Protocol for wireless multi-hop ad-hoc networks

[0001] The present invention generally relates to the field of wireless peer-to-peer networks, in particular to a service discovery protocol and a corresponding method for providing low-profile, low-overhead service discovery in-formation needed for determining the availability of requested remote services in a wireless multi-hop ad-hoc network designed for operation in e.g. home environments and personal usage scenarios.

[0002] Due to the widespread use of new personal network-enabled mobile devices (such as MD player and camcorder) and the emergence of near-ubiquitous communication infrastructure based on wired and wireless networks, it is becoming increasingly important to allow users and applications to interact with existing services without any need for configuration or administration. The key requirement for such a mechanism is to enable new services to advertise their existence and properties and allow service users to discover and utilize the services they need. These tasks are addressed by service discovery protocols.

[0003] Service discovery protocols which enable service providers to advertise capabilities to potential clients, thereby providing to clients and service providers a means for entering into a relationship, play a key role in mobile and wireless networks. Existing protocols addressing service discovery include e.g. IETF's Service Location Protocol (SLP), Sun Microsystems' Jini, Bluetooth's Service Discovery Protocol (SDP), Salutation—a nonproprietary service discovery protocol which is currently utilized by several shipping products—and the Universal Plug and Play (UPnP) protocol promoted by Microsoft.

[0004] Finding services within a wireless multi-hop ad-hoc network is a resource-consuming task with today's state-of-the-art service discovery protocols as these protocols often apply broadcast or centralized control structures which are not applicable to the special characteristics of wireless multi-hop ad-hoc networks. Therefore, it is necessary to implement some smart, distributed networking functionality where e.g. an application uses remote resources (services) to provide some functionality towards a user.

[0005] In the following, the requirements towards a service discovery protocol to be deployed within a home- or personal-type network scenario shall be explored, and a possible protocol for that purpose shall be discussed.

[0006] For service discovery in a wireless multi-hop ad-hoc network some boundary conditions have to be taken into account. Any of these boundary conditions translates to a requirement item that needs to be covered by any service discovery protocol that has to operate within a target environment. Even though a service discovery protocol might be operable in another scenario than the one given in a target environment, it should be tailored to provide optimal operation within said scenario.

[0007] Due to the ad-hoc nature of the target environment, any service discovery protocol applied has to be operational even with a varying network topology where nodes are allowed to permanently join or leave the network. In spite of having some indications on how stable a node's network affiliation is—stationary or mobile—, its presence can not be taken for granted over a certain period of time.

[0008] Hence, any service discovery protocol should operate without a single point of failure (i.e. without any central component).

[0009] In a typical multi-hop ad-hoc network sending network-wide broadcast messages is costly in terms of used resources. Furthermore, bandwidth is a scarce resource and its usage should be minimized while still allowing a high responsiveness to discovery requests. Any service discovery protocol needs to consider this by avoiding the transmission of broadcast messages and reducing the amount of background traffic.

[0010] Another problem that might probably occur in an ad-hoc networking environment refers to network partitions and rejoins. A service discovery protocol has to handle this scenario efficiently and quickly.

[0011] In e.g. a home network any configuration by the user side is to be avoided. Software can not easily be upgraded without having a mobile code. Once a system is deployed, it is assumed to be operational for a longer period of time.

[0012] As for every service discovery protocol there are some typical requirements that have to be fulfilled. For each of the following items it has to be ensured that they are working under the above-stated boundary conditions.

• • [0013] Discovery of services: The very basic (and most obvious) requirement towards a service discovery protocol is to find any service of interest on the network. For some applications it is additionally required to keep an up-to-date view on the services on the network, thus demanding not only a discovery of service but also monitoring services and possibly their states. Furthermore, it should be possible to perform the query for services in a diffuse way, e.g. by finding a service by only giving an incomplete list of attributes that the service shall met.
  • [0014] Consistent and simple service description: As a part of the service discovery not only communication protocols but also data formats need to be specified. These formats are needed to prevent inconsistencies and ambiguities.
  • [0015] Fast setup and reaction to changes: Any service discovery protocol shall strive to provide quick responsiveness to changes in the network, i.e. a newly introduced service shall be made known as fast as possible to other nodes. Minimizing the latency will influence the perceived agility of the network (eventually by the user). This problem is especially evident in an ad-hoc environment where changes in the network topology are probably more the rule than the exception.

