VIRTUAL CONNECTION TREE BASED ALGORITHMS FOR 3G MOBILE COMMUNICATION SYSTEMS

Transcription

1 VIRTUAL CONNECTION TREE BASED ALGORITHMS FOR 3G MOBILE COMMUNICATION SYSTEMS A. L. Heath and R. A. Carrasco School of Engineering and Advanced Technology, Staffordshire University PO BOX 333, Beaconside, Stafford ST18 0DF, Staffordshire, U.K. s: Abstract An algorithm for mobile Asynchronous Transfer Mode(ATM) handover that offers service adaptability and efficient allocation of wired resources for connection rates is evaluated in this paper. In future 3G systems there is a requirement for wireless connection rates of up to 2Mbits/sec, it is the objective of this paper to investigate the effects of this increase in rate on a service adaptable handoff algorithm. The algorithm is based on the Virtual Connection Tree (VCT) and provides fast rerouting for delay sensitive services and cell sequencing for cell loss sensitive services. An ATM handoff model is used to analyse the service adaptable algorithm in terms of cell loss and end-to-end delay. Results are discussed proving the advantages of the algorithm for future ATM mobile communications. 1. Introduction In the near future bandwidths of up to 2Mbits/s will be available using 3G mobile communication systems through IMT-2000 [1, 2], this will allow the most demanding of today s internet applications to be carried out by portable devices allowing world wide roaming. With this trend it is important to have an efficient signalling system and provision for handoffs to take place as the user travels from one cell to another. Resources in a mobile environment are limited and therefore require efficient administration if multimedia services are to be provided. The mobility that gives users freedom to roam complicates the management of calls at the central processor. Handoff or handover is the process of changing a user s reference BS, this makes transmission of information difficult since a new path must be provided each time this process takes place [4]. Seamless handover of ATM virtual connections is not currently supported in the ATM layered protocol; a new route is allocated each time a mobile changes BS. The delay introduced by the reestablishment of the routes is unacceptable for broadband services or delay sensitive connections since it can lead to disruptions or degradation of the connection [4]. In [4] a service adaptable handover algorithm was evaluated that avoids signalling and queuing if the service requires fast re-routing and assures the delivery most cells in order if the service can not support cell loss or cell delineation. The algorithm is based on the VCT which has been recognised as the fastest method but requires incorporation of a lossless handover mechanism to assure cell sequencing. An ATM handover model is developed in [4] to simulate the effects of the algorithm in terms of cell loss and cell delay. Connection rates of 64, 128 and 256kbits/sec were considered for constant and variable rate services, advantages were provided at these values, however with future rates increasing the ATM model will be used to evaluate the performance of the algorithm at these higher rates. This paper is organised as follows; In section 2 the Virtual Connection Tree with a Service Adaptable Handoff Algorithm is described. In Section 3 the model used to simulate the Handoff Algorithm is described. Results are obtained and analysed in Section 4 and the conclusions are drawn in Section 5.

