WO2010083661A1

WO2010083661A1 - Method and device for improving the management of wireless mesh networks

Info

Publication number: WO2010083661A1
Application number: PCT/CN2009/070323
Authority: WO
Inventors: Huimin Zhang; Wenjun Li; Yuan Zhou; Sebastian Max; Yunpeng Zang; Bernhard Walke
Original assignee: Huawei Technologies Co., Ltd.
Priority date: 2009-01-24
Filing date: 2009-01-24
Publication date: 2010-07-29
Also published as: CN102379135B; CN102379135A

Abstract

The present invention discloses a method for calculating transmission rate for wireless mesh network. The comprising: getting the interference probability distribution according to the traffic load and the current rate of links; getting SINR according to the interference probability distribution; determining the MCS according to the SINR; and determining the transmission rate according to the MCS.

Description

METHOD AND DEVICE FOR IMPROVING THE MANAGEMENT OF WIRELESS

MESH NETWORKS

Field of the Invention

[0001] The present invention relates to the field of Wireless Local Area Networks, and in particular to a method and device for Improving the Management of Wireless Mesh Networks.

Background of the Invention

[0002] The standard IEEE 802.11 for Wireless Local Area Networks (WLANs) was initially designed for small unmanaged networks, consisting of a single Access Point (AP) and one or more Stations (STAs). Recently, they are more and more used to provide wireless Internet access to larger areas, e. g. city centers. As the service area of one AP is limited by high pathloss and low transmission power, multiple APs are deployed to enlarge the service area. To reduce the costs of the wired backbone among the APs, wires are replaced by radio, introducing the Wireless Mesh Network (WMN). In a WMN, Mesh Points (MPs) serve to forward data to or from the nearby AP (possibly across other MPs) multi-hop from or to a STA. From the viewpoint of the STA, an MP appears to be a regular AP; hence, the concept is fully backward-compatible to legacy IEEE 802.11.

[0003] Intelligent network management is a must before and during the operation a large-scale WMNs. As WMNs have a static topology it becomes possible to manage the network, taking into account the requirements of the provider, the users, and the service area. Due to their special multi-hop structure and the characteristics of IEEE 802.11, network.

[0004] Management approaches as known from cellular networks cannot be applied directly to WMNs. The major difficulty towards the management of large-scale IEEE 802.11 networks is the quality estimation of a network, either during the planning or during the operation stage. To be useable, this estimation should have high accuracy, given that the initial parameters (traffic distribution, propagation conditions) are correct. Furthermore, the model should have low complexity to be applicable for the optimization of the network.

[0005] Dynamic models capture all necessary characteristics of IEEE 802.11 by implementing the protocol stack into an event-driven simulation. As several random sources influence the simulation (e. g. frame errors, random backoff, packet arrival times, etc.), statistical sound results are obtained by averaging over long simulation runs. In comparison to a static model, event-driven simulation enables models with high detail. For the optimization procedure during RNP, the long runtime makes this approach infeasible..

[0006] Markov-Chains are a prominent method to convert the dynamic attributes of the IEEE 802.11 Medium Access Control (MAC), e. g. the random backoff, into a static model. By deriving state probabilities, the interaction between the participating nodes and thus successful and failed packet transmissions can be predicted.

[0007] The prior art considers only one MCS per link, which is in contrast to the eight available MCS of IEEE 802.1 la/g; this simplification can only be used in networks without Rate Adaptation (RA) and leads to an underestimation otherwise.

Summary of the Invention

[0008] In this disclosure, we go beyond the existing work in three ways. First, we allow for a set of MCSs (MCS, Modulation- and Coding Scheme) with different susceptibility to interference; the selection of the MCS by the transmitter is encapsulated in a RA model. Hence, the multi-rate ability of IEEE 802.1 la/g with eight different MCS can be included appropriately. Second, the model covers not only single-hop networks, but is applicable to any multi-hop network with static topology, including WMNs. Third, in addition to the saturation throughput our model is able to estimate the occupancy of any node in the network. Therefore, it becomes straightforward to identify the bottleneck of a network, which enables a precise RNP optimization process.

[0009] An embodiment of the invention provides a method for calculating transmission rate for wireless mesh network, comprising: getting the interference probability distribution according to the traffic load and the current rate of links; getting SINR(SINR, Signal to Interference plus Noise Ratio) according to the interference probability distribution; determining the MCS according to the SINR; and determining the transmission rate according to the MCS.