[0016] The problem to be solved can be very quickly described as the need to implement the required functionality as described above under the aforementioned boundary conditions.

[0017] Publish/subscribe- versus request/response-type service discovery protocols: When a publish/subscribe-type discovery protocol is applied, the network registers at a certain source or channel for dedicated service discovery messages. These messages are then sent in case a service becomes available or is no longer present, respectively. In a request/response-type discovery protocol a certain service is only searched if it has explicitly been requested by a certain client. Information on service availability is not announced. Hence, monitoring the availability of services requires periodic polling of all (or some specific) services in the network. Of course, hybrid approaches (mixtures of the above two) are also conceivable.

a) Universal Plug and Play (UPnP)

[0018] Universal Plug and Play is an open network architecture developed by an industry consortium leaded by Microsoft Corporation. UPnP offers ad-hoc peer-to-peer network connectivity of different services and devices.

b) Service Location Protocol (SLP)

[0019] The Service Location Protocol (see http://www.srvloc.org/) is the service discovery protocol proposed by the IETF. It is being developed by the Srvloc Working Group and is vendor-independent. SLP is designed for TCP/IP networks and intended to become the standard in the Internet community. The current version of SLP is SLPv2. The SLP architecture is basically composed of three members: User Agents (UAs), Service Agents (SAs), and Directory Agents (DAs).

c) Salutation

[0020] The Salutation architecture is an industry consortium's solution to the service discovery and utilization problem. The architecture provides a standard method for applications, services and devices—so-called Networked Entities—to advertise their capabilities or request the desired ones. It is claimed to be processor-, operating-system and communication-protocol independent. Its key piece is the Salutation Manager (SLM). Every Networked Entity has an SLM or uses a remote SLM by means of the Remote Procedure Call (RPC) protocol. The SLM provides services and clients with a transport-independent interface (SLM-API). A Networked Entity can act as service, client or both. The different SLMs communicate among themselves using the Salutation Manager Protocol, which is based on remote procedure calls.

d) Jini

[0021] Jini is a service-oriented Java-based architecture (infra-structure and programming model) developed by Sun Microsystems. The architecture of Jini is based on the Jini Lookup Service (JLS) component. The services have to locate a JLS server—by using the discovery protocol—and then register themselves in the JLS by using a join protocol. The clients also discover a JLS and can then query it about the available services. The matching between queries and services can be made comparing Java interfaces or a list of characteristic attributes. Each service will be maintained by the JLS only for a certain period of time, i.e. for a lease period. In this way services which are not registered are eliminated from the register.

e) Bluetooth

[0022] A Bluetooth-enabled personal area network (PAN) consists of large-sized multi-hop networks in which mobile devices can communicate not only via a master node with other mobile devices located in a single piconet but also with wireless nodes which can be reached via a multiplicity of intermediate nodes. Mobile devices are thus able to communicate and use services provided by other mobile devices or infrastructure systems. To allow mobile terminals to use these services, service providers have to publish all available services together with some basic configuration information, and service users have to be equipped with means for searching these services and selecting a specific service provider. The Bluetooth Service Discovery Protocol (SDP) thereby defines how a Bluetooth client's application shell acts to discover services offered by available Bluetooth servers and their characteristics. SDP thereby allows client applications to access services by using other discovery protocols such as SLP, Salutation, etc., but it does not need them. The protocol defines how a client terminal can search for a remote service based on specific attributes without knowing anything about the availability of said service. It provides means for discovering new services that become available when a client terminal enters an area where a Bluetooth server is operating. SDP also provides functionality for detecting when a service is no longer available.

[0023] A device containing an SDP client can search services specifying its class or some of its attributes, and it also can retrieve services without knowing its characteristics. In turn, a remote SDP server will respond to these inquiries. Devices must be aware of the availability of new services and know about the unavailability of known ones. The use of intermediary agents as caches to improve the efficiency of the system is allowed.

[0024] US 2002/0120750 A1 describes a method, a wireless network device and a computer program product for performing service discovery in a pervasive wireless local area network, e.g. in an ad-hoc Bluetooth PAN consisting of a number of multi-hop networks.

[0025] U.S. Pat. No. 6,397,061 B1 refers to a method and apparatus for reprioritizing data transfer in a short-range mobile ad-hoc network (MANET) applied to a wireless communication device capable of communicating with a local wireless network within a predetermined communication range.

