WO2012077010A1

WO2012077010A1 - Method of determining a length of a reservation interval in a mesh network, node and network therefor

Info

Publication number: WO2012077010A1
Application number: PCT/IB2011/055294
Authority: WO
Inventors: Theodorus Jacobus Johannes Denteneer
Original assignee: Koninklijke Philips Electronics N.V.
Priority date: 2010-12-07
Filing date: 2011-11-25
Publication date: 2012-06-14

Abstract

The present invention relates to a method for determining a length of a reservation interval in a node of a wireless mesh network, the method comprising the following steps : - defining a maximum length M of the reservation interval, - determining the prime factorization of M, as (I), - determining a subset of intervals length as (II), - picking one element of the subset. The invention also relate to a node and a mesh network.

Description

METHOD OF DETERMINING A LENGTH OF A RESERVATION INTERVAL IN A MESH NETWORK, NODE AND NETWORK THEREFOR.

FIELD OF THE INVENTION

The present invention relates to wireless networks, and more particularly to wireless mesh networks.

This invention relates, more particularly, to a method for determining a length of a reservation interval in a mesh network. BACKGROUND OF THE INVENTION

Wireless mesh networks use multi-hop routes to establish communication between the nodes in the network as well as to applications residing in the Internet. Thus, they reduce the use of costly, wired infrastructure.

Current 802.11 (WiFi) networks (as described in Reference 1) are based upon a centralized architecture. In this standard, some central infrastructure must be available in the form of an access point (AP) and stations can associate with an AP that is located in their vicinity. Stations communicate via such an AP. Thus, stations in these Infrastructure Networks are dependent on an infrastructure that must be available.

Mesh networks, however, are based upon a decentralized architecture. In this architecture, stations can associate and communicate without the availability of any (centralized) infrastructure. This enables various usage scenarios that are not possible with the centralized technology.

In an IEEE 802.11s mesh network, all mesh stations beacon at regular intervals. The beacon carries important information for link setup and maintenance.

However, the beacon is also very important for the power save mechanism. In particular, if a mesh station is in light sleep mode with respect to another peer mesh station, it will wake up for its own beacons, as well as the beacons of this peer. If a mesh station is in deep sleep mode, it will only wake up for its own DTIM beacons. The exchange of packets between peer mesh stations that are in sleep mode with respect to each other are scheduled at intervals following these wake up moments.

The length of the beacon interval, in combination with the choice of the sleep mode, thus determines how often a mesh station in power save has to wake up. Consequently, this length is an essential parameter to regulate the power consumption of a mesh station.

A standard or protocol for mesh networking should therefore provide sufficient flexibility for mesh stations to select an appropriate beacon interval length. One of the main causes of the lack of scalability in current mesh networks is the "flow in the middle" problem.. A node in the center (middle) of a mesh network has many neighbouring nodes with which to compete for transmissions on the shared wireless medium. If the access to the medium is arranged via the standard CSMA/CA protocol, then such a node in the middle will have fewer opportunities to transmit. This can, with increasing network load, starve some of the flows in the network

PROBLEMS OF THE PRIOR ART AND DESCRIPTION OF THE INVENTION

Reservation protocols are introduced in mesh networking protocols to solve this flow in the middle problem E.g. in IEEE802.11s. a reservation protocol, termed MCCA, has been included in the specification. In such a reservation protocol, station reserve time slots for transmissions to neighbouring stations. Such new reservations are first verified against existing reservations and then signalled to all neighbouring stations. Thus, stations can transmit in reserved time slots.

Often, these reservations are specified relative to some reservation interval. First, assume, for convenience of exposition, a slotted time medium. Transmissions take place in the individual time slots; these slots are grouped into frames, and stations select a reservation interval that is expressed in number of frames.

Figure 1, highlights the structure of such a frame, which is composed of 10 time slots. In figure 2, the reservation interval of a station is shown. It is composed of three frames, within each frame, a number of time slots are selected and used for reserved transmissions to another station. Thus, the reservation schedule is specified relative to the reservation interval. Within frame 1, there are no reserved time slots; within frame 2, there is one reserved time slot, and within frame three, there are two reserved time slots.