2 2. Virtual Connection Tree with a Service Adaptable Handoff Algorithm VCT is an ATM oriented strategy that avoids involving the central processor during handoffs. A VCT consists of cellular BSs with radio transceivers connected to switching nodes through an ATM wired infrastructure and mobile units, which transmit information over the shared radio channel. The root of the tree is the MSC and the leaves are the BSs, the area covered by the VCT is the Neighboring Mobile Access Region (NMAR). The service adaptable algorithm [4] provides service adaptable handoffs, two options are considered, once the BS accepts the connection in it s cell. If the service requires fast re-routing and the loss of several cells is not inconvenient, then the standard VCT can be applied. Conversely, if the service requires all cells to arrive in the correct order, then the VCT with extended cell sequencing algorithm is employed. Forward handoff has been considered, this is when the MS stops the wireless connection with the BS and establishes a link with a new BS. The backward handoff where the new wireless link is established through the old BS is similar. 3. ATM Based Handoff Model A model was created in [4] to simulate the ATM handoff procedure, it simulates traffic in both the wired and wireless environments. ATM cells are transmitted, simultaneously in both directions from the end points of the system according to ATM technology to the other end point. Connections are interrupted momentarily and re-routed due to handoffs, the general system is shown in Figure 1. If initially BS1 was the serving BS, when a handoff occurs the connection is switched to BS2, and for the next handoff switched back to BS1, this is repeated several times during the simulation. This simulates a MS travelling across multiple BSs using only two BSs. Figure 2 indicates the path taken by the information and the possible delays that can be encountered in the simulated model, for more information refer to [4]. The model was built using BONeS Designer, a software package that provides an ATM library, that has been improved to allow ATM representation in the mobile environment. The performance in terms of cell loss, loss of sequence and end-to-end delay have been modelled for the two options considered in the VCT algorithm. Simulations were run for the VCT with and without the extended cell sequence algorithm, since the objective is to investigate the effect of handoff the simulations do not contemplate the effect of other users, either in the air link or the wired part. The general parameters used for the simulations are shown in Table 1. Mobile connections with a call duration of 360 seconds over a microcellular system were simulated. The number of handoffs considered during the simulation is high in order to obtain abundant results to complete a thorough analysis, a high velocity for the MS and a small cell size is therefore assumed. Different types of connection (constant and variable rate services) at different rates are considered. 4. Results and Discussion Firstly, the standard VCT was analysed with the intention of detecting if cells were lost during handover. The simulations were run for both constant and variable transmission rate connections between 64kbits/sec and 2Mbits/sec the results obtained are shown in Figure 3. The lost cells are downlink cells that have been sent to the old BS when the MS is at the new BS, since there is no algorithm to re-route these cells. With constant rate connections the relationship between the number of cells lost due to handoff and transmission rate is linear, whereas for variable rate this relationship can be seen to be non-linear as shown in Figure 3. The reason for this is the bursty manner that the variable rate traffic is generated since in some handoffs no cells are lost and in others many are lost. There are more cells lost for constant rate traffic (average values), but more cells are lost for bursty traffic in one instance for bursty traffic (peak values) these types of loses are more detrimental to the QoS. The results show that cells can be lost during handoffs when the standard VCT is applied. If a low constant rate service such as voice is considered then the number of cells lost per handoff is

3 negligible. For services at higher rates the number of cells lost more cells are lost and for services such as data this is not acceptable, so applying the VCT can present problems. Secondly, the VCT with extended cell sequencing was examined, the same conditions were applied as in the previous case. This option maintains all cells, however the delay incurred due to the forwarded cells is important to some services. The algorithm prevents the loss of any cells by forwarding the wrongly routed cells to the new location. The delay for these cells is the largest since they are buffered and sent through a longer path, forwarding of the cells occurs in the downlink and therefore this is the direction of study. Next, the cell sequence was examined both with the standard VCT and the VCT with extended cell sequencing, this is illustrated in Figure 4. The are more cell sequence errors for variable rate traffic than constant rate traffic, this is also due to the bursty nature of the variable rate cell creation. The distribution of sequence errors for variable rate traffic is very similar for the standard VCT and extended cell sequence VCT, with the quantity of errors increasing with transmission rate, this concept requires further investigation. With constant rate traffic there are no sequencing errors obtained for the extended cell sequence VCT, but there are random errors for the standard VCT. Even though the cell sequence errors for variable rate traffic are similar for both VCTs, they do not have the same overall amount of errors, since with extended VCT there are no lost cells, therefore the cell error rate (CER) has been calculated for each condition. The CER was calculated, using Equation 1. no.of lost cells + no. of sequencing errors no.of cell errors Cell Error Rate, CER = = (1) no. of cells transmitted no. of cells transmitted The CER for the different conditions is illustrated in Figure 5a, the CER for constant rate traffic is relatively constant with the standard algorithm ( slightly increasing with transmission rate) and no errors occurred with the extended VCT. The CER for variable rate traffic was higher than at consequent transmission speeds than the constant rate traffic. The variable system is more unstable due to the bursty nature of the traffic generation, there is not a linear relationship between CER and transmission speed, this requires further investigation. Next, the end-to-end cell delay is considered for the four different conditions, this is illustrated in Figure 5b. For constant rate connections the mean delays incurred are similar for the different transmission rates. The mean delay of the standard VCT is 340µs for the at all transmission rates and for the extended VCT a 470µs delay at 64 kbits/sec to a 570µs delay at 2M bits/sec this rate increases linearly. The reason that the extended VCT has higher delays is that the forwarded cells have to be buffered waiting for the algorithm. The first cell forwarded is the one with the highest delay since it was the first cell to be buffered and had to wait longer for the algorithm than cells arriving later. The difference between the end-to-end delay for successive forwarded cells is almost constant, at higher transmission speeds more cells are forwarded. In variable rate connections the average and maximum delays are increased, this is due to the algorithm not only on the out of band signalling but also on the generation of ATM information cells (i.e. first cell in the uplink). The generation of the cells is variable so it may be possible that no cells are transmitted in one direction when a handoff takes place, this increases the algorithm duration time and consequently the end-toend delay of forwarded cells. The relationship between the transmission rate and the delay is not linear. These results show that if the VCT with extended cell sequencing algorithm is applied to every type of service connection, significant delays may be introduced, this delay will not be suitable for delay sensitive types of service such as voice.