[0010] An embodiment of the invention provides a device for calculating transmission rate for wireless mesh network, the device includes: a MAC Layer unit, a channel model unit, a rate adaptation unit and a Physical model unit. [0011] The MAC unit is adapted to calculate the link interference probability distribution according to the traffic load and the current rate of links. The channel model unit is adapted to calculate the SINR according to the interference probability distribution. The rate adaptation unit is adapted to determine the MCS according to the SINR. And the Physical model unit is adapted to determine the rate of link according to the MCS and SINR. [0012] In the embodiment of the invention, first, we allow for a set of MCSs with different susceptibility to interference; the selection of the MCS by the transmitter is encapsulated in a RA model. Second, the model covers not only single-hop networks, but is applicable to any multi-hop network with static topology. Third, in addition to the saturation throughput, our model is able to estimate the occupancy of any node in the network. Therefore, it becomes straightforward to identify the bottleneck of a network, enabling a precise RNP optimization process. Brief Description of the Drawings

[0013] The drawings required in the descriptions of the embodiments of the invention or the prior art will be introduced briefly in order to explain the technical solutions in the embodiments or the prior art more clearly. Evidently, the drawings described below are merely illustrative of some embodiments of the invention, and those ordinarily skilled in the art can further derive from these drawings other drawings without any inventive effort.

[0014] Figure 1 illustrates a flowchart of an embodiment of the invention which provides a method for calculating transmission rate for wireless mesh network;

[0015] Figure 2 are exemplary network and corresponding interferer tree for one link; [0016] Figure 3 illustrates the curves between effective rate v.s. SINRs under the different MCSs;

[0017] Figure 4 illustrates a chart of an embodiment of a device for for calculating transmission rate for wireless mesh network.

Detailed Description of the Invention [0018] The technical solutions in the embodiments of the invention will be described below clearly and fully with reference to the drawings in the embodiments of the invention. Evidently, the described embodiments are only a part but not exhaustive of embodiments of the invention. Any other embodiments which will occur to those ordinarily skilled in the art in light of the embodiments in the invention here without any inventive effort shall fall within the scope of the invention.

[0019] The major challenge for an IEEE 802.11 -based WMN model is the complex interactions between links. One important property is the effective rate r(i, j) of a link i to j, defined as the mean number of data bits that i is able to transmit successfully to j per second. First, this rate depends on the selected MCS plus the overhead from the channel access procedure of IEEE 802.11, e. g. frame headers and the acknowledgment packet. Second, r(i, j) depends on the interference of surrounding nodes, especially if those nodes are hidden from i. If interference occurs frequently, i may select a more robust MCS, decreasing its effective rate. Therefore, it has to transmit longer for the same amount of data, possibly increasing interference to other nodes. These, in turn, may also decrease their MCS, and so on. Either, this procedure converges at a stable point or some links are blocked completely and stop operating.

[0020] Figl shows an embodiment of the invention which provides a method for calculating transmission rate for wireless mesh network.

[0021] SlOl : Get the interference probability distribution according to the traffic load and the current rate of links. [0022] S 102: Get SINR according to the interference probability distribution.

[0023] S 103 : Determine the MC S according to the SINR.

[0024] S 104: Determine the transmission rate according to the MCS.

[0025] The process from SlOl to S 104 is iterated, and the initial effective rates are selected as the rates under optimal condition, i. e. without interference.

[0026] An embodiment of getting the interference probability distribution is as follows. The Distributed Coordination Function (DCF) in the IEEE 802.11 MAC uses a variant of Carrier Sense Multiple Access/Collision Avoidance (CSMA/CA) to reduce the probability of interfering transmissions. A node with a packet in its transmission queue waits until the channel is idle before starting the transmission attempt. To reduce the probability of a simultaneous access of multiple nodes, a random waiting time is chosen after the channel becomes idle; if the channel stays idle during the countdown, the transmission starts; otherwise the node backoffs and repeats the attempt at the next opportunity.

[0027] A central element in this procedure is the Clear Channel Assessment (CCA) methods that determine the channel condition. Two different techniques are used: First, the Physical

Clear Channel Assessment (P-CCA) relies solely on the received signal energy; if this exceeds a threshold (-82dBm in IEEE 802.1 Ia), the channel is determined to be busy. Second,

Virtual Clear Channel Assessment (V-CCA) uses the frame- and Network Allocation Vector

(NAV)-length transmitted in the PHY and MAC header which indicate the length of the current frame and the current frame exchange, respectively. If a node overhears such information, it defers from the channel access for the indicated duration.

