AU2014100684A4 - Method for scheduling a random-access communication system with absolute and proportional throughput guarantees - Google Patents

Abstract

A method for scheduling a random-access communication system having high priority (HP) and low-priority (LP) nodes is provided, where the scheduling is configured such that the system provides an absolute throughput guarantee for the HP nodes, and a proportional throughput guarantee for the LP nodes. The method is based on obtaining a length of a contention window assigned to an individual node, which is either a HP or a LP node, from a per-slot attempt rate of this individual node. In particular, the attempt rate of each individual node is determined by an algorithm configured such that a resultant sum of the per-slot attempt rates of all the individual nodes is a fixed value independent of the number of the individual nodes. By this scheduling method, the maximum system throughput is approached. The method can be advantageously used in a wireless local area network (WLAN). Abstract figure: FIG. 4. Page 18 - I CL, C/, 0 LP C,,) I 0) Z 0 00 0)0 0) a) - a) 0 a) 00 ELao0 C:E(D ( c 0 m 0o cz 0 0- a

Description

Method for Scheduling a Random-Access Communication System with Absolute and Proportional Throughput Guarantees Inventors: Qinglin ZHAO and Zhijie MA FIELD OF THE INVENTION [00011 The present invention relates to scheduling a random-access communication system for providing absolute throughput guarantee to one or more high-priority nodes and proportional throughput guarantee to one or more low-priority nodes. BACKGROUND [00021 There follows a list of references that are occasionally cited in the specification. Each of the disclosures of these references is incorporated by reference herein in its entirety. List of references [1] 0. Bejarano, E. W. Knightly, and M. Park, "IEEE 802.1lac: From channelization to multi-user MIMO," IEEE Communications Magazine, vol. 51, pp. 84-90, 2013. [2] E. Perahia, C. Cordeiro, M. Park, and L. L. Yang, "IEEE 802.1 lad: Defining the next generation multi-gbps Wi-Fi," in Consumer Communications and Networking Conference (CCNC), pp. 1-5, 2010. [3] D. Yoon, S. Lee, J. Hong, and K. Chung, "Weighted bandwidth sharing scheme to guarantee the video quality in home networks," in International Conference on Information Networking (ICOIN), IEEE, pp. 423-427, 28-30 Jan. 2013. [4] A. Banchs, X. Perez, and D. Qiao, "Providing Throughput Guarantees in IEEE 802.11 e Wireless LANs," Proceedings of the 18th International Teletraffic Congress (ITC-18), pp. 1001-1010, 2003. [5] M. Nassiri, M. Heusse, and A. Duda, "A Novel Access Method for Supporting Absolute and Proportional in 802.11 WLANs," Proceedings of2008 IEEE Conference on Computer Communications (INFOCOM), pp. 709-717, 2008. [6] G. Bianchi, "Performance analysis of the IEEE 802.11 distributed coordination function," IEEE Journal on Selected Areas in Communications, vol. 18, no. 3, pp. 535 547, Mar. 2000. Page 1 [7] Q. L. Zhao, D. H. K. Tsang, and T. Sakurai, "A novel CAC scheme for homogeneous 802.11 networks," IEEE Transactions on Wireless Communications, vol. 9, no. 3, pp. 1168-1174, 2010. [8] S. Wiethoelter, M. Emmelmann, C. Hoene, and A. Wolisz, "TKN EDCA Model for NS2," Technische Universitat Berlin, Tech. Rep. TKN-06-003, Jun. 2006. [9] "The Network Simulator: Building Ns" (http://www.isi.edu/nsnam/ns/ns-build.html). [10] "Intermediate value theorem" (http://en.wikipedia.org/wiki/Intermediatevaluetheorem). [00031 From WiFi to the latest IEEE 802.11 ac [1] and 802.11 ad [2], wireless local area networks (WLANs) have already been widely deployed in thousands of families. In a typical scenario of wireless home networks, some people watch online high-definition digital TV (HDTV), while the others desire to browse online hot news. Watching online HDTV consumes a huge bandwidth, which might cause a significant delay in accessing the Internet. This situation calls for a reasonable bandwidth allocation to provide absolute throughput needed for inelastic traffic (such as online TV) and relative throughput for elastic traffic (such as text message), at the same time to maximize the system utilization. Consider a random access communication system with one high-priority (HP) class and one low-priority (LP) class. It is desired to schedule the system by setting different contention windows (CWs) for nodes of different classes so as to achieve the following three objectives simultaneously: * providing an absolute (or a fixed) throughput guarantee for the HP nodes; e providing a proportional throughput guarantee for