4 5. Conclusions A model has been used to evaluate the handover procedure in terms of cell loss, cell sequencing and end-to-end delay at transmission speed of up to 2 Mbits/sec. The results obtained show that different services need to be treated differently when handoffs take place, for example, on average about 9 cells are lost in every handoff when the standard VCT is applied over a constant connection at a rate of 64 kbits/sec. This can be acceptable for voice services where the loss of a few cells is permitted as long as the delay is kept low. However data connections can not allow the loss of information and will be damaged. When the VCT with extended cell sequencing option is applied over the same connection no cells are lost at the cost of incurring peak delay of 0.075s in comparison to the average delay measured to be 470 µs for this option and 340 µs for the standard VCT. In this situation the opposite occurs voice can not tolerate delay and consequently the connection will be affected. When the VCT extended cell sequencing is employed no cells are lost at all at any speed or with constant or variable rate connections, however some cell sequencing errors do occur for variable transmission rates at high speeds (2 Mbits/sec and Mbits/sec), these have a similar distribution as the standard VCT at the same connection rates and service type, this phenomena requires further investigation. The performance of the extended VCT in terms of cell error rate is much better than the standard VCT. 5. References [1] Samukic A., UMTS universal mobile telecommunications system: Development of standards for the third generation IEEE Transactions on Vehicular Technology v 47, n 4, (Nov 1998), p [2] Godara, L.C.; Ryan, M.J.; Padovan, N. Third generation mobile communication systems: Overview and modelling considerations Annales des Telecommunications/Annals of Telecommunications v 54, n 1, (1999), p [3] Karol, M., Veeraraghavan M. and Eng, K.Y., Implementation and Analysis of Handoff Procedures in a Wireless ATM LAN IEEE Globecomm London [4] Larrinaga, F. and Carrasco, R.A., Virtual Connection Tree Concept Application over CDMA Based Cellular Systems IEE Colloquium on ATM Traffic in the Personal Mobile Communications Environment, Savoy Place, London 11 Feb Table 1 Simulation Parameters Parameter Value Service Type Variable or Constant Mean Rate 64, 128, 256, 394, 512, 1024, (kbits/sec) Call Duration 360 (sec) Air Link Capacity (bit/sec) Radio Capacity Delay, δ air rate ATM cell size/capacity air link (sec) S Fabric Delay (Switch) δ i (sec) Capacity or Rate of the Link 100 or 150 (Mbit/sec) linkrate Link Rate Delay δ i ATM cell size/capacity of link (sec) Link Propagation delay per unit length (sec/miles) linktrans. Transmission Link Delay δ Link propagation delay per unit length * Length i of the link (sec) Wireless Connection Establishment Between 25 and 75 (msec) Delay δ wirelessest Velocity of Mobiles Between 20 and 70 (miles/hour) Diameter of Cells 0.1 (miles)