[0028] For any link in the network, the CCA operation controls (i) which sets of nodes are able to transmit simultaneously with this link, creating interference, and (ii) which nodes may block the link by their transmission. The computation of (i) and (ii) together with the corresponding probability is the task of the MAC model.

[0029] The following explanation of the algorithm (given as Algorithm 1) uses the link "Transmitter to Receiver" in the network in Figure 2a as an example. To compute the probabilities, the load of each node must be given; in the example it is indicated as small number next to the node. [0030] Algorithm 1 Static MAC model: computation of interference sets and blocking probability. Initialization: ComputelnterfererSets (tx, Icand, {}, 1.0)

Input: Active transmitter tx, candidate interferers I_cand, active interferers I_active, probability p

Output: II = {(I, pi)|I: Interferer set with probability pi}, Pbusy 1 : if lcand = {} then

No candidate interferers left 2: return {({}, p)} ₃ 0.0

Choose candidate interferer 3: I <— one candidate interferer from I_cancj

4: if tx is blocked by transmission from i then

/ contributes only to pbusy 5 : II , pbusy «— ComputelnterfererSets

(tx, Icand — i, Iactive, p) 6: Pbusy <^~ Pbusy + P ' Ptx(i)

7: return II , pb_USy

Recursion for the case that i is not active 8: II ^~, Pbusy^" <— ComputelnterfererSets

(tx, Icand — i, Iactive, p ^■ (1 — ptx(i)))

i f_!- T new T , ^•

I U. I active * — ^active ^"+^" 1

11 : Icand ^new -*— {j G Icandlj is not blocked by i}

Recursion for the case that i is active 12: I+, pbusy+ <— ComputelnterfererSets

(tX, IcaruT, Iactive™™, P ' Pt_x(O)

13 : Append i to all sets in 1+

14: return II ^{+ 1}J II^", p _busy ⁺ + p_buSy ^~

[0031] In a first step, the nodes which affect the link are stored into the candidate set I_cand; they are characterized by having a load of more than zero and being inside a defined distance from either the transmitter or the receiver, which is set to -105dBm. In the example, I_cand ⁼ {1,

2, 4, 5, 7}.

[0032] Then, Algorithm 1 proceeds in a depth-first search over I_cand to enumerate all subsets of nodes which can be simultaneously transmitting with the selected link, called "interference sets". The corresponding search tree of the example is given in Figure 2b. In each recursion step a candidate node i is picked from I_canu, depending on the node two cases are distinguished: i cannot be transmitting simultaneous to the transmitter and adds to its channel busy fraction

(Line 5). In the example, this holds for nodes 1 and 2; i can transmit simultaneous; this creates two children of the current vertex; one for a passive and one for an active i (Lines 8 and 12, respective). In the case that i is active, the set of candidate interferers is updated to contain only the nodes which are not blocked by i. An example of this case can be found in the example search tree where node 4 is active and blocks node 5.

[0033] A leaf of the tree is found if I_cand is empty; it represents one interference set, containing all nodes selected as active on the way to the leaf. After all leafs of a vertex are found the recursion is finished and the parent collects the different interferer sets in II . The probability of each set is calculated as the product probability of each case in the path during the tree, which is determined by the transmission probabilities of the nodes on the path. In the same way, p_busy of the transmitter is calculated by summing up the transmission probabilities of the blocking nodes .

[0034] So, the steps summarize as follows. Store the nodes information which affects the link, and making a candidate set. Get the interference set from candidate set, wherein the nodes in interference set can transmit data at the same time with transmitter. Get the interference probability distribution according to the active probability. [0035] Figure 2.4b is an example to get the the interference sets and the corresponding probabilities by the depth-first search under the example network in 2.4a. (l)The initial I_cand of the Transmitter and the Receiver is {1, 2, 4, 5, 7} and the initial busy probability p = 1.0; (2) Nodes 1 and 2 are moved from I_cand because the cannot be transmitting simultaneous to the transmitter and new I_cand = {4, 5, 7} with p =1.0; (3)Node 4 can transmit simultaneous and its offered traffic load is 0.2, so the active probability of node 4 is 0.2; (4) Node 5 will be blocked if node 4 is active, so node 5 is moved and the I_cand = {7} when node 4 active and p = 0.2; (5) Node 7 can transmit simultaneous and its offered traffic load is 0.2, so the the active probability of node 7 is 0.2; (6) One interference set {4, 7} is determined, the p = 0.04 by product the active probability of node 4 and node 7. [0036]

[0037] The following is explanation of process for calculating SINR.