LP nodes, where all LP nodes share the available bandwidth according to the desired ratios; and * maximizing the system throughput. [0004] In the art, only subsets of these three objectives are achieved by prior-art scheduling techniques. For example, [3] only considered providing weighted bandwidth allocation to video streams; [4] just provided absolute throughput guarantee, not considering proportional throughput guarantee; and [5] focused on supporting absolute and proportional priorities, rather than absolute and proportional throughput guarantees. There is a need in the art to have a technique for scheduling a random-access communication system such that the three aforementioned objectives are achieved. Page 2 SUMMARY OF THE INVENTION [00051 An aspect of the present invention is to provide a scheduling method for achieving the aforementioned three objectives. In particular, the method is for scheduling a random access communication system having one or more HP nodes, one or more LP nodes and a hub. The one or more HP nodes and the one or more LP nodes contend for sending data packets to the hub. Each individual HP node has a pre-determined throughput value for the system to provide by the scheduling. The scheduling is configured such that the system provides an absolute throughput guarantee for the one or more HP nodes, and a proportional throughput guarantee for the one or more LP nodes. The method comprises obtaining a length of a contention window assigned to an individual node, which is either a HP or a LP node, from a per-slot attempt rate of this individual node. In addition, the method further comprises determining the attempt rate of each individual node by an algorithm configured such that a resultant sum of the per-slot attempt rates of all the individual nodes is a fixed value independent of the number of the individual nodes. [0006] The random-access communication system may be a WLAN so that the hub is an access point of the WLAN. The WLAN may be an IEEE 802.11-compliant WLAN. [00071 Other aspects of the present invention are disclosed as illustrated by the embodiments hereinafter. BRIEF DESCRIPTION OF THE DRAWINGS [0008] FIG. 1 is a plot obtained from a simulation experiment for illustrating that the absolute throughput guarantee is achievable by the present invention, the plot showing a per node throughput for a HP class versus the number of LP nodes where the HP class has n = 2 nodes with the following fixed throughput requirements: a, = 0.24Mbps and a 2 = 0.48Mbps. [00091 FIG. 2 is a plot obtained from the simulation experiment for illustrating that the proportional throughput guarantee is achievable by the present invention, the plot showing a per-node throughput for a LP class versus the number of LP nodes where the LP class has m nodes, and the required throughput ratio between the first half of and the second half of m nodes is 1:2. [00101 FIG. 3 is a plot obtained from the simulation experiment for illustrating that the system throughput is maximized by the present invention, the plot showing the system throughput versus the number of the LP nodes, where n = 2 and m = 2, 3, ... , 18. Page 3 [0011] FIG. 4 depicts a flowchart illustrating the determination of contention windows for nodes of different classes according to an exemplary embodiment of the present invention. DETAILED DESCRIPTION [00121 In the present invention, a random-access communication system with one HP class and one LP class is considered. By setting different contention windows (CWs), it is desired to achieve the aforementioned three objectives simultaneously, namely: providing an absolute throughput guarantee for HP nodes; providing a proportional throughput guarantee for LP nodes; and maximizing a system throughput. An aspect of the present invention is a scheduling method for achieving these objectives by determining CW lengths for the HP nodes and the LP nodes. [0013] The present invention is exemplarily illustrated by considering a WLAN as an example to the random-access communication system. The exact and approximate solutions to the optimal CWs are presented. These solutions are obtained based on the IEEE 802.11 DCF protocol, which is the most practical, robust, and popular WLAN protocol. The existence and uniqueness of the approximate solution are proved. Simulation experiments done by the Inventors have verified that the approximate solution can well achieve the three objectives. For example, the mean relative errors from simulation are 1.14%, 0.9 1%, and 1.18% for the targeted fixed-throughput, proportional throughput ratio, and