International Journal of Computer Trends and Technology (IJCTT) – volume 7 number 3– Jan 2014
Multi Attribute Scheduling using Min Potential Delay Fairness having Reconfigurable Devices in Heterogeneous Wireless System 1,3
R.Sambathkumar¹, V.Bharathi², Ravi Gunaseelan³ PG Scholars, 2Assistant Professor, Sri Manakula Vinayagar Engineering College, Puducherry.
Abstract In this paper, we present a scheduling algorithm that is used to manage resources in a heterogeneous wireless network that works with reconfigurabl e devices. The system model supports device that works on both unified and heterogeneous networks. Based on the prior multi-attribute scheduling algorithm which was implemented by Global Resource Controller (GRC) resources are managed that support various autonomous wireless systems. The various attributes considered here are spectral efficiency, minimum potential delay fairness, instantaneous and long term fairness and overall energy consumption by considering the battery life of each user. These attributes are balances using a weighted sum values and using Analytical Hierarchy Process (AHP) which collects details from various network providers. Using Matlab si mulator, we compare previous separately studied algorithms like Max-Sum Rate, Max-min Fair, Proportional fair and min power with our proposed min potential delay fairness algorithm. Keywords— AHP-Analytical Hierarchy Process, IP-CANs- IP Connectivity Access Networks, IMSIP Multi-Media Service, GRC-Global Resource Controller, RR-Reconfigurable Radios
1. Introduction Cellular industries which are driven by the economi c forces are beginning t o reduce the number of cellular providers by increasing their wireless network to grow larger based on various wireless technologies that are being introduced recent days. Recently, one of the most significant way of achieving high efficient joint wireless is by allocating fair spectrum policies that are regularly maintained. Due to the more usage of unlicensed spectrum over for commercial purposes, the usage of licensed spectrum has is likely to be reducing even though the demand for spectrum is increasing as early as 2014[1, 2]. This has sparked the interest in developing techniques which focus on improving the spectral efficiency which includes cognitive radios and networks
ISSN: 2231-2803
that adopt to the behaviour to make the efficient use of open or unlicensed spectrum. While this attempts to process the efficient spectrum utilization, the attempt to focus on the energy efficiency was neglected by network and device manufactures. Later only the wireless operators have learned that mobile device battery life efficiency is an important attribute that should be provided by operator’s services [3]. Another important parameter that needs to be concerned while going for resource allocation is the fairness allocation across all users that are present in the network.
2. System model We consider a system model based on 3GPP IMS architecture [4] as shown in Fig. 1. In our model, a cognitive user equipment (cUE) as an end-user with cognitive and reconfigurable capabilities which has the ability to access multiple IP Connecti vit y Access Networks (IP-CANs) separately or simultaneously. We use Gl obal Resource Controller (GRC) whi ch manages the resources that are allocated by the network providers. Based on these objectives, the cUE-IP-CAN link and the rate assignment is determined per mapping.
Fig.1. System Model The working operation is based on the following pattern. First, the cUE senses the nearby IP-CANs that are available and thereby register the available network with the GRC before transmitting and data to the IP-CAN. By selecting one of the available IP-
http://www.ijcttjournal.org
Page169
International Journal of Computer Trends and Technology (IJCTT) – volume 7 number 3– Jan 2014
CANs that is present the cUE obtains the IP Network connection which is used to communicate with the external hosts. We obtain IP-CANs in the following order Wi-Fi, 4G(LTE/WiMAX), 3G(HSPA/EVDO). This preference mode is taken into account if the user is not able to establish a connection to his/her preference due to any technical reasons like very high network load or interference, then he/she will try to connect to the second preference and this continues on till the cUE establishes a connection with any available IP-CAN. Then the cUE discovers, registers and communicates with the GRC application server which are described in [4]. The above procedure is shown in Fig. 2. Once the cUE establishes connection with the GRC, the cUE updates the details regarding the IPCAN to the GRC. The GRC will then be able to calculate the cUE-IP-CAN route mapping and rate assignment per mapping. This mapping information is then sent to the cUE so as to reconfigure the Reconfigurable Radios (RRs) to the corresponding IP-CANs.