This reservation interval (with a reservation schedule), translates into a transmission schedule, where the reserved slots are used for transmissions. In figure 3, the transmissions are shown in a succession of time slots. The figure displays a succession of time slots comprising two superframes each corresponding to a reservation interval. The location of the reserved time slots is fixed within each superframe and corresponds to the schedule that is specified relative to the reservation interval. Thus, in Figure 3, there are two superframes, and the reservation schedule specified in Figure 2 is repeated. In this figure, the black slots correspond to transmissions in reserved slots, the grey slots correspond to other transmissions. In the MCCA reservation protocol in IEEE 801.11s (2), this reservation interval overlaps with the DTIM interval, which is a multiple of the beacon interval. The situation is illustrated in Figure 4. Here, the reservations are within a DTIM interval, and repeated in successive DTIM intervals.

In order to make (or accept) a reservation, a mesh node has to verify that the specific reservation does not overlap with any existing reservation. In a mesh network, this means that

1. a new reservation shall not overlap with an existing reservation of a neighbour

2. a new reservation shall not overlap with an existing reservation of a neighbour's neighbour

Thus, a mesh station has to do a comparison with the reservations of (possibly many) other mesh stations.

If stations share a reservation interval, the verification procedure easy. To verify that a reservation made in a time slot in Frame 1, station A need only check the corresponding time slot of frame 1 of station B. This is the same for station B.

In Figure 5, the superframes of stations A and B have been shown aligned to start at the same instant. However, for the verification, the alignment is not essential. When the starts of the superframes are shifted relative to each other, the individual time slots shift by the same amount, so that the verification must be done against a time slot that is shifted by the same amount. This simple verification procedure is possible because station A and B have selected reservation intervals that have the same length.

Below, however, we show that the verification of new reservations may be difficult in case the stations do not have reservation intervals of a common length. In particular, in this case,

1. the amount of checks to be carried out by a station will increase, and

2. the probability of finding a free time slot for a new reservation will decrease.

Next, we propose a method to select reservation intervals that increases the probability of selecting a free interval and the decreases the amount of computation needed for checks.

In reservation protocols in mesh networks, as exemplified by the MCCA protocol in the IEEE 802.11s mesh standard, each station selects a fixed reservation interval, and the reservations are scheduled relative to this fixed reservation interval. Each selected interval shall be a subinterval of a maximum reservation interval. The current state of the art admits of three methods to select these subintervals:

1. each node selects an interval independent of the other nodes 2. the standard defines a mandatory (sub)interval length, to which each node adheres

3. the nodes execute some protocol to agree on some subinterval length.

It has been noticed that each method has an essential drawback. In short, with method 1 , the subintervals may be chosen so that the verification must be done against many frames of neighbouring stations. In this case, the verification of a new reservation is computationally demanding and the probability of selecting a free slot for reservation is small. Method 2 limits the flexibility to select the subintervals, whereas method 3 induces protocol overhead. We now expand on the drawback of Method 1. Assume that we have a slotted medium and that station A selects a reservation interval of 3 frames, whereas station B selects a reservation interval of 5 frames. The situation is depicted in Figure 6, where the frames are synchronised, so as to have the same starting position. It can be observed that the frames shift relative to each other, due to the fact that 3 and 5 are primes, and, consequently, relatively prime. Consequently, if A intends to schedule a reserved time slot in frame 1 of its reservation interval, it needs to check all frames 1, 2, and 3 of station B in order to find out whether the intended slot is indeed available and not already reserved by station B. Conversely, B need do the same checks against all frames of station A.

Figure 6 shows frame sequences of stations A and B, where station A has a reservation interval of 5 frames, and station B has a reservation interval of 3 frames.