system throughput, respectively. [0014] The Inventors have observed that in existing techniques such as [3]-[5], supporting absolute priority is such that absolute-priority nodes are benefited from all available throughput even if some non-absolute-priority nodes already exist, consequently lowering the bandwidth utilization because the absolute-priority nodes may acquire more bandwidth than is required. In the present invention, on the other hand, supporting absolute throughput guarantee is realized by that the HP nodes are guaranteed to acquire the bandwidth for providing the required throughput only, rather than to occupy more bandwidth than is necessary. [00151 For simplicity in presentation, herein in the specification and the appended claims, the terms "node i " and "i th node" are used interchangeably with the same meaning. A. Model Assumptions and Problem Formulation [00161 We consider a one-hop WLAN with two classes: HP class and LP class. The HP class has n nodes and the LP class has m nodes so that the total number of nodes is N = n Page 4 + m. Each node i, 1 < i < N, always generates a random back-off count uniformly distributed in [0, CW, -1] for each new transmission or retransmission, where CWJ > 1. All nodes send data to an access point (AP), where the AP only acts as a receiver. It is assumed that: 1) all nodes hear each other and run in the basic mode; 2) there are ideal channel conditions (that is, the transmission errors are a result of packet collision only); and 3) all nodes are in saturated operation (that is, each node always has packets to transmit) and have the same packet size L. [0017] The problem is formulated as follows. Let p A (p,#82, '', pN), where p, 1 i N, represents the attempt rate of node i per slot (namely, the mean number that node i attempts to transmit a packet in a slot). In the saturated operation, p, is calculated from CW, by [6] A= 2 2. (1) CW +1 It follows that finding the optimal (CW 1 , CW 2 ,. -, CWN) is equivalent to finding the optimal p8. [00181 Let F, A F(,), 1 i N , be the throughput of node i under p8. Let aj, 1 i N , represent the fixed throughput required by each HP node i. Let ri , 1 i m , represent the proportional throughput ratio between the LP nodes n + i and n + 1. In the considered WLAN, it is desired to find the optimal p3 so as to guarantee each HP node's fixed throughput requirement a and each LP node's proportional throughput requirement rj, and at the same time maximize the system throughput = F, . That is, N the optimal / = arg max T, (p8) (2) # i= over Fj,)= aj, 1:! i:! n , (3) l~imr,, 1: i m. (4) F,1 (p3) In the next section, we express the throughput F, and find the optimal p3. B. The Optimal Attempt Rate [0019] This section first expresses the per-node throughput F,, and then calculates the exact and approximate solutions to the optimal p3. Page 5 B.1 Per-Node Throughput [00201 Let P' be the probability of a successful transmission of node i, 1 < i i N. Since node i can successfully transmit a packet only if all other nodes do not transmit, we have N P, = p jj(1 - p), 1 i<!N. (5) j#i [0021] Let P and P, be the idle probability and collision probability of a slot, respectively. We have N Pe= (l1 -A) (6) and N i=1 [0022] Let Q be the mean time that elapses for one decrement of the back-off counter. Note that the back-off counter decreases by one for each idle slot and is suspended when the channel is busy. Then Q can be calculated by N Q = orP + T P + TP .(8) In (8): o is the duration of one time slot; and T, (T,) >> o- is the mean time of a successful (an unsuccessful) transmission, where T, is given in Table 1. We further assume that T, = T to simplify analysis. This assumption has widely been used in previous work such as [7], and can be removed easily. Under this assumption, (8) reduces to Q = (o- - T)P + T,. [0023] The throughput of node i, Fi, 1 < i < N, is defined as the number of bits that node i successfully transmits in a time duration of Q. Then we have F.