Fig.2. cUE-GRC Connectivity Only after RR is configured according to cUE-GRC mapping, radio link is established with corresponding IP-CANs for data transmission. The IP-CAN base station uses the resource request from ( each cUE as 2) guidance to overcome their scheduling decisions. The GRC scheduling decision is based on long term scales (seconds or minutes) while the local abase station scheduling is based on short term scales i.e, milliseconds only. Also while scheduling for LTE and WiMAX are customizable, the IPCAN can generate a schedule in every 10 milliseconds and HSPA in 2 milliseconds and EVDO in every 26.67 milliseconds but Wi-Fi assigns channel to user for every 0.5 milliseconds to send one data frame. So, in
ISSN: 2231-2803
order to minimize the actual overhead and to be sure that cUEs and base stations of various IP-CANs uses the decision given by GRC by a scheduling interval period of 1 second is proposed for GRC.
3. Attribute function A multi-attribute resource allocation algorithm is used by GRC to determine the cUE-IP-CAN mapping and rate assignment per mapping for every schedule time t per mapping. The attributes considered here are spectral efficiency, max-min fairness, minimum potential delay fairness, proportional fairness and battery life of each user in the system. The notations used are extensions from [5, 6 ]. a)
SPECTRAL EFFEICENCY The spectral efficiency for the time interval [t, t+1] is denoted by γt which is represented by (1). It is the ratio of aggregate rate allocated to each user at time t to the total spectrum user by the system. The rate allocation to user denot ed by u ∈ U at time t, is represented in (2). This depends on : i) cUEIP-CAN assignment parameters at time t, and ii) the aggregation rate allocated to user u ∈ U by base station or access point at time t. Different IP-CANs uses different terminologies like Wi-Fi which uses CSMA/CA mechanism for resource allocation similarly OFDMA based WiMAX and LTE group 12 consecutive subcarriers in frequency domain and 6 or 11 symbols in time domain so as to form a minimum resource allocation unit. (1)
Since most of the spectrum allocated is constant to each IP-CAN, the total spectrum k is constant. Max-Sum Rate (MSR) problem occurs when the need to maximize the sum of rates allocated to each user in order to improve the spectral efficiency. If γt max and γt min represents the maximum and minimum achievable spectral efficiency for time period[t, t+1] obtained by solving MSR optimization problem respectively. The normalized spectral efficiency utility is given by,
http://www.ijcttjournal.org
Page170
International Journal of Computer Trends and Technology (IJCTT) – volume 7 number 3– Jan 2014
( 3) (7) b)
MAX-MIN FAIRNESS e)
The fairness can be calculated for each scheduling time step be checking the minimum and maximum data rate that is required. This fairness is determined by transmitting data when there is maximum and minimum numbers of users are working when the system is in use. For each GRC interval the user achieves a data rate of T bits/sec. For the admission control mechanism, if no users are blocked than the utility becomes 1 and if all users are blocked the utility becomes 0. The utility is represented by,
(4)
c)
MIN POTENTIAL DELAY FAIRNESS
When an user r tries to send a file of size w then the time taken to transfer the file is given by x[6].
BATTERY LIFETIME
The last metric in our multi-attribute resource allocation algorithm is battery lifetime. Our energy consumption model is similar to linear energy consumption [8, 9]. The energy consumption of an user during time period [t, t+1] is given by,
(8) Pt,a(x) represents to the total energy component and depends on the maximum number of data bytes ηtua which can be transferred by the radio a ∈ A of user u ∈ U during time period [t, t+1]. Then the 2nd component Po,a has two energy components which represents the extra energy spent by RRs in reconfiguring the hardware components and the extra energy that is spent for establishing a connection with the new IP-CAN. If ωt max and ωt min represents the maximum and minimum achievable overall energy consumption for time period [t, t+1]. Then the battery llife utility function is denoted by,
(5) If we decrease the total transfer time of all the sources in network, then the maximum utility can be given by,
(9)
4. Multi-Attribute resource allocation (6)
d)
PROPORTIONAL FAIRNESS
The 2nd step in resource allocation takes into account the long-term fairness. The direct mapping of Jain’s Fairness Index [7] is applied to long-term fairness utility function shown in (7). The proportional fairness utility is computed using aggressive rates allocated to each user while considering all time steps. The utility function is normalized in the interval [0, 1].
