An Efficient Resource Allocation Protocol For Multimedia

Published Date: 02 Nov 2017

An Efficient Resource Allocation Protocol for Multimedia Wireless NetworksAbstract: In the next generation high-speed wireless networks are required to support multimedia applications (video, voice, and data). As such, it is important that these networks provide Quality-of-Service (QoS) guarantees. The communication entities in wireless networks change their connectivity through handoff when user moves from one cell to another. Once a connection is admitted into the wireless network, resources must be allocated at the negotiated level, for the duration of the connection. It is important to realize that, in a wireless mobile network where the user may move through a sequence of cells, this commitment is for all cells. If the connection is to be maintained after the user crosses the boundary between neighboring cells, the network must guarantee an appropriate level of resources in each new cell that the user traverses. In this paper a new resource allocation protocol for multimedia wireless networks is proposed that uses a combination of bandwidth reservation and bandwidth borrowing to provide QoS in terms of guaranteed bandwidth, call blocking, and call dropping probabilities.

Keywords: Multimedia, wireless network, cell, handover, call blocking, and call dropping probabilities, bandwidth reservation.

INTRODUCTION

The wireless networks to support both data and real-time interactive multimedia traffic will be highly in demand in next generation. In the face of this more complex traffic mix, where each connection may have different requirements, providing QoS guarantees poses a difficult challenge for network providers. While best-effort service may be sufficient for the data traffic produced by simple email and web applications, newer real-time interactive multimedia applications require more stringent QoS guarantees. Admission control and bandwidth allocation schemes can help provide these guarantees in wireline networks, however the problem is much more complex due to bandwidth limitations and host mobility in the wireless networks [1][ 2].

SYSTEM DESCRIPTION

The main components of the system are Cells, Base Stations, Mobile Switching Centers, and Wireline Links. The geographic area of interest is assumed to be tiled by a collection of regular hexagons referred to as Cells. The wireless communication in a cell is supported by a Base Station (BS). The base stations are connected to each other by the wireline links. Several base stations are connected to a Mobile Switching Center (MSC) that acts as a gateway from the cellular network to existing wireline networks, the Internet, and the PSTN. With the active participation of the mobile hosts and the MSC, the base stations are instrumental in initiating and finalizing hand-offs [4].

Mobile Switching Center

Cells

Base Stations (BS)

Mobile Hosts (MH)

Figure: 1 Architecture of cellular network.

The mobile hosts in a cell communicate directly with the corresponding BS that has the responsibility of handing all demands for service originated in the cell. The BS is in charge of negotiating QoS parameters, of performing admission control, and of reserving resources for ongoing connections. This could mean denying access to the new connections in order to provide an acceptable level of service to active connections. Every new connection would like to be accepted regardless of the current demand. The base station must be able to manage the load efficiently to permit continual cellular network traffic.

From the end user’s perspective, an initial blocking of the connection is far more acceptable than to terminate an ongoing connection due to lack of resources during hand-off.

CALL ADMISSION CONTROL AND RESOURCE RESERVATION SCHEME

Unlike wired networks, communication entities in wireless networks change their connectivity via handoff when they move from one cell to another. The use of micro or pico -sized cells makes the role of handoff procedures very important in maintaining the service continuity and QoS guarantees to the multimedia applications. Due to the limited bandwidth resources in wireless multimedia system there should be efficient Call Admission Control (CAC) and efficient Resource Reservation (RR) schemes in order to maintain desired QoS. CAC schemes enable the system to provide QoS to new incoming as well as existing calls [5][6][7][8]. The RR scheme, such as the use of Guard Channels (GC), is adopted to reserve resources for certain higher priority calls. Obtaining a right balance between the two opposing criteria is a big challenge.

The idea is to increase the access probability for the higher priority calls, while ensuring high overall system efficiency, in the presence of multiple QoS classes such as priority, rate adaptivity as well as different mobility. The basic idea here is of the GC scheme, which gives preferential treatment to the handoff calls by reserving a fixed number of channels exclusively for them. However, such a scheme may lead to poor channel utilization because it decreases the handoff dropping rate at the cost of increasing the blocking rate for other users. To deal with this problem, the scheme used is a dynamic resource reservation algorithm to efficiently estimate resources needed to be reserved for high priority calls, by using the distance information of mobile users in neighboring cells.

