Dynamic Resource Allocation in Hierarchical ... - Semantic Scholar

1 downloads 0 Views 185KB Size Report
hybrid TD/CDMA scheme by comparing the call blocking, handoff failure, as well as the ... bit rate for a specific service is fine-tuned by selecting an appropriate ...
DYNAMIC RESOURCE ALLOCATION IN HIERARCHICAL CELLULAR SYSTEMS FOR MULTI-MEDIA TRAFFIC A S Jahan, R G Hernandez, A H Aghvami Centre of Telecommunications Research, King’s College London,WC 2R 2LS London, UK E-mails: [email protected], [email protected]

Abstract - This paper examines the performance of a new Dynamic Resource Allocation (DRA) Strategy for a 3G Hierarchical Cellular Structure (HCS). With rising demands on wireless communication systems there is a need to improve the resource utilisation of mobile network. This can be achieved with the implementation of HCS along with a suitable self-adaptive DRA strategy.

within the individual layers and the layers’ hierarchy. The resources can include the spectrum bandwidth as well as channels based on frequency, code, time slots or a combination of these. Satellite

The main focus of attention is to evaluate the ability of the new DRA strategy to deal with multimedia traffic. Simulation results confirm that the modified DDRA performs adequately in the HCS under Multimedia traffic conditions.

M acrocell

M icrocell

I. INTRODUCTION

A

S mobile teletraffic demands increase, better methods are needed to utilise and manage available network resources. One method, which is generating a great deal of interest, is the implementation of cellular system structures with hierarchical cellular overlays in future networks. These Hierarchical Cellular Structures (HCS) have become a major requirement in 3G mobile systems such as UMTS and IMT2000. A HCS, Fig.1, consists of two or more layers of small cell clusters overlaid by larger cells. It can be formed of a picocellular layer to provide service to indoor environments and a microcellular layer to provide service to indoor/outdoor environments. Both layers can be further overlaid by a macrocellular layer to deal with requests from users in outdoor areas, e.g. cities or rural areas. At the highest hierarchical level there can finally be communication satellite beams overlaying all the terrestrial layers. Such a structure permits the features of different cell types to be taken advantage of simultaneously for: • improving coverage of the network • increasing capacity of cells • balancing the loads between layers • providing services to users efficiently with different mobility characteristics, i.e. slow or fast For a HCS to function properly the resources of the structure need to be managed and allocated efficiently

0-7803-7589-0/02/$17.00 ©2002 IEEE

Fig 1 Three-level Hierarchical Cellular Structure 3G mobile telecommunications systems promise high quality multimedia services as well as traditional voice telephony. This requires that resource allocation strategies used in the HCS are capable of supporting a variety of services for different types of users (i.e. fast or slow). To this end, research has been carried out [1] to develop a suitable self-adaptive Distributed Dynamic Resource Allocation (DDRA) strategy for a multi-layer Hierarchical cellular structure. The aim of this paper is to examine the performance of the modified DDRA strategy in the presence of multimedia traffic. The performance evaluation is carried out using hybrid TD/CDMA scheme by comparing the call blocking, handoff failure, as well as the force call termination probability. II. RESOURCE ALLOCATION STRATEGY For the optimum sharing of resources Fixed Resource Allocation (FRA) strategies have been used effectively. But, with the ever increasing requirements and changing load distributions in the network, Dynamic Resource Allocation (DRA) strategies are more suitable. DRA schemes can be classified according to the type of cost function used to allocate resources within the network. The parameters employed in the cost functions can be used to arrange the strategies into three main groups which can adapt to the interference, traffic variations and channel reusability. For

PIMRC 2002

IV. TRAFFIC MODEL The web traffic model used, describes the typical World Wide Web browsing session, which consists of a sequence of web pages requests. During a web session, the user requests a certain number of web pages. When the document is completely downloaded, the user spends a certain amount of time for studying the information received. This time interval is also considered and it is called the thinking time or reading time, Fig.2.

Th inking Tim e Do wn l oad Ti m e Re ques t for a Web Page

Bandwidth

HCS in the third generation systems the most appropriate are those adapting to traffic variations. But, perhaps, a more significant classification of the DRA strategies is the one that divides them into centralised and distributed strategies. In the centralised DRA strategies resources are assigned to incoming calls by a central controller from a central pool. This adds to signalling load produced by communication between each Base Station (BS) controller and the central controller while seeking the status of channels in every cell at the time of channel requests. In the distributed DRA strategies the base stations assign resources to incoming calls following information about the current status of resources in their vicinity [4]. This allows for minimum communication between base stations and, being the easiest to implement, is one of the best options for use at small cell layers and, hence, the best choice for a 3G HCS.

...

