Dynamic channel allocation scheme for mobile satellite systems

4th National Conference on Telecommunication Technology Proceedings, Shah Alam, Malaysia Dynamic Channel Allocation Scheme for Mobile Satellite Systems Angeline Anthony, Mohd. Hadi Habaebi, Sahar Talib and Borhanuddin Mohd. Ali Department of Computer & Communications Systems, Faculty of Engineering, UniversitiPutra Malaysia 43400UPM Serdang,Selangor, Malaysia Abstract - A new dynamic channel allocation scheme (DCAS) has been proposed for mobile satellite systems with the aim of improving the utilization of the network resources by reducing the handoff call dropping probability (HODP) while guaranteeing a certain quality of service for the new call blocking probability (NCBP). The arriving calls are given channels based on their priority. The handoff calls have higher priority compared to the new calls, and real time traffic calls have higher priority over non real time calls. The DCAS is then combined with the shortest path routing for a adaptive channel management scheme. 1. Introduction Low Earth and Medium Earth Orbit (LEO/MEO) satellite constellations are foreseen as appropriate alternatives to the geostationary satellite systems for providing global personal communications services (PCS). Compared to geostationary satellites, these constellations offer a significantly smaller round trip delay between earth and space segments. Furthermore, the use of inter-satellite links (ISLs) has been identified as a means to provide global connectivity in space, thereby, enhancing system autonomy and flexibility, and has been retained in the design of systems like Teledesic and SkyBridge [2]. Building a network in space can decrease the dependency of the networks on the earth gateway stations. The ability of direct terminal accessibility makes it much easier to provide global roaming and to connect to any existing networks when compared with a cellular system. Also, the onboard processing capabilities make it possible to be applied in highspeed broadband signal and data transmission [5]. From a network point of view, a major benefit of a developed ISL subnetwork in space lies in the possibility to transport long distance traffic over reliable and high capacity connections. Thus forming a good base for ATM (asynchronous transfer mode) operation with the third generation (3G) and fourth generation (4G) wireless networks. 0-7803-7773-7/03/$17.00 © 2003 IEEE. ATM is regarded as one of the potential promising candidates for providing QoS guaranteed broadband telecommunication services. The concept of WATM over LEO/MEO satellite networks is considered recently because of several particular features of a LEO/MEO satellite network such as global coverage and flexible remote accessibility etc. It is possible to integrate the mobile satellite and ATM by applying the inter/intra satellite links (ISLs) as the physical layer of WATM and considering a connection between two satellites as a virtual path circuit (VPC) in ATM [2,3]. Thus we can utilize the mobility and handoff schemes of WATM to deal with routing and handoff issues in a WATM based mobile satellite network. In a mobile satellite network as shown in Figure 1, routing and handoff is extremely important in supporting the global seamless roaming and satisfying the network QoS requirements. However, in a LEO mobile satellite network, the satellite’s fast movement causes the frequent change not only in the topology of the network-in-space, but also in the footprints which is defined as the servicing area of a corresponding satellite on the earth. The latter one may result in changing the uplink/downlink, which is defined as the connection between terminals on the earth and their corresponding satellites. However, either of the two factors will cause a handoff event. Normally, there are two types of handoff defined in a mobile satellite network [6,7]. One is connection handoff, which is caused by the relative movement between end users and their corresponding satellites and this type of handoff cannot be avoided. The other one is link handoff, which is caused by the releasing of some ISLs when the satellites are in or approaching some particular situations (e.g. either of the earth poles), this type of handoff can be avoided to some degree by proper routing scheme. Obviously, a handoff event will cause re-allocation on the affected ISL channel resources. As the increasing of the arrival rate of handoff calls, channel management becomes a very important issue in providing QoS [8,9]. Satellite Satellite Satellite Satellite Satellite ATM (terrestrial network) Satellite Satellite dish ATM (terrestrial network) Satellite Satellite Satellite Satellite dish Satellite dish ground segment user