[0038] For embodiment of the invention, we consider that either on-site measurements or at least a modeling of the specific propagation conditions in the service area have been performed. Therefore, for a given network configuration of n nodes, the average reception power at node j during transmission from node i is known or can be approximated. In the embodiment, we will denote this power as P(i, j).

[0039] As an important element in the IEEE 802.11 model the wireless channel is used to convert the output of the MAC model (a set II consisting of tuples (I, Pi ), where I is a set of nodes and pi is the associated probability that these nodes transmit simultaneously) into an SINR histogram for a specific link tx to rx. This is done by converting each set in I to the corresponding SINR value: Pi fdWΛ'

SINR_J =

where N represents background and receiver noise. Then, all SINRi with weight pi are sorted into a histogram, approximating the SINR distribution of this link. [0040] The following explanation process for determine the MCS according to the SINR.

[0041] Rate Adaptation (RA) in wireless networks, specifically in IEEE 802.11, deals with the problem that the transmitter has to estimate the optimal MCS for a transmission using limited knowledge of the current SINR condition at the receiver. Especially in the case of an IEEE 802.11- based WMN, hidden nodes and the nondeterministic channel add to the complexity of the problem.

[0042] The SINR distribution determines the selected MCS, using a RA model. In the embodiment, we consider the transmitter can knows via feedback the mean SINR of previous frames. Therefore, its functionality in the static case is reduced to computing the mean SINR from the SINR histogram (generated by the MAC and channel model) and selecting the MCS with the highest expected throughput at this SINR level. It has to be noticed that this expected throughput differs from the effective rate, as this computes the average rate of the selected MCS, given the SINR distribution. This difference reflects the fact that the transmitter is not able to react to fast-changing SINR, but only to changes in the mean.

[0043] Each possible SINR value is calculated in the candidate set to get the mean SINR weighted with the probability of all the said possible SINR value in the candidate set. The mean SINR values determine the MCS, as an embodiment, illustrated by figure3. We can get the relation between SINR and MCS in advance as illustrated by figure 3. The selected MCS has the highest expected throughput at this SINR level

[0044] The following explanation of process for determining the transmission rate according to the MCS.

[0045] While the embodiment is not specific to the characteristics of a given physical layer, we will use IEEE 802.11a, which uses OFDM to provide 8 different MCS in the license-exempt 5.5 GHz band. In reality, a successful frame reception depends on several parameters: Interference from other sources during the transmission must be limited such that all transmitted data bits can be decoded correctly using the applied error-correcting code; the receiver has to be correctly synchronized to the transmission, using the frame preamble; Effects from multiple reflections or Doppler-shift must be still tolerable.

[0046] As an outcome, the physical layer determines the Packet Error Rate (PER), the expected number of retransmissions for a successful reception and the effective rate r (i, j) between two nodes i and j . [0047] In the static model r(i, j) depends on the interference distribution and the selected MCS only. For a single SINR value, the effective rate can easily be computed from the Bit Error Rate (BER) of the selected MCS, the frame length, the number of data bits per symbol of the MCS and the constant IEEE 802.11 overhead, including average backoff, Interframe Spaces (IFSs) and the Acknowledgment (ACK) duration.

[0048] As an embodiment, the rate can be searched in a diagram of relation between SINR and MCS, as in figure3. For all rates, one rate corresponds to one SINR value. The SINR values are all possible SINR values in the candidate set. Then get the mean rate weighted with their probability. [0049] The described models in the previous sections allow for implementing the iterative estimation of the effective rates To compute the saturation throughput of a given network an intermediate step convertes the effective rates to the occupancy, and then finally to the saturation throughput.

[0050] In the embodiment, the dynamic selection and maintenance of paths is simplified to a static weighted graph, which is used as input for the Floyd- Warshall all-pairs shortest path algorithm. To compute the edge weights, the expected rate r(i, j) of each link is required; these rates are computed once for the optimal case without interference.

[0051] Using the output of the Floyd- Warshall algorithm, the offered end-to-end traffic can be converted to the offered traffic per link o(i, j) by going through all selected paths and adding up the corresponding load.

[0052] Similar, the initial effective rates are selected as the rates under optimal condition, i. e. without interference. After the iterative process from Figure 1.2 has converged, the final effective rates r(i, j) and the channel busy fraction pbusy(i) are known. Therefore, the fraction of time blocked by and transmitting to its neighbors can be computed as the occupancy of the node: '

[0053] The network is in saturation if at least one node has an occupancy greater than 1.0, this node is the bottleneck of the network.