= LP' (9) (--T )P + T where L denotes the packet size. B.2 Exact Solution to the Optimal / [0024] The exact solution to the optimal 8 can be found in five steps below. [00251 Step 1: Express #,,,, 1 < i m, in terms of #,1. From (4), (5) and (9), we have Page 6 r. ,1- = fl"+ G " . (10) i,1 p1+ (1 - p9,,)( Then p,,i can be expressed in terms of p8,, , namely p+, 1im. (11) 1 - p +± + , [0026] Step 2: Express pA, 1 ! i ! n, in terms ofpln, 1 . Note that F,/F 1 = ai/a, and regard a,/a, as the throughput ratio between nodes i and 1. From (11), we have , = , a , 1 s i ! n. (12) 1- pi+ 'A al [00271 Step 3: Setup a relationship between p, and pA . After substituting (5) and (6) into (9), we rewrite FT = a, as Lp 1 x 1-p)(1-A) 1-p) (13) (aT ~X I (1-,) X ",(1-A,8+) + T Further, substituting (11) and (12) into (13), we obtain an implicit relationship between p,, and p . [00281 Step 4: Express E in terms of 8, and 1A. With (3) and (4), the system throughput _N F, is written as N n 1 1 jT,= la, + T,,r (14) 1=1 i=1 i=1 =_a,+1"* A (15) j=1 A 1 n,+1 i=1 where from (14) to (15), we use the expression of F, 1 /T , which can be obtained according to (5) and (9). [0029] Step 5: Find the optimal p . We first search all pairs of p, and pA that satisfy (13), then choose their optimal values maximizing (15), and finally calculate other ,'s by (11) and (12). Page 7 [0030] In general, it is not easy to find the exact solution to the optimal p because there are infinite pairs of 8, and /. In addition, we even do not know whether the exact solution exits and is unique. We therefore seek the approximate solution in the next subsection. C. Approximate Solution to the Optimal#f3 [00311 In order to find the approximate solution to the optimal #, we adopt a key approximation, pA << 1, which is widely used in the related literature such as [5]. The approximation holds true since p, represents the per-node attempt rate in a very short slot and therefore it is generally much small. [00321 In the following, we first consider a simple case of n = 1 and m > 1, and then utilize the derived result to find the approximate solution for the general case of n 1 and m 1. C.1 The Case of n = 1 and m 1 [0033] With the approximation , << 1, we can calculate the per-node attempt rate p, by following the steps as in Section B.2. The derived result is summarized in Theorem 1 below. Theorem 1 states that the approximate solution must exist and be unique, and provides a method to compute it quickly and easily. [0034] Theorem 1: When n = 1 and m > 1, with the approximation that , << 1, we have the following results. [00351 (a) The optimal solution p, to (2)-(4) is uniquely given by B +Tee" A~j , = p,= r... (16) y , z = 2 3 , -- N , where: A = L/a 1 +U-T ; B = -- T;and 9 = p8 2 1".' r is the total attempt rate of LP nodes and it is the only solution to h(9) = 0 in which h(p) A B + Te (1- p) and 0 < p < 1. Page 8 [00361 (b) Given the packet size L and the fixed HP throughput a,, the total attempt rate of all nodes and the maximum system throughput are constants, independent of the total node number and the proportional LP throughput ratios. [00371 Proof Please refer to Section C.3. [0038] From Ah 1 and the expression of pA given by (16), we can obtain the upper bound of al , namely, a 1 L/(Te''). C.2 The Case of n I and m 21 [00391 Theorem 1(b) states that for the given L and a 1 , the total attempt rate of all nodes is independent of the total node number. This reveals the idea of computing the approximate solution when n > 1, as explained below. [0040] First, compute HP node l's attempt rate pA and the total attempt rate Y§= p, by regarding all HP nodes except node 1 as the LP nodes, according to (16) in Theorem 1(a) and the parameter values of L and a 1 . [00411 Next, compute HP node i's attempt rate, pA, 1 i < n , according to (19) shown in Section C.3 and the throughput ratio ai/a between node i and node 1. p,= a'p, 1 ! i ! n . (17) a, [00421 Finally, compute LP node n +i 's attempt rate, p,, 1 < i m, by allocating all remaining attempt rates to the LP nodes according to their ratios, rid's: p3 = (Z p pZy), 15ism. (18) Zir. i,+ 1 = I E- =,j,(8 C.3 Appendix: Proof of Theorem (1) [0043] Theorem 1(a) is proved by two steps below. [0044] Step 1: Express p, by (16). When n = 1, (10) reduces to ri ~1 /p62. Since p, << 1, it follows that p+, = =r 2 = P , 1! i ! m . (19) Substituting (19) into (13), we get an explicit relationship between #A and #2: Page 9 B+ 'e pi=- ~=j 82 B + Te (20) A A where we apply the approximations: A << 1; and (1 - x) ~ e'' for x << y . Combining (19) and (20), we obtain (16). [0045] Step 2: Prove that p exists and is unique. First, we can express Y_ E, in terms of p . According to (19), (15) reduces to i1 , = al+ aA B + Te .