terminals Figure 1: Satellite ATM Network Methods to keep the happening frequency of handoff as low as possible, and how to alleviate the side effects of handoff events on the whole network are very important in providing QoS promised service. In a mobile satellite network, we normally use three QoS elements to measure the network performance; they are end-to-end delay, new call blocking probability (NCBP) and handoff call dropping probability (HODP). There are several feasible routing schemes, e.g. Modified Dijkstra Shortest Path (MDSP) Algorithm and Rerouting Nodes Handoff (RNH) algorithm which are based on the network traffic detection, moving prediction and path reservation schemes [2,3]. In order to best utilize the network resource and provide better network QoS service, a dynamic channel allocation scheme (DCAS) is considered, which is based on the proposal of more reasonably allocating the network resource, to increase the network performance. 2. System Model 2.1 Loss Model We assume that the network traffic is symmetric (uniform distribution), and new calls can be generated with equal probability from all satellites. The interarrival time of new calls and the duration time of calls follow the exponential distribution. The whole system obeys the M/M/C/C model as shown in Figure 2. N(t) 1 Many Lines Limited number of trunks Poisson arrivalsλ λ(1-P(b)) λP(b) 2 C E[X] = 1/µ Figure 2: M/M/C/C loss model The first and second ‘M’ represent that both user’s arriving and leaving rates obey the Markov process policy; the first ‘C’ represents the maximum number of calls that can be handled concurrently in the system. The second ‘C’ represents the system capacity, i.e. no waiting queue is allowed in the system, and a call can only be either accepted or rejected. Call rejection may be due to two reasons. One is that currently there are already ‘C’ users in the system and the system cannot accept any more new calls. The other reason is that although there are less than ‘C’ users in the system, the new calls are still rejected because the ISL channels are insufficient in providing the bandwidth requirement of the new calls at some particular positions (polar regions), denoted as PN. We use PC to denote the probability of C users in the system. Where PC is given by C    λ  / C! µ  PC = C C ∑  λ µ  / i ! i= 0 (1) where, λ is the new call arrival rate, and µ is the average call duration time. The new call blocking probability is given by (2) NCBP = 1 – (1- PC ) (1 - PN ) where, PC is a fixed system parameter, but PN can be decreased by proper channel management to improve the network performance from the CBP point of view. If we assume that PN = 0, i.e. there is no releasing of ISLs at particular positions, then the new call blocking probability would be NCBP(PN=0) = 1- (1- PC )(1-0) = PC (3) 2.2 Channel Allocation Scheme Making efficient use of the ISL channels is very important in improving the network performance from the viewpoints of QoS of NCBP and HODP. There are several schemes currently used for channel reservation. The fully shared scheme (FSS) allows both handoff calls and new calls to access all available channels equally. This scheme does not protect the handoff calls and therefore cannot guarantee QoS of handoff calls. Next the guard channel scheme (GCS) reserves part of channels (guard channels) only for handoff calls, and rejects new calls when the number of occupied channels exceed the given threshold. This can improve QoS of the handoff call, but has negative effect on QoS of NCBP when the arrival rate of new calls is high. The dynamic channel reservation scheme (DCRS) handles handoff calls and new calls equally if the number of occupied channels is below the predefined guard channel (GC) threshold. The GC can also be accessed by new calls with a request probability instead of immediately blocking when the occupied channels exceed the predefined threshold. This can improve QoS of NCBP because new calls have more chances in obtaining channels compared with it in GCS. The dynamic channel allocation scheme (DCAS) proposed in [1], doesn’t separate channels into guard channels and new channels. A handoff call can access any available channels, where else a new call uses the channel resource by an access probability (PNEW ) which depends on the mobility of call (ratio of handoff call arrival rate to new call arrival rate) and the number of currently occupied channels. When the number of occupied channels is low, both handoff call and new call have almost same opportunity in using channel resource; when the number of occupied channels increase, PNEW decreases smoothly; when number of occupied channels reach threshold (H) (which dynamically depends on the QoS requirements), PNEW decreases