There are two approaches to resource reservation:

Fixed Reservation:

A certain percentage of available resources in a cell are permanently reserved for hand-off connections.

Statistical Reservation:

Where resources are reserved using a heuristic approach. These approaches range from allocating the maximum of the resource requirement of all connections in neighboring cells to reserving only a fraction of this amount. Numerous approaches for reserving bandwidth are there in the literature [9][10][11]. The traffic offer to a cellular network may be classified into two classes based upon that if it is real time or not. The two types are:

Class I: Real time multimedia traffic, such as interactive audio and video.

Class II: Non Real time data traffic, such as e-mail and web applications.

When a request for a connection is made to the network, the following parameters are provided:

The traffic class (I or II).

The desired amount of bandwidth, the level at which the application performs the best.

The minimum acceptable amount of bandwidth, the level of bandwidth below which the service would become unacceptable to a user.

One of the significant features of the call admission control and bandwidth reservation schemes is that in order to admit the connection, bandwidth must be allocated in the originating cell and at the same time reserved for the connection in all the neighboring cells. For a new connection to be admitted in a cell, the cell must be able to allocate to the connection in the desired bandwidth. For Class I connections, the call will be blocked unless the desired bandwidth can be allocated to it in the original cell and some bandwidth can be reserved for it in each of its six neighboring cells. During a handoff an established Class I connection is dropped if its minimum bandwidth requirement cannot be met in the new cell or if appropriate reservations cannot be made on its behalf in the new set of neighboring cells. However, Class II traffic has no minimum bandwidth requirement in the case of a handoff and a call will be continued if there is any free bandwidth available in the new cell. Numerous approaches for reserving bandwidth are there. The scheme used in statistical based on number of connections in neighboring cells, the size of connections in neighboring cells, the predicted movement of mobile hosts, and combination of these factors. In this scheme the dropping probability of Class I connections very low since the mobile hosts should find bandwidth reserved for it, regardless of the cell to which it moves. But, bandwidth may be wasted in the neighboring cells (the host can move only to one neighbor) and the blocking probability in those cells may increase because unused bandwidth is being kept in reserve. In general this scheme minimize the Call Dropping Probability (CDP) at the expense of the Call Blocking Probability (CBP) and give Class I traffic precedence over Class II traffic [12][13][14].

RESOURCE ALLOCATION SCHEMES

There are various resource allocation schemes available in the cellular networks.

Rate Based Borrowing Scheme

It is clear that keeping a small pool of bandwidth always reserved for hand-offs yields low CDP. However, in this scheme the size of the reserved pool is not determined by requests from neighboring cells, but is fixed at a certain percentage of the total amount of bandwidth available in the cell. In this scheme there is no overhead of communication between neighboring base stations to request and release reservations. This scheme does not allow bandwidth from reserved from pool to be allocated to incoming hand-offs unless the bandwidth is needed to meet the minimum bandwidth requirements of the connection. This scheme also gives precedence to Class I connection. Class II traffic does not make use of reserved bandwidth. In order to lower the CDP this scheme allows the borrowing resources from existing connections especially bandwidth. The important points to note about this scheme is that

This scheme guarantees that the bandwidth allocated to a real time connection never drops beyond the minimum bandwidth requirement specified by the connection at call setup time. This is very critical to ensuring that the corresponding application can still function at an acceptable level.

This scheme also guarantees that if bandwidth is borrowed from a connection, it is borrowed in small increments, allowing time for application level adaptation.

This borrowing scheme is also fair in the sense that if bandwidth is borrowed from one connection, it is also borrowed from the existing connections. If borrowing is necessary in order to accommodate a requesting connection (new or handoff), every existing connection will give up bandwidth in proportion to its tolerance to bandwidth loss. That is the reason this scheme is called rate based fair.

The borrowed bandwidth is returned to the degraded connections as soon as possible. Thus, the degradation in the QoS is transient and limited to a minimum.

This scheme features low call dropping probability, low call blocking probability, good bandwidth utilization and reasonable success with keeping both classes of connection operating nearly their desired bandwidth level.