(21) Second, setting the first-order derivative of E= F, with respect to p to 0, we obtain h(P) = 0. Third, the solution to h(9) = 0 must exist uniquely from the intermediate value theorem [10], because (i) h ((p) is continuous and decreasing over the range (0,+o) as a result of h'(p) = - poTe < 0; (ii) h(0) = a > 0; and (iii) h(1) = B < 0. [0046] We next explain Theorem 1(b). From (16), the total attempt rate of all nodes is given by A +B+Tee+P. (22) For the given packet size L and the HP throughput ai, A, B and T are constants while P is the unique solution to h(9) = 0. Consequently, the total attempt rate Ei p, in (22) and the maximum system throughput _E=, F, in (21) are constants, irrespective of what the total node number and the proportional LP throughput ratios are. D. Model Verification [0047] In this section, we demonstrate the efficiency of the disclosed scheduling scheme for WLANs. We use the 802.11 e EDCA simulator [8] in ns2 version 2.28 [9] as a validation tool. In the simulation, we differentiate CW parameters only, so we set AIFS = DIFS, TXOP = 0, and CW = CW. = CW; for node i. The other protocol parameter values are listed in Table 1 and are set according to the IEEE 802.1 lb specification. Each simulation run lasts for 200 seconds. Table 1. Parameters for IEEE 802.1 lb. Page 10 m/M 5/7 Header 220 ps = Mheader + Pheader o 20 ps T, = Header + L + SWS+ ACK + DIFS SIFS 10 ps L, = L bytes@ Raa DIFS 50 ps ACK 304 ps = 24 bytes @ Rbase + 14 bytes @ Rbade S I1 Mbps Mheader 20 ps = 24 bytes @ Rd,, + 4 bytes @ Raa, lRba.i 1 Mbps Pheader 208 ps = 26 bytes @ R a [00481 In our experiment, the HP class has n = 2 nodes with the fixed throughput requirements: ai = 0.24Mbps and a2 = 0.48Mbps. The LP class has m nodes with the proportional throughput ratios: r, = 1 for 1 i m/2, and r, = 2 for m/2 + 1 i m, where m = 2, 4, ..., 18. Each node has the same packet size, L = 1500 bytes. We set CWJ = [2/pA -1] by (1) in simulation, where [-] denotes a rounding function. The exact solution to p, is calculated by (11) and (12) in Section B.2. The approximate solution to , is calculated by (17) and (18) in Section B.3. Table 2 shows the obtained values of CW,'s. Table 2. Exact and approximate solutions to CW and system throughput Exact solution [Eqns. (11) and (12)] Approximate solution [Eqns. (17) and (18)] m System System (n=2) CWi CW 2 CWmi CWm2 throughput CWi CW 2 CWmi CWm2 throughput (Mbps) (Mbps) 2 265 133 35 18 6.3889 337 168 43 21 6.1912 4 313 157 82 42 6.2648 337 168 88 43 6.1912 6 327 164 129 65 6.2279 337 168 132 66 6.1912 8 333 167 175 88 6.2102 337 168 177 88 6.1912 10 338 170 222 112 6.1997 337 168 221 110 6.1912 12 341 171 269 135 6.1928 337 168 266 132 6.1912 14 344 173 317 159 6.1879 337 168 310 155 6.1912 16 344 173 363 182 6.1843 337 168 355 177 6.1912 18 345 173 409 205 6.1815 337 168 399 199 6.1912 Note. ml e={3, ., 2+m/2} and m2 e {3+m/2, ...,2+m}. Page 11 [0049] We now explain that the derived CW 's can guarantee the fixed throughput requirement of the HP nodes (as shown in FIG. 1), guarantee the proportional throughput ratios of the LP nodes (as shown in FIG. 2), and maximize the system throughput (as shown in FIG. 3). In FIGS. 1 to 3, the labels "anaexact" and "ana appx" denote the theoretical exact throughput (15) and the approximate throughput (21), respectively; the labels "sim exact" and "sim appx" denote the simulation results based on the exact and the approximate solutions, respectively. Define the relative error to be the ratio of Simulation value - target value$j to target-value. We use the mean relative error and the maximum relative error to evaluate the accuracy of the proposed model. [00501 FIG. 1 plots the per-node throughput of HP class versus the number of LP nodes (note that the y-axis is not equally divided). There are 2 nodes in the HP class. The fixed throughput requirements of node 1 and node 2 are a, = 0.24Mbps and a2 = 0.48Mbps, respectively. From FIG. 1, we can observe that the sim appx simulation curves closely match the corresponding sim exact simulation curves, respectively; and we can calculate that for the target value of either a, = 0.24Mbps or a 2 = 0.48Mbps, the mean relative error and the maximum relative error are 1.14% and 2.33%, respectively. This manifests that the proposed scheme can guarantee the fixed throughput requirements of the HP class. In addition, FIG. 1 compares the throughput of the absolute-priority nodes in [5] with the throughput of the HP nodes obtained by the present invention. The labels "IS-nodel" and "IS-node2" denote the throughputs of the two absolute-priority nodes. From FIG. 1, we can see that 1) two absolute-priority nodes obtain almost the same throughput; and 2) the obtained throughput is more than is required (i.e. a, = 0.24Mbps or a 2 = 0.48Mbps). The reason is that 1) in [5], the absolute-priority nodes are always benefited from all available throughput even if some non-absolute-priority nodes exist; and 2) [5] did not differentiate the service for the absolute-priority nodes. [0051] FIG. 2 plots the