5 Endpoint Fixed Networks MSC BS 1 BS 2 Figure 1 ATM Handover simulation Model δ air int δ air rate δ retrans. δ air trans. S δ BS Q δ BS linkrate δ BS n i = 0 δ + δ S i n Q δ i i = 0 linkrate i + δ linktrans i. linktrans δ BS. MS Radio Channel BS Link Intermediate Nodes & Links S δ MSC δ wirelesses t. Intermediate Nodes & Links Q δ MSC linkrate linktrans. δ δ MSC MSC l S δ + δ k k k = 0 l Q δ k k = 0 linkrate + δ linktrans k. Link old BS S δ oldbs Q δ oldbs linktrans δ oldbs linkrate δ oldbs. Link Intermediate Nodes & Links Delays MSC Downlink Forwarded Cells Link m j = 0 j Intermediate Nodes & Links S δ + δ m Q δ j j = 0 linkrate j + δ linktrans j Fixed-end Terminal. Normal Routing Figure 2 End-to-End mobile connection and delay

6 Cells Lost per Handoff 400 Number of Cells Lost Transmission Rate (kb/s) AVG-Const PEAK-Const AVG-Var PEAK-Var Figure 3 Cells lost due to handover, standard VCT Total Cell Sequence Errors Cells Lost 2000 Transmission Rate (kbits/sec) Std Alg (Const) Ext Alg (Const) Std Alg (Var) Ext Alg (Var) Figure 4- Cell sequencing errors Cell Error Rate for ATM Handoff System Mean End-to-End Cell Delay Cell Error Rate Time (µs) Transmission Rate Transmission Rate (kbits/sec) a) Std Alg (Const) Ext Alg (Const) Std Alg (Var) Ext Alg (Var) Figure 5 a) Cell Error Rate, b) End-to-end Cell Delay b) Std Alg (Const) Ext Alg (Const) Std Alg (Var) Ext Alg (Var)

7 VIRTUAL CONNECTION TREE BASED ALGORITHMS FOR 3G MOBILE COMMUNICATION SYSTEMS Alison Heath MEng, BEng (Hons), AMIEE and Professor Rolando Carrasco BSC(Hons), PhD, CEng, FIEE School of Engineering & Advanced Technology 11 April 2000

8 Index 2 Introduction, to 3G communication systems Research Objectives Background Theory Asynchronous Transfer Mode,Virtual Connection Tree, Bandwidth ATM Model Mobility and Handoff Problem, Algorithm and Delay Improve QoS by measuring cell loss, delay and sequencing errors Results Conclusions and Future Work

9 Introduction 3 Evolution to 3G mobile communications High bw, broadband systems USA Korea Voice IS-95 Cdma2000 Single set of services Anytime, anyplace communications Packet based transport mechanisms Data IS-136 IMT-2000, UMTS Voice GSM EDGE WCDMA Data HSCSD GPRS Europe Japan Korea Second Generation Systems Third Generation Systems ATM TCP Transport Mechanism IP Transport Medium

10 Research Objectives 4 Investigate the issues affecting the access to mobile ATM networks Accommodate mixed types of information Voice, data, images - Multimedia, and Mobility Simplify re-routing of information (handoffs) Minimise time delay, and loss of cells

11 Asynchronous Transfer Mode (ATM) 5 Different types of services at different traffic rates using the same unique Universal Network Common Network Layer for all types of traffic Intelligent Network that assures QoS UMTS and Wireless ATM (Mobile) Control Plane Higher Layers Management Plane ATM Adaptation Layer ATM Layer User Plane Higher Layers Physical & Convergence Layer ATM Protocol Reference Model