sharply to give higher priority to the handoff call. When the mobility of call is less than 1 (more incoming new calls than handoff call), PNEW would get closer to 1, to give more channels to new calls and guarantee QoS of NCBP. But when mobility of call is higher than 1 (arrival rate of handoff call more superior to that of the new call), PNEW will drop smoothly so as to reserve more channels for handoff calls to avoid too much handoff blocking. PNEW is described as follows,       1  (0 ≤ k ≤1) PNEW =  1− exp − C     k * M + (1− k ) *    C − i  (4) where, M is the mobility of call, C is the total number of ISL channels, i is the number of currently occupied channels, and k is the traffic adjustment coefficient. We proposed a new dynamic channel allocation scheme, where the arriving calls are given channels based on their priority. As the handover arrival rate is usually greater than the new call arrival rate in mobile satellites, we give more priority to the handover calls than the new calls. The handoff calls and new calls are subdivided into real time and non real time calls. In this way we give highest priority to the real time handoff calls, followed by the non real time handoff calls, real time new calls and finally the non real time new calls. In this scheme, the handoff request of real time traffic can be executed in any available channels at any time. However, the handoff request of non real time calls, new real time calls and new non real time calls can be executed according to an access probability, which depends on the mobility of call and the number of available channels. The probability generator calculates the probability of the incoming call getting a channel and compares that value with a threshold value before deciding to accept or reject the call. We assume that the maximum capacity is 50 channels on each ISL link. Later the the capacity was increased to match the actual Teledesic model. As long as less than 60% (A1 region) of the channels are occupied, all calls will be allocated a channel as shown in Figure 3. When the number of occupied channels range between 60% to 80% (A2 region), the traffic type will be checked and the only given a channel if it satisfies a certain acceptance probability. When the number of occupied channels range between 80% to 100% (A3 region), only handover calls will be allocated, giving priority to real-time handover calls. Acceptance Probability A1 A2 60% A3 80% 100% Figure 3: Call acceptance probability vs Occupied Channels The dynamic channel allocation scheme was later combined with the shortest path routing. The new channel management scheme is shown in Figure 4. A G A1 C N B1 H A2 E A3 I O R Q B2 B3 F M B J P C1 L C2 K C3 D Figure 5: Nine node model of Teledesic system The system model is based upon the LEO ATM satellite system proposed by Teledesic. Only nine satellite nodes are considered for the simulation as shown in Figure 5. Each satellite node has 8 ISL links (4 short links and 4 long links). In this paper we assume the short links have a weight value of 1 and the long links are 1.5. The best or optimum path would be the path with the least total link weight and has available channels. YES ch <= 60% C a ll A r r iv a l P H O _ n rt NO R e je c t P r o . YES D e t e r m in e S o u r c e a n d D e s t in a t io n , a n d C a ll T y p e 60% <=ch <= 80% Id e n t if y S h o rte s t P a th 80% <= ch <= 100% N e w C a ll? NO A c c e p t P ro . g e n e ra to r NO YES NO C h e c k e a c h lin k ’s c h a n n e l a v a ila b ilit y f o r s e le c t e d S P H a n d o ff C a ll? YES N e w C a ll? NO YES R e a l T im e H O C a ll? NO YES YES R e a l T im e N e w C a ll C a ll B lo c k in g P N E W _ rt YES Channel A llo c a t io n A c c e p t P ro . g e n e ra to r NO R e je c t P r o . R e je c t P r o . P N E W _ n rt A c c e p t P ro . U p d a t e C h a n n e ls g e n e ra to r In fo r m lin k is d o w n & s t o p s e le c tin g p a th s w ith t h a t lin k C o n t in u e C a ll N ext s h o r te s t p a th E n d C a ll R e le a s e C h a n n e l a n d U p d a te Figure 4: Procedure for Channel Management Scheme The mechanism of the channel management scheme can be summarized as follows: Step 1: When the call arrives, the source and destination is determined. Step 2: The traffic type and the region of the number of available channels is determined. Step 3: The possible virtual paths are checked for channel availability, choosing from the least weighted paths to most weighted. Channels are assigned based on the Dynamic Channel Allocation Scheme. Step 4: If all the links in the virtual path have enough channels, the virtual path is setup and the connection is successful. Step 5: The call blocking probability and delay is measured to analyze the QoS of the system. probability compared to the New_RT and New_NRT calls. 