Resource allocation procedure is represented here that is used y GRC that is used to determine that mapping for cUE-IP-CAN and rate assignment per mapping. There are two steps in resource allocation problem i.e. an iterative admission control policy designed to satisfy minimum data rate requirements for real time traffic and a weightage sum of the various attributes mentioned above for best effort traffic. The method is derived from [10]. The relationship between rtua and rtua,norm is shown below,
(10)
ISSN: 2231-2803
http://www.ijcttjournal.org
Page171
International Journal of Computer Trends and Technology (IJCTT) – volume 7 number 3– Jan 2014
The optimization problem P* is resolved using,
(11) The second step in resource allocation is cUE-IPCAN mapping and rate assignment per mapping based on optimization function M* derived in (13). A method to maximize the long-term fairness utility function is described below,
Fi network. Use case 2 has users that connect to both carrier’s network and Wi-Fi network. For both cases we use 100 nomadic users having 50 using carrier 1 and other 50 using carrier 2. All users are assumed to be equipped with 2 reconfigurable radios for case 1 and 4 for case 2. The mobility is considered to be 2 mph and the user can connect to any IP-CAN [12].
6. Result We first present the result for best effort traffic condition and then for various scheduling algorthms. The over all utility function depends on the values of spectral efficiency, fairness and energy consumption which are α = 0.649, β = 0.072, and τ = 0.279 derived from the table. The overall utility result (13)for case 1 is provided in Fig. 3.
(12) M*:
(13) Where α, β and τ represents the corresponding weightage that is provided by table I through AHP [11]. Analytical Hierarchy Process (AHP) determines the weightage of different utility attributes using stake holders by pairwise comparison and ratings. TABLE I AHP Table Matrix Fig. 3 Utility for Use Case 1Ttu=0 Because of more connectivity in case 2, the avg spectral efficiency of case 2 is more than case 1. Fig.4 and Fig.5 shows the overall utility of combination of best effect and real time traffic for Use Case 1 and Use case 2.
5. Simulation and Description A MATLAB based simulation model is developed to shoe the properties of our multi attribute resource allocation algorithm for a heterogeneous wireless system. Two major cellular carriers are considered that has multiple IP-CANs in 2*2 km area. Carrier 1 has EVDO, WiMAX and Wi-Fi IPCANs and Carrier 2 has HSPA, LTE and Wi-Fi IPCANs. We consider two cases where Use case 1 has users that connect to their own carrier’s network and Wi-
ISSN: 2231-2803
http://www.ijcttjournal.org
Page172
International Journal of Computer Trends and Technology (IJCTT) – volume 7 number 3– Jan 2014
[5] T. Bu, L. Li, and R. Ramjee, “Generalized proportional fair scheduling in third generation wireless data networks,” in Proc. IEEE INFOCOM,2006. [6] Congestion Control 2: Utility, Fairness and Optimization in Resource Allocation Lecturers: Laila Daniel and Krishnan Narayanan Date:11th March 2013 [7] R. Jain, D. Chiu, and W. Hawe, “A quantitative measure of fairness and discrimination for resource allocation in shared computer systems,” DEC Research Report TR-301, Sep. 1984. Fig.4. Utility for Use Case 1Ttu
[8] L. M. Feeney and M. Nilsson, “Investigating the energy consumptionof wireless network interface in an ad hoc networking environment,” in Proc. IEEE INFOCOM, 2001. [9] N. Balasubramanian, A. Balasubramanian, and A. Venkataramani, “En- ergy consumption in mobile phones: a measurement study and implications for network applications,” in Proc. 9th ACM SIGCOMM Internet Measurement Conference, 2009. [10]
Balancing Spectral Efficiency, Energy
Consumption, Heterogeneous
Fig.5. Utility for Use Case 2Ttu
and
Fairness
in
Future
Wireless Systems with Reconfigurable Devices.. IEEE journal VOL.31 NO.5, MAY 2013
The over all utility function is given below,
[11] T. Saaty, “How to make a decision: the analytic hierarchy process,”European Journal of Operational Research, vol. 48, 1990.
(14) The overall utility of an algorithm in both cases varies significantly than the other algorithms.
7. References
[12] J. Martin, R. Amin, A. Eltawil, and A. Hussien, “Limitations of 4G wireless systems,” in Proc. Virginia Tech Wireless Symposium, June 2011.
[1] FCC, “ET Docket No. 03-222, Notice of Proposed Rule Making and Order,” Dec. 2003. [2] R. Research, “HSPA to LTE advanced,” Sep. 2009. [3] I. Mansfield, “Smartphone battery life has become a significant drain on customer satisfaction and loyalty,” April 2012. [Online]. Available: http://www.cellularnews.com/story/53523.php [4] 3rd Generation Partnership Project, “3GPP TS 23.228 V11.3.0,” IP Multimedia Subsystem (IMS); Stage 2 (Release 11), Dec. 2011.
ISSN: 2231-2803
http://www.ijcttjournal.org
Page173