PROPOSED SCHEME

The new proposed scheme is similar to the rate based borrowing scheme in the sense that it also keeps a pool of bandwidth reserved for handoffs. This reserved pool is used for only Class I handoffs, with the assumption that real time connections have more stringent QoS requirements than Class II connections. This is different from the Rate based borrowing scheme in the sense that some bandwidth is reserved dynamically in each of the neighboring cells. It reserves the bandwidth in all the neighboring cells whenever a new connection request comes to any of cell. That means whenever a connection request comes to any of the cell what it does it first checks that if the bandwidth can be reserved in all the neighboring cells in which the connection can handoff, if yes then only it will accept the connection otherwise block the connection.

The important features of this scheme are:

No Class I connection will ever have to give up bandwidth beyond the minimum level of bandwidth negotiated at the call time setup.

At the time of admission in any cell, some bandwidth is reserved dynamically in each of the neighboring cells. This reserved pool of bandwidth is used only for the handoff connections. Also, connections belonging to Class I only are allowed to use this reserved pool of bandwidth.

If the cell does not have enough residual bandwidth to accommodate an incoming call, the existing connections will temporary have to give up a certain amount of bandwidth.

While reserving the bandwidth in the neighboring cells, a cap is implemented on the size of reserved pool. Not more than certain amount of total bandwidth can be reserved in a cell.

If bandwidth must be borrowed, it is borrowed in small increments to allow time for application level adaptation.

As soon as bandwidth becomes available due to a terminating call or to a mobile host leaving the cell, the borrowed bandwidth will be returned to the degrade connections.

This scheme is fair in the sense that if bandwidth is borrowed, all connections will have to give up an amount of bandwidth proportional to their tolerance to bandwidth loss.

CELL AND CONNECTION PARAMETERS

Each cell maintains a pool of bandwidth reserved for Class I hand-offs, which represents r percentage of the total bandwidth. At the setup time, each connection specifies to the cell in which it originates a maximum bandwidth M, call desired bandwidth, and a minimum bandwidth m. The difference between these two values is the Bandwidth_Loss Tolerance (BLT) of the connection. Thus,

BLT = M – m.

Since for Constant Bit Rate (CBR) connection, M = m, indicating no bandwidth_loss tolerance and, thus, BLT=0.

Each cell maintains a local parameter, f, which represents the fraction of the BLT that a connection may have to give up in the worst case. This fraction is the Actual Borrowable Bandwidth (ABB) of the connection. Thus,

ABB = f * BLT = f (M-m).

By accepting a new call, the cell agrees that the supplied bandwidth will not fall below a certain level that we call the Minimum Expected (MEX) bandwidth that the connection is guaranteed to receive during its stay in its starting cell [4][13].

By definition,

MEX = M – ABB.

So it is clear that MEX >= m.

By doing some mathematical calculation this can be shown that

MEX = (1-f) * M + f * m .

Share_1

Share_2

Share 3

Share_4

m

M

 = 4

MEX = (1-f) M + fm

ABB = f (M-m)

Figure: 2 Main connection parameters.

To prevent borrowing from producing noticeable changes in a connection’s QoS, there is another parameter n. The ABB is divided into λ shares, each share being equal to M-m/n. This provides the basis for a method of borrowing bandwidth gradually from a set of connections whose allocated resources may be quite different. A cell is said to be operating at level L (0 ≤ L ≤ λ) when all its ongoing connections have had L (or more) shares borrowed from them.

New Call Admission

When a new call requests admission into the network in a cell at operating at level L, the cell first attempts to provide the connection with an amount of bandwidth equal to its desired bandwidth minus L shares of its ABB, that is

M – L* ABB/λ = (1-Lf/λ)* M + Lf/λ * m

If the amount of bandwidth requested exceeds the amount of bandwidth available, the cell tests to see if the call could be admitted if the cell progresses to level L+1. If transition to level L+1 will provide enough bandwidth to admit the call, the bandwidth is borrowed, the level is incremented, and the call is admitted. When the cell is operating at level L=λ no more borrowing is allowed

Resource Reservation

At the time of admission in any cell, some bandwidth is reserved dynamically in each of the neighboring cells. This reserved pool of bandwidth is used only for the handoff connections. Also, connections belonging to Class I only are allowed to use this reserved pool of bandwidth. When a connection hands off to other cell then the bandwidth reservation should be made to all the neighboring cells of the new cell in which connection has moved. While reserving the bandwidth in the neighboring cells, a cap is implemented on the size of reserved pool. Not more than 15% of total bandwidth can be reserved in a cell. The amount of bandwidth being reserved is 20% of desired bandwidth of the connection. Bandwidth reserved in each cell = 20% of desired bandwidth. Maximum bandwidth that can be reserved = 15% of total bandwidth in the cell. When the bandwidth is reserved in all the neighboring cells only then the connection is accepted otherwise call is blocked.