[0054] The skilled person in the art will readily appreciate that the present invention may be implemented using either hardware, or software, or both. Embodiments within the scope of the present invention also include computer-readable media for carrying or having computer-executable instructions, computer-readable instructions, or data structures stored thereon. Such computer-readable media can include physical storage media such as RAM,

ROM, other optical disk storage, or magnetic disk storage. The program of instructions stored in the computer-readable media is executed by a machine to perform a method. The method may include the steps of any one of the method embodiments of the present invention.

[0055] The following explanation of a device for calculating transmission rate for wireless mesh network.

[0056] See Fig 4, the device50 includes a MAC Layer unit 501, a channel model unit 503, a rate adaptation unit 505 and a Physical model unit 507.

[0057] The MAC unit 501 is adapted to calculate the link interference probability distribution according to the traffic load and the current rate of links.

[0058] The channel model unit 503 is adapted to calculate the SINR according to the interference probability distribution. [0059] The rate adaptation unit 505 is adapted to determine the MCS according to the SINR value. Calculate every possible SINR in the candidate set. Every possible SINR value is calculated in the candidate set to get the mean SINR from all the said possible SINR value in the candidate set. The mean SINR value determines the MCS, as an embodiment illustrated by figure3. We can get the relation between SINR and MCS in advance as illustrated by figure 3

[0060] And the Physical model unit 507 is adapted to determine the rate of link according to the MCS and SINR.

[0061] The device 50 includes a judge unit 504; adapted to judge the channel occupancy rate when the device is in iterative state and then determine whether to quit the iterative state. [0062] The device 50 includes a path selection unit 502, the path selection unit adapted to get route from mesh point (MP) to access point (AP). The AP can connect to the intent.

[0063] The embodiments both method and device allow for the computational efficient estimation of the saturation throughput of a given WMN, taking into account the complex interplay of different factors. The evaluation shows a reasonable mean relative error of less than 15%, measured in different typical WMN scenarios. As a special feature, our model allows for identifying the bottleneck of the network, which can be used to guide the optimization of the network. This allows for several different applications during the different stages of the RNP process of WMNs.

[0064] The above embodiments are provided for illustration only and the order of the embodiments can not be considered as a criterion for evaluating the embodiments. In addition, the expression "step" in the embodiments does not intend to limit the sequence of the steps for implementing the present invention to the sequence as described herein.

[0065] Additional advantages and modifications will readily occur to those skilled in the art.

Therefore, the invention in its broader aspects is not limited to the specific details and representative embodiments shown and described herein. Accordingly, various modifications and variations may be made without departing from the scope of the invention as defined by the appended claims and their equivalents.

Claims

LA method for calculating transmission rate for wireless mesh network , comprising: getting the interference probability distribution according to the traffic load and the current rate of links; getting SINR according to the interference probability distribution; determining the MCS according to the SINR; determining the transmission rate according to the MCS.

2.The method according to claim 1, further comprising, iterating process from SlOl to S 104, wherein the initial effective rates are selected as the rates under optimal condition.

3. The method according to claim 1, wherein the getting the interference probability distribution according to the traffic load and the current rate of links comprises: storing the nodes information which affects the link, and making a candidate set; getting the interference set from candidate set, wherein the nodes in interference set can transmit data at the same time with transmitter; getting the interference probability distribution according to the active probability.

4.The method according to claim 1, wherein the getting SINR according to the interference probability distribution comprises: calculating every SINR values in the candidate set; getting the mean SINR from all the said SINR values in the candidate set.

5. The method according to claim 1, wherein the determining the transmission rate according to the MCS comprises: searching rate in a diagram of relation between SINR and MCS; getting all rates, wherein one rate corresponds to one SINR values; getting the mean rate weighted with the said rate.

6.A device for calculating transmission rate for wireless mesh network, comprising a MAC Layer unit, a channel model unit, a rate adaptation unit and a physical model unit; the MAC unit, adapted to calculate the link interference probability distribution according to the traffic load and the current rate of links; the channel model unit, adapted to calculate the SINR value according to the interference probability distribution; the rate adaptation unit, adapted to determine the MCS according to the SINR value; the physical model unit, adapted to determine the rate of link according to the MCS and SINR.

7.The device according to claim 6, further comprising: a judge unit, adapted to judge the channel occupancy rate when the device is in iterative state and then determine whether to quit the iterative state.

8. The device according to claim 6, further comprising: a path selection unit adapted to get route from mesh point to access point.