III. MODIFIED DDRA STRATEGY The basic way of supporting multiple bit rates is to assign multiple slots to a user. The number of slots inside a frame can be varied dynamically in order to adapt to the transmission of changing service needs. For operation with spreading, multiple codes on a slot can also be assigned. The bit rate for a specific service is fine-tuned by selecting an appropriate combination of slot length, burst type, number of slots and coding rate. In this study a modified DDRA strategy designed for a two layer microcell-macrocell HCS is used. This has an added capability of sharing resources between layers and of actually changing the spectrum partitioning according to differing or time varying conditions of traffic. When a request for service arrives at a cell in a particular layer of the HCS, the DDRA gets the required resources from that cell. If requirements are not satisfied then it finds additional resources by: • borrowing a channel (according to SDCA[2]) in the same hierarchical level, • applying the Hand Down Procedure (HDP), or • borrowing resources from the upper hierarchical layer. The last point deals with spectrum sharing between layers in the HCS and can be seen as the ultimate step in the allocation of resources by a cell. This, however, is avoided whenever possible to minimise blocked calls in the upper layers. DDRA makes use of channel reallocation, hand-up and hand-down (similar to channel assignment seen in [3]) to decrease the blocking and dropping probability in the system. DDRA only requires that each BS has limited information about resource usage, thus making this strategy distributed. This decreases the communication between BS’s and therefore the signalling load in the system.

time

Fig 2 Web Session The aim of having the model traffic is to calculate the total time spent by a user in a session. Hence, the session can be modelled in order to define the typical behaviour explained before: • Number of pages per session • Size of each page • Downloading time for each page • Thinking / Reading time spent for each page When an arrival occurs in a cell, the first thing calculated is the number of pages the user is going to browse (N). This is done following a geometrical distribution with a certain mean. Secondly, in order to know the downloading time of each page, the size of the pages (Si) is calculated using a lognormal distribution. The downloading time is calculated with the size of the page and a fixed bit rate (Br), thus:

Td =

Si Br

(eq.1)

The thinking time (Ti) is obtained using a pareto distribution. The reading time starts when the web page is totally downloaded, and ends when the user makes a request for the next web page call. The total time can be found thus: Total time =

N

∑ Td i =1

+ Ti =

N



i =1

Si + Ti Br

(eq.2)

The process is summarized in the following scheme, Fig.3:

Arrival of the user

Number Of Pages To Visit

Size Of Each Web Page

Downloading Time Of Each Page

Thinking Time Of Each Web Page

Total Time Obtained

Fig 3 Calculation of total session time V. SIMULATION RESULTS The simulations to assess the strategy performance used an evaluation environment based on a Manhattan-like urban model with a wraparound topology for micorocells in the first layer of the structure overlaid by macrocells, with a large number users traversing the area covered, each of them with a different speed. The service used for testing with the web traffic model was based on data transmissions of 64Kbps. All other parameters and conditions were set to those used in [1], for DDRA evaluation. The performance of the Modified DDRA strategy was compared with three different resource allocation strategies, Fixed Resource allocation (FRA), Hierarchical Fixed Resource allocation version 1 and 2 (HFRA1 and HFRA2). These Strategies are defined below:

Fig 4 Overall Blocking Probability in HCS

FRA strategy: This strategy is basically the classic FRA but with an intracell channel reassignments capability. Mobility of the users is considered and, using the classification of slow and fast users, the calls are directed into the microcell and macrocell layers respectively. There are no resources shared between the two layers. HFRA1 strategy: Since the macrocell layer has a lack of resources and must be updated according to the traffic load, hand-up capability is added to the classic FRA strategy forming HFRA1. When there is no channel available to accommodate a new call in the microcell, a handover petition is performed to a neighbouring microcell, Again, if there are no resources in the neighbouring cells, the call is redirected to its preferred macrocell (hand-up).

Fig 5 Total forced Termination Probability

HFRA2 strategy: HFRA2 is similar to HFRA1. The main addition is the hand-down procedure. When a slow user is attended by a macrocell due to overflow, is sent back to a microcell, new resources are made available. Obviously, this technique reduces the blocking and the forced termination probability in comparison to FRA and HFRA1, but increases the complexity of the system. Fig 4 shows the overall blocking probability vs. the load increase. It can be seen from these blocking results that under multimedia traffic conditions the modified DDRA strategy performs much better than the FRA strategy and even better than the FRA strategy with HDP and channel reallocation Fig 5 and Fig 6 show that the forced termination probability (and therefore the number of dropped calls) as well as the probability of handover failure in the HCS is significantly reduced using DDRA rather than FRA or FRA based strategies.

Fig 6 Overall Probability of Handover Failure in the System

VI. CONCLUSIONS Performance evaluation of the modified DDRA strategy in the presence of multimedia traffic shows that the DDRA strategy provides significant reduction in dropped calls and signaling load for the network. Further that the DDRA has a clear advantage over FRA, HFRA1 and HFRA2 in terms of blocking probability, probability of force termination and handover failure probability, making its implementation worthwhile for use in a 3G HCS. VII. ACKNOWLEDGEMENTS The research reported in this paper was supported by the EPSRC under the grant no. GR/M14722. Their funding is gratefully acknowledged by the authors. VIII. REFERENCES [1] Lauro Ortigoza-Guerrero and A. H. Aghvami, “Resource Allocation in Hierarchical Cellular Systems”, Artech House Publishers 2000 [2] Kevin A. West and Gordon L. Stuber, “An Aggressive Dynamic Channel Assignment Strategy for a Microcellular Environment,” IEEE Trans. VehicularTechnol., Vol. 43, No. 4, pp. 1027 - 1038, Nov. 1994 [3] Kuen-Rong Lo, et. al., “A Combined Channel Assignment Strategy in a Hierarchical Cellular Systems,” In Proc. IEEE Int. Conf. on Universal and Personal Commun., ICUPC’97, pp. 651 - 655, San Diego, CA, USA, Oct. 1997 [4] Kwan Laurence, Yeung and Tak-Shing, Peter Yum; “Compact Pattern Based Dynamic Channel Assignment for Cellular Mobile Systems”, IEEE Transactions in Vehicular Technology, Vol.43, No 4, pp. 892-896, 1994