[0026] A communication device and a software for operating multimedia applications in at least one communication network is described in US 2001/0003191 A1.

[0027] EP 1 227 689 A1 pertains to an entry gateway server provided to support mobile devices in the discovery process of local services.

[0028] EP 1 022 876 A1 refers to a method for advertising services offerings in wireless local area networks comprising at least two mobile terminals and an apparatus for exchanging service information with other mobile terminals.

[0029] A scheme and an apparatus for distinguishing services offered by a service-providing device in adjacency of the apparatus from services offered by a service-providing device not being in the apparatus' adjacency is described in EP 1 024 628 A1.

[0030] An advanced system and method for dynamically discovering, provisioning and accessing host services on wireless data communication devices needed for sending a service book to a mobile device is described in WO 02/084975 A2.

[0031] WO 02/23826 A2 pertains to a service framework supporting service discovery and connection, in particular to an information appliance system with a user device comprising a client platform that includes a service framework to discover and connect with a variety of services, both remote and local, transient and persistent, and to disconnect from said services when they are no longer of interest or become unavailable.

[0032] WO 02/45382 A2 is directed to a method and device for providing a service record for an application (e.g. a legacy application) running on a virtual serial port of a wireless transceiver device, such as a Bluetooth-enabled device.

[0033] OBJECT OF THE PRESENT INVENTION

[0034] It is the object of the present invention to propose a technology for improving peer-to-peer service discovery in wireless multi-hop ad-hoc networks.

[0035] This object is achieved by means of the features of the independent claims. Advantageous features are defined in the subordinate claims. Further objects and advantages of the invention are apparent in the detailed description which follows.

[0036] The present invention is basically dedicated to a service discovery protocol and a corresponding method for providing low-profile, low-overhead service discovery in-formation needed for determining the availability of requested remote services in a peer-to-peer-based wireless multi-hop ad-hoc network organized according to the store-and-forward messaging principle which is designed for operation in e.g. home environments and personal usage scenarios.

[0037] According to one embodiment of the present invention, received service announcement messages are transmitted, said messages referring to remote services offered by service providers within said network. Thereby, outdated messages which are identical with old service announcement messages that have already been received by this peer are discarded, and new service announcement messages are accumulated in a local message pool assigned to said peer and sorted according to their potential relevance before being propagated to said neighboring peers. First, each service announcement message received by this peer is tagged with a relevance value. After that, relevance values of all service announcement messages stored in said local message pool are summed up, thus yielding a cumulative relevance value. Once this cumulative relevance value exceeds a predefined relevance threshold value, all service announcement messages stored in said local message pool are aggregated and sent to all neighboring peers of said peer.

[0038] Further advantages and possible applications of the present invention result from the subordinate claims as well as from the following description of different embodiments of the present invention, accompanied by the following drawings:

[0039] FIG. 1 shows different characteristic features, advantages and disadvantages of conventional service discovery mechanisms and protocols according to the state of the art,

[0040] FIG. 2 is a diagram showing different peer protocol layers used for the handling of a metadata-based service description according to the present invention,

[0041] FIG. 3 is a first UML message sequence chart, which shows the interactions for registering a remote service according to the present invention,

[0042] FIG. 4 is a second UML message sequence chart, which shows the interactions for registering a local service according to the present invention,

[0043] FIG. 5 is a third UML message sequence chart, which shows the interactions for handling a service message indicating the availability of a remote service according to the present invention,

[0044] FIG. 6 is a fourth UML message sequence chart, which shows the interactions for deregistering a local or remote service according to the present invention,

[0045] FIG. 7 is a UML state chart illustrating the service discovery mechanism executed by the service discovery protocol 208 according to the present invention, and

[0046] FIG. 8 is a UML class diagram showing the classes for an object-oriented implementation of the service discovery protocol according to the present invention.

[0047] In case a peer requires a certain service for any of the client applications and/or services it is running, it has basically two different possibilities: proactively querying for any service which might be of interest to a user or listening to some channels for service announcements.