per-node throughput of the LP class versus the number of the LP nodes. Since the absolute-priority nodes in [5] has shared almost all available throughput, then we just discuss the throughput of the LP nodes in this work. There are m nodes in the LP class, where m varies from 2 to 18. The required throughput ratio between the first half of and the second half of m nodes is 1:2. From FIG. 2, we can see that 1) the sim appx simulation curves almost overlap with the corresponding sim exact simulation curves, respectively, and 2) with the number of LP nodes increasing, the simulated throughput ratio Page 12 between the first half of and the second half of m nodes is still about 1:2. For the target value of 1:2, the mean relative error and the maximum relative error are 0.91% and 2.23%, respectively. This manifests that the disclosed scheduling scheme can guarantee the proportional throughput requirements of the LP class. [0052] FIG. 3 plots the system throughput versus the number of the LP nodes. From FIG. 3, we can see that 1) the simexact and simappx simulation curves closely match the ana exact theoretical curve which plots the maximum system throughput limit, where the difference between ana exact and simexact is due to the rounding errors of CW, and 2) the ana_appx theoretical curve matches the anaexact theoretical curve well with the nodes number increasing. For the targeted approximation solution (21), the mean relative error and the maximum relative error are 1.18% and 2.06%, respectively. This manifests that the proposed throughput guarantee scheme can maximize the system throughput. In addition, the simulation shows that the maximum system throughput is a quasi-constant regardless of how the node number varies, which has been approximately predicted in Theorem 1(b). E. The Present Invention [0053] The results obtained in Sections B and C are used in the development of the present invention. [0054] According to the present invention, a method for scheduling a random-access communication system having one or more HP nodes, one or more LP nodes and a hub is provided. The one or more HP nodes and the one or more LP nodes contend for sending data packets to the hub. Each individual HP node has a pre-determined throughput value for the system to provide by the scheduling, and the scheduling is configured such that the system provides an absolute throughput guarantee for the one or more HP nodes, and a proportional throughput guarantee for the one or more LP nodes. [0055] The random-access communication system may be a WLAN, but the present invention is not limited only to a WLAN. In case of a WLAN, the hub is an AP. The WLAN may be an IEEE 802.11-compliant WLAN. As is used in the IEEE 802.11-compliant WLAN, the contention method is based on a carrier sense multiple access with collision avoidance (CSMA/CA) protocol. [00561 Exemplarily, the method comprises obtaining a length of a CW assigned to an individual node, which is either a HP or a LP node, from a per-slot attempt rate of this individual node. The CW length is computed from the per-slot attempt rate by (1). In addition, the method further comprises determining the attempt rate of each individual node Page 13 by an algorithm configured such that a resultant sum of the per-slot attempt rates of all the individual nodes is a fixed value independent of the number of the individual nodes. The last feature is based on Theorem 1(b) above. As a result, the algorithm generates 8', the per-slot attempt rate of one of the one or more HP nodes, by B+Tfe '= ' (23) and CW', the length of the CW assigned to said one of the one or more HP nodes, by CW' = [2/p' -1], (24) where: A' = L/a'+ a - T in which a' is the pre-determined throughput value of said one of the one or more HP nodes; and[z] is a rounding function for rounding the value of z to an integer value. Numerical results of Section D demonstrate that the CW lengths computed by the disclosed method lead to a system throughput that approaches the maximum one. [00571 FIG. 4 is a flowchart showing the steps carried out in the method according to an exemplary embodiment of the present invention. A per-slot attempt rate of a first HP node is first determined in a step 410 by an expression given in (16). If there are more than one HP nodes in the random-access communication system, the attempt rates of the remaining HP nodes are computed in a step 420 according to (17). After the attempt rate(s) for all the HP node(s) are obtained, the attempt rates of the LP nodes are determined in a step 430 by (16) if there is only one HP node in the system, and