12 Virtual Connection Tree (VCT) 6 Endpoint Switch A Look Up Table VCN-in Port-in VCN-out Port-out vc1 1 vc2 3 vc9 5 vc10 3 vc7 4 vc8 2 vc13 4 vc14 6 BS 1 Neighbouring Mobile Access Region vc1 vc8 vc9 MS 1 associated to BS A 5 6 vc14 vc10 vc2 vc7 vc13 vc11 vc3 vc12 MSC root vc6 vc4 vc5 BS 5 Static Part Dynamic Part Pre-established Routes MS 1 BS 2 BS BS leaves 4 3

13 VCT Wired Bandwidth Allocation 7 MSC 20 Mbit/sec 20 Mbit/sec 8 Mbit/sec 8 Mbit/sec 8 Mbit/sec 8 Mbit/sec BS 1 BS 2 BS 3 BS 4 4 Mbit/sec 4 Mbit/sec 4 Mbit/sec 4 Mbit/sec

14 New Bandwidth Allocation 8 MSC VP switches Reserve PVP for BSs 20 Mbit/sec vp1 vp2 20 Mbit/sec vp3 vp4 PVP between BSs vp1-2 8 Mbit/sec 8 Mbit/sec 8 Mbit/sec 8 Mbit/sec vp3-4 BS 1 BS 2 BS 3 BS 4 4 Mbit/sec 4 Mbit/sec 4 Mbit/sec 4 Mbit/sec

15 Bandwidth Comparison Band-Width Allocation(MSC) Band-Width (Mbit) BSs 50 Channels Old VCT New VCT Active Users in VCT Band-Width Allocation(BS) Band-Width (Mbit) Old VCT New VCT Mbit/s per connection Active Users in VCT

16 ATM-Based April 2000 Handoff Model Endpoint MSC Fixed Networks Objectives Test Algorithm Performance Compare to Simple VCT Measurements Cell Loss Cell mis-orderring Cell Error Rate Cell Delays BS 1 BS 2 Conditions Diff. Call Rates Diff. Services Mobility

17 Service Adaptable Handoff Algorithm 11 Handoff request at BS Wireless bw available? Yes No Blocked Connection Cell Loss Sensitive VCT with extended cell sequencing Algorithm Type of service? Delay Sensitive Standard VCT Algorithm Transmission from new BS (VPI)

18 End-to-End mobile connection and delay 12 MS Radio Channel BS Link Intermediate Nodes & Links Downlink Forwarded Cells Normal Routing Intermediate Nodes & Links Link MSC Link Intermediate Nodes & Links Fixed-end Terminal old BS Link Intermediate Nodes & Links

19 Simulation Parameters 13 Parameter Service Type Mean Rate Call Duration Air Link Capacity Radio Capacity Delay, δ air rate Fabric Delay (Switch) δ S i Capacity or Rate of the Link Link Rate Delay linkrate δ i Link Propagation delay per unit length Transmission Link Delay linktrans δ i Wireless Connection Establishment Delay δ wirelessest Velocity of Mobiles Diameter of Cells. Value Variable or Constant 64, 128, 256, 394, 512, 1024, (kbits/sec) 360 (sec) (bit/sec) ATM cell size/capacity air link (sec) (sec) 100 or 150 (Mbit/sec) ATM cell size/capacity of link (sec) (sec/miles) Link propagation delay per unit length * Length of the link (sec) Between 25 and 75 (msec) Between 20 and 70 (miles/hour) 0.1 (miles)

20 Cell Loss Measurement Lost Cells Due to Handover Std Alg (Constant) Time (sec) kb/s 394kb/s 512kb/s 1024kb/s 1444kb/s 20000kb/s At first handover bits/sec Constant Variable 256k k M Lost Cells Due to Handover Std Alg (Variable) Time (sec) Lost Cells No cells are lost using Extended VCT Algorithm

21 Cell Loss Measurement 15 Average number of Cells Lost per Handoff Lost Cells Linear relationship Transmission Rate (kb/s) AVG-Const AVG-Var

22 Cell Sequence Loss (Constant) 16 Cell Sequence Errors - Std Alg (Constant) Handoff Time 1024 kb/s 512 kb/s 394 kb/s Random loss of sequence Time (sec)