[0048] To find a specific service with given attributes in the network, a reactive service discovery mechanism can be executed. For this purpose, the data link control (DLC) layer already implements some low-level service discovery means. To utilize this functionality, the following three-phase approach is proposed:

• • [0049] 1. First, a local service table is queried for a specific service of interest.
  • [0050] 2. If locally none is available, a DLC service discovery protocol is executed.
  • [0051] 3. Once some services of interest are known, more detailed information on them is retrieved and forwarded to the requesting peer.

[0052] To further enhance the responsiveness of the applied service discovery mechanism, not only proactive service querying but also service announcement is supported. The design target of the proposed service discovery protocol is to provide some robust and efficient service announcement means while minimizing the number of messages to be sent. Thereby, broadcast messages are avoided as they are rather costly in a wireless ad-hoc networks.

[0053] According to one embodiment of the present invention, every peer N_i implements a so-called service announcement message pool P_i. Thereby, any service announcement message M_ik received by this peer N_i is tagged with a relevance value r_ik. After that, all relevance values r_ik in the pool are summed up, thus yielding a cumulative relevance value $\begin{array}{cc}{R}_i:=\sum _{k=0}^{K_i-1}{r}_{\mathrm{ik}}\mathrm{with}{r}_{\mathrm{ik}}\in 0_{},& \left(1\right)\end{array}$
wherein K_i denotes the total number of service announcement messages M_ik stored in said peer's pool P_i. Once this cumulative relevance value R_i exceeds a predefined threshold value R_th,i, all messages in the pool P_i are aggregated and sent to any neighboring peer N (≠i). Said relevance values r_ik can be very small for service announcements that are not of high importance and higher for more important ones. The speed with which messages are distributed in the network can thus be controlled in a fine-grained way.

[0054] In case the applied transmission media support one-hop broadcasting, a single service announcement message pool P_i can be used for the entire neighborhood of a wireless peer N_i. If this is not the case or in case the message delivery should be controlled in a more fine-grained manner, another pool P_j can be implemented. This allows e.g. to set the threshold values R_th,i of pools P_i assigned to mobile peers N_i to a higher value than the threshold values R_th,j of pools P_j assigned to stationary peers N_j. It is also possible to define a non-fixed threshold value R_th,i, which means a time-variant threshold function R_th,i(t) that can be changed according to external conditions. For example, when determining that a neighboring peer N is running low on energy and/or other information concerning varying resource availability within said network and/or changing link characteristics between a peer N_i and its neighboring peer N, the threshold value R_th,i of the according service announcement message pool P_i of peer N_i is increased in order to avoid sending unnecessary messages M_ik to the neighboring peer N.

[0055] While the cumulative relevance values R_j:=Σ_kr_jk on stationary peers N_j must not exceed a predefined threshold value R_th,j, mobile peers N_i are free to choose any value for their threshold value R_th,i. Setting this threshold value R_th,i to “infinite” would be equal to not sending any messages at all while a value of zero would force every message M_ik to be resent immediately.

[0056] The basic protocol for propagating service announcement messages is rather simple in order to allow an efficient execution. In principle, the following steps need to be executed by any peer participating in the service announcement procedure according to the present invention:

• • [0057] 1. Any service announcement is done with an additional integer relevance parameter r_ik as well as a real-valued relevance degeneration rate d_ik between 0 and 100%.
  • [0058] 2. If a service announcement message M_ik has been seen before on a peer N_i, it is discarded. In any other case, the message is locally stored on this peer by putting it into a pool P_i of received service announcement messages M_ik.
  • [0059] 3. After that, all relevance values r_ik assigned to service announcement messages M_ik stored in this pool P_i are added, thus yielding a cumulative relevance value R_i:=Σ_kr_ik. If the cumulative relevance value R_i exceeds a predefined threshold value R_th,i, the service announcement messages M_ik are aggregated and sent to all peers in the neighborhood of said peer N_i. The same happens if the number K_i of service announcement messages M_ik in the pool P_i exceeds a certain number threshold K_th,i or if the time Δt_ik passed since the reception of the last service announcement message M_ik exceeds a predefined time period threshold Δt_th,i. These conditions can be summarized as follows: $\begin{array}{cc}\forall i:\left[\left(R_i>R_{\mathrm{th},i}\right)\bigvee \left(K_i>K_{\mathrm{th},i}\right)\bigvee \exists k:\left(\Delta t_{\mathrm{ik}}>\Delta t_{\mathrm{th},i}\right)\right]\Rightarrow \mathrm{Send}\underset{k=0}{\bigcup ^{K_i-1}}{M}_{\mathrm{ik}}\mathrm{from}{N}_i\mathrm{to}\mathrm{all}\mathrm{neighbors}{N}_l\left(l\ne i\right).& \left(2\right)\end{array}$
  • [0060] 4. Any peer N_i that receives a service announcement message M_ik puts it into its local pool P_i. Furthermore, said peer N_i recalculates the relevance value r_ik of the incoming message by subtracting the percentage of degeneration given by the relevance degeneration rate d_ik from the respective relevance value r_ik: $\begin{array}{cc}{r}_{{\mathrm{ik}}^{\prime }}:=r_{\mathrm{ik}}·\left(1-\frac{d_{\mathrm{ik}}\left[\%\right]}{100}\right)\in 0_{+}^{}\mathrm{for}{r}_{\mathrm{ik}}\in 0_{}\mathrm{and}{d}_{\mathrm{ik}}\in \left[0\%,100\%\right].& \left(3\right)\end{array}$