by (18) after substituting (22) thereinto if there are more-than-one HP nodes. Finally, the CW lengths of all the HP and the LP nodes are computed from the attempt rates by (1) in a step 440. [0058] The method disclosed herein can be implemented in an AP having a function of scheduling a WLAN. The AP comprises one or more processors configured to execute a process for scheduling the WLAN to provide an absolute throughput guarantee for one or more HP nodes and a proportional throughput guarantee for one or more LP nodes by a method according to an embodiment of the present invention. [0059] The embodiments disclosed herein may be implemented using general purpose or specialized computing devices, computer processors, or electronic circuitries including but not limited to digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA), and other programmable logic devices configured or programmed according to the teachings of the present disclosure. Computer instructions or software codes running in the general purpose or specialized computing devices, computer Page 14 processors, or programmable logic devices can readily be prepared by practitioners skilled in the software or electronic art based on the teachings of the present disclosure. [0060] The present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The present embodiment is therefore to be considered in all respects as illustrative and not restrictive. The scope of the invention is indicated by the appended claims rather than by the foregoing description, and all changes that come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Page 15

Claims (5)

1. A method for scheduling a random-access communication system having one or more high-priority (HP) nodes, one or more low-priority (LP) nodes and a hub, the method comprising: obtaining a length of a contention window assigned to an individual node, which is either one of the HP nodes or one of the LP nodes, from a per-slot attempt rate of the individual node; and determining the attempt rate of the individual node by an algorithm configured such that a resultant sum of the per-slot attempt rates of all of the individual nodes is a fixed value independent of the number of the individual nodes; wherein the one or more HP nodes and the one or more LP nodes contend for sending data packets to the hub; wherein each of the HP nodes has a pre-determined throughput value for the system provided by the scheduling; and wherein the scheduling is configured such that the system provides an absolute throughput guarantee for the one or more HP nodes, and a proportional throughput guarantee for the one or more LP nodes.

2. A method for scheduling a random-access communication system having a hub and N nodes contending for sending data packets to the hub, the N nodes consisting of n high-priority (HP) nodes and m (LP) nodes where n > 1, m > 1 and N = n + m, the N nodes being arranged such that an i th node is a HP node for n > i > 1 and is a LP node for N > i > n + 1, the scheduling being configured such that the system provides an absolute throughput guarantee for the n HP nodes and a proportional throughput guarantee for the m LP nodes, the method comprising: for N > i > 1, computing CW, a length of a contention window assigned to the i th node, from pi,, a per-slot attempt rate of the i th node, by CW =[2/pB - 1); computing p, by p-'1= A; A when n > 1, computing pi for n > i > 2 by Page 16 #=- #; and al computing 8,, for m > i > 1 by r. _ n, n=1 ,n'r ( YEj=2p ), n >1; - j = 1 wherein: [.] is a rounding function; ai, n > i > 1, is a pre-determined throughput value for the system to provide to the i th node by the scheduling; r, m > i > 1, is a pre-determined proportional throughput ratio between the (n + i)th node and the (n + 1)th node; A = L/a, +o--T; B = -- T; p e (0,1) is a solution to h( 9 ) = 0 in which h((9) = B + Tee(1-(); L is a packet size adopted in the system; o- is the duration of one time slot; and T is the mean time of an unsuccessful transmission.

3. The method of claim 1 or 2, wherein the random-access communication system is a wireless local area network (WLAN) and the hub is an access point of the WLAN.

4. The method of claim 3, wherein the WLAN is an IEEE 802.11-compliant WLAN.

5. An access point having a function of scheduling a wireless local area network (WLAN), comprising one or more processors configured to execute a process for scheduling the WLAN to provide an absolute throughput guarantee for one or more HP nodes and a proportional throughput guarantee for one or more LP nodes by the method of claim 3. Page 17