More generally, the maximum reservation interval is represented by a value M, where M is expressed in some convenient basic unit. Thus the maximum reservation interval can be expressed in frames consisting of mini slots. Alternatively, it can be expressed in beacon intervals. Assume that the choice of reservation interval is left free, and subject only to the limitation that it must be smaller than the maximum reservation interval. In this case, neighbouring mesh stations may select reservation intervals that are poorly compatible and this leads to increased computational complexity. In particular, assume that station A takes a reservation interval of length I A = M, and that nodes B takes h = M-l. In this case, for node A to make a reservation in one of its frames, it needs to verify this reservation against M-l frames of node B. Thus the computational complexity to arrange for 1 reserved transmission in a superframe of M blocks equals M-l.

The situation is aggravated in mesh networks where the node will have not one, but multiple neighbours. In this case, a new reservation must be verified against the existing reservations of all neighbouring stations. Indeed, assume that node A has s neighbours. Moreover, assume that all neighbours select reservation intervals of length M-l . In this case, station A needs to verify its reservation against all frames of the neighbouring stations. It follows that the computational complexity to set up for one reserved transmission in a superframe of M blocks equals s(M-l).

Additionally, selecting poorly compatible reservation intervals, also leads to a reduced success rate in finding a location for a new reservation. Indeed, if a new reservation in a frame must be verified against M-l frames of a neighbouring station, then each of these frames is a competitor frame for the new reservation and may contain an existing reservation that conflicts with the new reservation to be set up. Thus, a high number of competitor blocks also leads to a reduced success rate. In fact, assume that the reservation interval contains a fraction of/ free slots and that these are randomly distributed over the frames that make up the reservation interval. Assume that there are c competitor blocks then the probability that the selected frame is free in all of these c blocks equals f ^C(f to the power c), so that the success rate decreases exponentially with the number of competitor blocks.

One way to decrease the number of verifications when setting up a new reservation is given by Method 2. Here, the standard selects one subinterval and it is mandatory or recommended for all nodes in the network to use this subinterval. However, this would limit the nodes considerably. E.g., if a fixed beacon interval is chosen the mesh nodes would beacon at the same rate, and no differentiation, would be possible that reflects their power supply situation. However, in practice, nodes that have a battery will be set up to beacon less often than nodes that are mains powered.

Alternatively, Method 3 proposes that nodes engage in a protocol to negotiate a common subinterval. However, this would result in additional overhead, both at network set up and when nodes join and leave the mesh network in a dynamic fashion. It is the object of this invention to define a large set of subintervals of the maximum interval that is allowed so that each node has sufficient flexibility to select its subinterval. Yet, that the computational resource needed to compare these subintervals is limited.

Let M be the maximum interval that is allowed. E.g., this can be the maximum interval allowed for beaconing, or this can be a maximum interval allowed to define the reservations (such as the maximum DTIM interval in IEEE l is). The purpose of the invention is to define a method to generate a (large) set of subintervals of the maximum interval that they are easily compatible.

To this end, let (p,, nj) be the prime factorization of M, i.e.,

i=l

Where pi is the sequence of prime numbers, and is their multiplicity.

We use the prime factorization to define the subintervals. Define:

With 0 < kj < rij. Then the collection of intervals {I(ki, k„)_ 0 < h < n , i = 1, .., n} defines a set of

subintervals. With the intervals chosen in this manner, the number of frames against which to verify a new reservation is substantially reduced. In fact, if a station has one neighbour, it can be shown that the number of frames against which to verify to arrange for 1 reserved transmission in a superframe of M blocks equals 1. More generally, assume that a station A has s neighbours and all stations that select reservation intervals from the restricted set, In this case, to arrange for one new reservation in the superframe of M frames, station A must verify the new reservation against s frames of neighbouring station.

Thus, by restricting the set of reservation intervals to a subset with lengths as above, the computational complexity needed to verify a new reservation is reduced with a factor M as compared to the worst case situation in case the selection of reservation interval lengths is left free.

Additionally, eliminating poorly compatible reservation intervals, also leads to an improved success rate in finding a location for a new reservation. Indeed, if a new reservation in a frame must be verified against just one frames of a neighbouring station, then this frame can only contain one existing reservation that conflicts with the new reservation to be set up. Thus, the reduced number of competitor blocks also leads to an increased success rate in finding new reservations.

