectively a lucky dip for call assignment to channels. The smallest call first system assigns more calls to channels compared to the largest call first system. Thus, the queue length is longer in the latter case and also the channel utilisation is less. Figures (3-6) show a sample of the results obtained from simulation. Figures (3 and 4) show the channel occupancy/utilisation (Erlang) for voice and data traffic for increasing number of data packet streams. Figures (5 and 6) show the average delay ratio of data calls and the average queue length (slots) for increasing number of data packet streams. The use of a multichannel system has the advantage of maximising the overall bandwidth efficiency. This may seen to contradict the Shannon theory principles, but it actully improves the overall system bandwidth efficiency in the mobile radio channel. It is also anticipated that using a multichannel system encourages the use of OFDM system implementation. The latter is proved to substantially boost the efficiency of the newly proposed protocol bandwidth utilisation. Since the contention protocols proved to be inefficient when these are used in a mobile radio environment as in [10,11], the newly proposed protocol takes the advantage of the contention protocols only for the reservation purposes and not for the provision of service for the data and voice mobile users. Thus, the deficiency of the contention protocol that is used on the R-Ch constitutes in a degradation of less than 5% of the overall system bandwidth. VI. CONCLUSION A new protocol based on the duality of reservation and contention protocols has been proposed. The procedure of its operation is described for different types of traffic requirement. A simplified and abstracted simulation model which mimics the operation of the protocol is given for different insertion

strategies. The results show that the inferiority of which strategy performs better depends on system requirements. The overall performance of the proposed protocol is still under investigation. VII. REFERENCES [1] D. Goodman, R. Valenzuela, K. Gayliard, and B. Ramamurthi, “Packet reservation multiple access for local wireless communications,” IEEE Trans. Comm., vol. COM-37, pp. 885-890, Aug. 1989. [2] S. Nanda, D. Goodman, and U. Timor, “Performance of PRMA: A packet voice protocol for cellular systems,” IEEE Trans. Vehicular Technology. Vol. VT-40, pp. 584-598, Aug. 1991. [3] A. Glass and R. Brewster, “Reservation with contentionbased traffic demand assignment protocol for land mobile radio communications,” IEE 5th International Conference on Mobile Radio and Personal Communications (Coventry, UK), Dec. 1989, pp. 6-9. [4] P. Roorda and V. Leung, “Dynamic control of time slot assignment in multi-access reservation protocols,” IEE Proc.Communications, vol-143, no.3, pp. 167-175, June 1996. [5] O. Kubbar, and H. Mouftah, “Multiple access control protocols for wireless ATM: problems definition and design objectives,” IEEE Comm. Magazine, pp. 93-99, Nov. 1997. [6] N. Amitay and S. Nanda, “Resource auction multiple access (RAMA) for statistical multiplexing of speech in wireless PCS,” IEEE Trans. Vehicular Technology., vol. VT-43, pp. 584-595, Aug. 1994. [7] G. Wu, K. Mukumoto, and A. Fukuda, “Analysis of an integrated voice and data transmission system using packet reservation multiple access,” IEEE Tran. On Vehicular Tech, Vol.43, No.2, pp.289-297, May 1994. [8] M. Frullone, G. Riva, P. Grazioso, and C. Carciofi, “PRMA Performance in cellular environment with self-adaptive channel allocation strategies,” IEEE Trans. Vehicular Technology., vol. VT-45, pp. 657-665, Nov. 1996. [9] G. Bianchi, F. Borgonovo, L. Fratta, L. Musumeci, and M. Zorzi, “ C-PRMA: A centralised packet reservation multiple access for local wireless communications,” IEEE Trans. Vehicular Technology. Vol. VT-46, pp. 422-435, May 1997. [10] S. Chakraborty and S. Wager, “A new approach for medium-access control for data traffic and its adaptation to the GSM general packet radio services,” IEEE Trans. Vehicular Technology, Vol. VT-48, pp. 240-248, Jan. 1999. [11] T. Woo, “FMAC: A highly flexible multiple-access protocol for wireless communications systems,” IEEE Trans. Vehicular Technology, Vol. VT-48, pp. 883-890, May 1999.

U- or D-Chs Slot 1 2 … … Voice Traffic

Ch. 1

Data Traffic Buffer (Queue)

9 10 Service Provided

2 . . . n

Figure 2. Schematic of protocol multichannel simulation model.

1

Frame No

2



m-1

m

m+1

ι-1



ι

R-Ch (MUs to BS) C-Ch (BS to MUs) U-Ch (MUs to BS) D-Ch (BS to MUs)

(a) Layout Frame m Slot No

1

R-Ch (MUs to BS)

2 v21

3

4

D1

Frame m+1

5

6

v22 v23

7 D2

8

v24

1

2

v25

3

4

D1

5

6

v21

7 v22

8 D2

va11,v112,D113,vb22,Da14,vc15,v433,vd16,…

C-Ch (BS to MUs) U-Ch (MUs to BS)

v1a

v1b

Da

v1c

v1d

D-Ch (BS to MUs)

(b) Operation Figure 1. Protocol channels layout and operation example.

Average Channel Voice Occupancy

Average Channel Voice Occupancy vs Data Streams

Average Data Call Delay vs Data Streams

12

0.7

10

0.6

8

0.5 0.4

6 0.3

4

0.2

2

0.1 0

0 1

2

3

4

5

6

1

-0.1

2

3

Data Streams

Figure 3. Channel voice occupancy vs. data stream for HOL SCC.

5

6

Figure 5. Average data call delay ratio vs. data stream for HOL SCC.

Average Channel Data Occupancy vs Data Streams

Average Channel Data Occupancy

4

Data Streams

Average Data Queue Length vs Data Streams

7 18

6

16 14

5

12

4

10 8

3

6

2

4

1

2 0

0 1

2

3

4

5

6

Data Streams

Figure 4. Channel data occupancy vs. data stream for HOL SCC.

-2

1

2

3

4

5

6

Data Streams

Figure 6. Average data queue (slots) vs. data stream for HOL SCC.