23 Cell Sequence Loss (Variable) 17 Cell Sequence Errors - Std Alg (Variable) Handoff Time 2 Mb/s More errors as speed increases with both variations of the Algorithm Time (sec) Cell Sequence Errors -Ext Alg (Variable) 1444 kb/s 1024 kb/s Handoff time 2Mb/s 144 kb/s Time (sec)

24 Cell sequence errors 18 Cells Lost Total Cell Sequence Errors 76 cells out of sequence for Standard Algorithm and 75 cells out of sequence for Extended Algorithm for variable rate traffic at 2Mbit/sec Transmission Rate (kb/s) Std Alg (Const) Ext Alg (Const) Std Alg (Var) Ext Alg (Var)

25 Cell Error Rate 19 Cell Error Rate Cell Error Rate for ATM Handoff System Higher CER, bursty sources max = Constant CER Transmissio n Rate Std Alg (Const) Std Alg (Var) Ext Alg (Const) Ext Alg (Var) no. of lost cells + no. of sequencing errors Cell Error Rate, CER = = no. of cells transmitt ed no. of cell errors no. of cells transmitt ed

26 End-to-End Cell Delay (Constant) 20 Delay (sec) Downlink Cell Delay (Constant 256 kb/s) (Alg) Time (sec) Handoff Time Forwarded cells cells 388 cells Delay (sec) Downlink Cell Delay (Constant 2 Mb/s) (Alg) Time(sec) Handoff Time Forwarded cells Std Algorithm has constant delay, Ext Algorithm delay increases with transmission rate, more packets are forwarded at higher rates

27 End-to-End Cell Delay (Variable) 21 Downlink Cell Delay (Variable 256 kb/s) Downlink Cell Delay (Variable 2 Mb/s) Delay (sec) cells at 1 st handover (Alg) Time (sec) Handoff Time Forwarded cells Delay (sec) cells (Alg) Handoff Time Variable traffic delays are higher, traffic is bursty Time(sec) Forwarded cells

28 Mean End-to-End Cell Delay 22 Mean End-to-End Cell Delay 800 Time (µs) Transmission Rate (kbits/sec) Std Alg (Const) Std Alg (Var) Ext Alg (Const) Ext Alg (Var) Variable traffic larger mean delay than constant traffic

29 Conclusions 23 Service adaptable ATM handover algorithm with reduce bandwidth reservation and seamless inter-vct handover (ATM level) Cell Loss Constant rate Std Alg number cells lost increases as transmission speed increases More cells lost for constant rate traffic but more cells are lost in each instance with variable traffic (387 const, 215 No cells lost with the Extended VCT At the cost of increased time delay

30 Conclusions 24 Cell Sequencing Errors Constant rate: Std Alg there are small number of errors and none for Ext Alg Variable rate: similar distribution for both Algorithms at high speeds, no errors at low speeds (75 Cell Error Rate Constant rate Std Alg, CER stays relatively constant as transmission speed increases (0.0045) Variable rate Std Alg, CER is variable, in the same order as constant Variable rate for Ext Alg has low CER (5.83x10-5 ) End-to-End Delay Higher values for Ext Alg than Std - due to forwarding of cells and higher values (780µs extended 610µs slight increase in delay as transmission speed increases

31 Future Work 25 Investigate reasons for cell sequence errors at high speeds Implement similar Algorithm for TCP Implement a packet switched B-ISDN system to run on top of IP Investigate multiple access schemes such as cdma2000, WCDMA, TDMA, FDMA

32 Any Questions? School of Engineering & Advanced Technology Thank you for Listening 11 April 2000

33 VIRTUAL CONNECTION TREE BASED ALGORITHMS FOR 3G MOBILE COMMUNICATION SYSTEMS Alison Heath MEng, BEng (Hons), AMIEE and Professor Rolando Carrasco BSC(Hons), PhD, CEng, FIEE School of Engineering & Advanced Technology 11 April 2000