In the embodiments below, the maximum M is defined in the standard. Hence, the standard can also define the set of allowable subintervals by explicitly restricting to the subintervals defined above. Hence, the nodes in the mesh network can utilize this set of intervals and do not need to embody tools to generate the prime factorization.

EMBODIMENTS OF THE INVENTION

In the mesh Draft standard PIEEE802.11s version 7.0, the DTIM interval is used to make reservations; i.e. reservations are defined within a DTIM interval, and relative to the start and duration of the DTIM interval.

Now, in order to make these reservations compatible they must have a smallest common multiple that is limited to the largest possible DTIM interval. Now for this case, the largest Beacon Interval that is needed equals 65535 msec. In this case,

65536 msec = 2¹⁶ 10³ )iSQC = 2⁶ 10³ 2¹⁰ )X,SQC = 2⁶ 10³ TUs,

Where 1 TU = 2¹⁰ \i&QC is a unit that is often used to express the length of DTIM intervals. Next, 2⁶ 10³ TUs = 2⁶ 10 100 TUs, Where lOOTUs is a "basic" beacon interval that is used by most APs that are in the market. It is then useful to take lOOTUs as the basic unit for a DTIM interval, to take M = 2⁶ 10 = 2⁷5, and to allow only intervals of the form 1(1, j) = 2^l 5¹, with 0 < I < 7 and 0 < <1.

Note that this provides a choice 16 different subinterval. Also, most common choices for subintervals like 1, 2, 5, 10, .. TUs have been covered.

More flexibility is given by allowing higher power of 5, taking M = 5*65536 and using intervals of the form 1(1, j) = 2¹ with 0 < I < 7 and 0 < <2. Because this set of intervals extends the choice to also include (o.a.) intervals of lengths 25, 50, and 100 TUs

As an alternative, it is possible to use an upperbound for the largest Beacon Interval that is needed. The largest interval equals 65536 msec. As before,

65536 msec = 2¹⁶ 10³ μβεΰ = 2⁶ 10³ 2¹⁰ μ₈εΰ = 2⁶ 10³ TUs 2⁶ 10 100 TUs < 2⁶2⁴ 100 TUs,

In this case, we take M = 2⁶2⁴ = 2¹⁰, and to allow only intervals of the form

1(1, j) = 2¹0≤l < 10.

As compared to the previous choice, this limits the amount of possible subintervals to 10.

In the present specification and claims the word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. Further, the word "comprising" does not exclude the presence of other elements or steps than those listed.

The inclusion of reference signs in parentheses in the claims is intended to aid understanding and is not intended to be limiting.

From reading the present disclosure, other modifications will be apparent to persons skilled in the art. Such modifications may involve other features which are already known in the art of wireless communications and which may be used instead of or in addition to features already described herein.

Claims

1. A method for determining a length of a reservation interval in a node of a wireless mesh network, the method comprising the following steps :

defining a value M that is greater than or equal to the maximum length of the reservation interval,

determining the prime factorization of M,

determining a subset of intervals length

picking one element of the subset.

2. A method according to claim 1, wherein the value M that is greater than or equal to the maximum length of the reservation interval is predetermined in the network.

A method according to claim 1 or 2, comprising the initial step of defining a Time Unit TU corresponding to a prime factorization, and comprising the step of dividing each element of the subset by TU;

4. A method according to claim 3, where in M is equal to 65536 msec, and wherein TU is equal to 2¹⁰

5. A method according to claim 3, where in M is equal to 5 *65536 msec, and wherein

TU is equal to 2¹⁰ and where in a subset of intervals length is chosen as

I(k₁,k₂) = 2"¹5"²

6. A node in a wireless mesh network, comprising means for carrying out a method according to any of claims 1 to 5.

7. A network comprising nodes according to claim 6.

A network according to claim 7, wherein the network is a mesh network according the standard IEEE802. l is.