Handoff Management

The handoff policies differentiate between the Class I connections and Class II connections. The reserved bandwidth is used for only CLASS I connections, which are admitted only when their minimum bandwidth needs can be met. When a class I connection requests admission into the cell as a handoff, the cell checks to see if the minimum bandwidth requirement can be met with the sum of available free and reserved bandwidth in the cell if such is the case, the call is admitted into the cell and given bandwidth from the free bandwidth up to its desired level minus L shares. The connection is given the bandwidth from reserved bandwidth pool only if it is needed to reach its minimum requirement. If the minimum cannot be met using the free and reserved bandwidth, the cell tests to see if scaling to level L+1 would free up enough bandwidth to admit the call. If so, the cell scales the other calls in the cell and provides the handoff call with bandwidth accordingly. On the other hand, Class II traffic will only be dropped if there is no free bandwidth left in the cell at all. The reserved pool is not available for these types of connections.

SIMULATION MODEL

The simulation model is composed of N cells, each cell keeping contacts with its six neighboring cells. Each cell contains a base station which is responsible for the connection setup and tearoom of new connection are handoff connections as well as the reservation of bandwidth in neighboring cells [3][6][9]. Two types of connections are assumed in the simulation: a new connection, which is initiated by a mobile user and a handoff connection, which occurs when a user crosses to another cell during an outgoing connection. The inter-arrival times of new connection requests are assumed to follow a geometric distribution with mean 1/. It is also assumed that each connection may experience multiple handoffs in its lifetime. The probability that a connection experiences it first handoff is assumed to be Pn, and this probability is assumed to decrease exponentially for successive handoffs of the same connection. So the handoff probability for a connection is equal to Pn / 2n, where n is the no of handoffs already experienced by the connection. The value of the handoff probability, Pn is high for the connections with the longer duration and lower for the connections with the shorter duration.

Modeling of Cell

The network is modeled as a grid of size 1010 consisting of 100 cells. Traffic is provided to each cell at the level being measured. If a host moves out of the 1010 grid, the connection is considered to end normally i.e. hosts do not bounce back into the network. Each cell has a maximum bandwidth capacity of ‘B’ b/s. The mobile user is simulated while assuming a random movement pattern i.e. the user moves to all possible directions with equal probability when the user hands off to some other cell, the bandwidth reserved on its behalf in its neighboring cells, is also released. The values of various simulation parameters shown in table 1 are summarized. These are chosen to closely represent the realistic scenario.

simulation Results

The simulation results of new proposed scheme are compared with the rate based borrowing scheme and fixed reservation in which 5% of total bandwidth is reserved in each cell. From Figure 3, we can observe that the new scheme utilized more bandwidth compare to other scheme. Figure 4 and Figure 5 are showing that the call dropping probability of new scheme is less than rate based borrowing scheme and fixed reservation scheme. Also call blocking probability in new scheme is less than others that is shown in Figure 6 and Figure 7.

Table: 1 Values of various Simulation Parameters.

Parameters

Value

Description

N

100 Cells

No. of cells in system

B

30 Mb

Maximum B/w capacity of a cell

1/

Variable

Mean inter-arrival time of

new connection

Pn

Variable

Handoff Probability



Variable

No of shares that can be borrowed

F

0.5

Fraction of the BLT that connection may give up

Figure: 3 Comparison of Bandwidth Utilization.

Figure: 4 Comparison of CDP for Class I traffic.

Figure: 5 Comparison of CDP for Class I and II traffic.

Figure: 6 Comparison of CBP for Class I traffic.

Figure: 7 Comparison of CBP for Class I and II traffic.

CONCLUSION

In this paper, we study various resource allocation protocols. A new resource allocation protocol for multimedia wireless networks is proposed that uses a combination of bandwidth reservation and bandwidth borrowing to provide QoS in terms of guaranteed bandwidth, call blocking, and call dropping probabilities. This is different from the rate based borrowing scheme in the sense that some bandwidth is reserved dynamically in each of the neighboring cells. In this protocol, the call dropping and call blocking probability of Class I connections is very low. Also the new proposed protocol increases bandwidth utilization.