1 P_accept 0.8 P_new_nrt ( i) Pnew_rt ( i) P_ho_nrt ( i) 0.6 0.4 P_ho_rt ( i) 0.2 0 60 70 80 90 100 i No. of Occupied Channels(%) Figure 6: Performance of Paccept when M =0.1 1 P_accept 0.8 P_new_nrt ( i) 3. Simulation Results and Discussion Pnew_rt ( i) P_ho_nrt ( i) The handoff call dropping probability (HODP) and new call blocking probability (NCBP) was obtained from the simulation results. Figures 6 and 7 show the relationship of the probability of acceptance of a call as the channel occupancy increases from 60% to 100%. The mobility rate was fixed as 0.1 in Figure 6 meaning that the new call arrival rate was ten times more than the handoff call arrival rate. In Figure 7, M is 5.0, meaning that the handoff call arrival rate is five times higher than the new call arrival rate. So the HO_RT and HO_NRT calls have a high acceptance 0.6 0.4 P_ho_rt ( i) 0.2 0 60 70 80 90 100 i No. of Occupied Channels(%) Figure 7: Performance of Paccept when M =5.0 The initial channel occupancy in each link was fixed to 50% and the HODP and NCBP were measured for the four traffic types. The performance of the scheme varies as the link occupancy varies. The HODP and NCBP was analyzed when the HO_RT incoming traffic amount was varied from 0% to 75%. Figure 8 shows the HODP and NCBP for the selected pair of nodes, A3 to C1, when there is 0% of the HO_RT type calls arriving and 75% of the arriving calls in Figure 9. longer routes. Handover optimization schemes should be combined to improve this channel management scheme. This system only considers the pure satellite ISLs. For a realistic LEO satellite system, the earth gateway stations must also be considered. 5. Blocking Probability 1 0.9 0.8 NCBP 0.7 0.6 HO_NRT 0.5 New_RT 0.4 0.3 New_NRT HODP 0.2 0.1 15 00 0 18 00 0 21 00 0 24 00 0 27 00 0 30 00 0 33 00 0 36 00 0 0 Call Arrival Figure 8: : Call Blocking vs Call Arrival ( 0% of HO_RT) Blocking Probability 1 0.9 0.8 0.7 HO_RT NCB 0.6 HO_NRT 0.5 New_RT 0.4 New_NRT 0.3 HODP 0.2 0.1 0 18000 21000 24000 27000 30000 33000 36000 Call Arrival Figure 9: : Call Blocking vs Call Arrival ( 75% of HO_RT) There is 0% dropping of HO_RT calls when there is sufficient amount of channel resources. HO_RT calls are only blocked when the system is overloaded. The HO_NRT calls have higher priority than the new calls, and are blocked at a lower rate. By dividing the new calls into New_RT and New_NRT, we have reduced the new call blocking probability by rejecting the New_NRT calls first and then the New_RT calls when the link occupancy reaches a critical state of about 80% or more. 4. Conclusion and Future Work The proposed channel management scheme allows the high priority calls to have a larger call acceptance probability and utilize the shorter routes while the lower priority traffic calls have a lower call acceptance probability and are passed through the References [1] J. Chen and A. Jamalipour, “Adaptive Channel Managemnet for Routing and Handoff in Broadband WATM Mobile Satellite Networks,” Proceedings of the IEEE International Conference on Communications, Helsinki, Finland, vol. 9, pp 2928 – 2932, June 11-14, 2001. [2] J. Chen and A. Jamalipour, “An improved handoff scheme for ATM-based LEO satellite systems,” Proceedings of the 18th AIAA International Communication Satellite Systems Conference and Exhibition, Oakland, CA, 10-14 April, 2000. [3] M. Werner, “ATM-based routing in LEO/MEO satellite networks with intersatellite links,” IEEE Journal on Selected Areas in Communications, vol. 15, no. 1, January 1997. [4] M. Werner, “Analysis of system parameter for LEO/ICO Satellite Communication Networks,”IEEE Journal on Selected Areas in Communications, vol. 13, no. 2, pp. 371-381, February 1995. [5] Prakash Chitre and Ferit Yegenoglu, “Next Generation Satellite Networks: Architectures and Implementations”, IEEE Communications Magazine, vol.37, no. 3, pp. 30-36, March, 1999. [6] John Naylon, “Low-Latency Handover in a Wireless ATM LAN,” IEEE Journal on Selected Areas in Communications, vol.16, no. 6, pp. 909921, August 1998. [7] Ian F. Akyildiz, “Mobility Management in Current and Future Communications Networks,” IEEE Network, pp. 39-49, July/August 1998. [8] Young Chon Kim, “Dynamic Channel Reservation Based on Mobility in Wireless ATM Networks,” IEEE Communication Magazine, pp. 47-51, Nov. 1999. [9] S. Chan, “A combined path-extension and rerouting handoff scheme for wireless ATM networks,” Proceeding of IEEE International Conference on Communications (ICC`99), June, 1999. [10]Sung Cheol Chang, “A Dynamic Wireless Resource Management Scheme for Fast Handoff in Wireless ATM Networks,” IEICE Transaction, vol. E82-B, No. 6, pp. 978-983, June 1999.