[0061] The degeneration of the relevance value r_ik allows to create messages M_ik that initially have a high relevance value r_ik (and are thus quickly sent throughout the network) but get “less important” after some hops. Thus, service announcement messages M_ik are quickly spread to near-by peers but only slowly propagated to peers farther away.

[0062] Every message M_ik that is sent contains a service identifier (or service description), a message identifier and the address of the peer N_H hosting the service. The message identifier is created from a peer internal counter that is incremented after a message M_ik was tagged.

[0063] Whenever a peer N_i receives a new message M_ik, it checks for the lastly seen message M_i,k−1 with the given service identifier from the given peer. If the stored message identifier is newer than the one of the new message M_ik, this new message is discarded.

[0064] Additionally, each available service is associated with a lease timeout before which the available service has to be re-announced. This mechanism is used to avoid old entries in the service tables.

[0065] Service availability announcements are triggered when a service which has to be made known to client terminals connected to the wireless multi-hop ad-hoc network registers itself on a local peer N_i. Thereby, any service availability announcement message M_ik is not only put into the peer's service announcement message pool P_i but also causes an update of the local service table.

[0066] In the same sense, the availability of a service has to be made known to other peers, whereas the loss of a service has to be propagated throughout the entire network. In principle, the reasons for announcing service loss will mainly be that a service is unregistered at a peer. The case that a peer is no longer present in the network and thus services hosted on that peer are also lost is extra handled by monitoring the availability of wireless peers. The overall process for announcing a message is the one described above.

[0067] In order to remove outdated entries from peers which are no longer present in the network, a mechanism for indicating peer loss is added to the service discovery protocol. This mechanism supports the leasing mechanism already introduced above.

[0068] Once a new neighboring peer N_j becomes visible, service tables and management data of the services offered by this peer (e.g. its message counter) are requested. The returned services are checked whether they have already been handled. If this is not the case, they are added to the local message pool P_i of the respective peer N_i. Thereby, network fluctuations are used to furthermore trigger message exchanges and to contribute to the overall availability of messages in the network. To avoid sending unnecessary messages, negotiation procedures as described below can be applied.

[0069] Since message overheads caused by said negotiation procedure should be avoided, a new peer joining said network is only allowed to clone the state of one (or some) of its neighbors.

[0070] So far we only described a pro-active service presence information dissemination. Of course, the given protocol can also be used to forward requests for services to other nodes.

[0071] Whenever the pool content is sent to anyone of a wireless peer's neighbors (or to a multiplicity of them), a short outline of the messages to be sent will be delivered. The information sent includes (at least) the service identifier, the address of the peer NH hosting the service and the message identifier. With this information any peer can decide if the message is of interest to it or not. Accordingly, there are the following options available in case peer N_i offers peer N_i+1 a message M_ik:

• • [0072] 1. In case this message M_ik is not of interest for peer N_i+1, it can request peer N_i to delete the message.
  • [0073] 2. If said message M_ik is possibly of interest to peer N_i+1 but N_i+1 does not yet want to receive it (i.e. because peer N_i+1 is going to initiate a sleep mode and does not process further messages), it can request peer N_i to store the message for later delivery.
  • [0074] 3. It might also be possible that peer N_i+1 has an updated message M_i+1,k that makes the one (M_ik) of peer N_i obsolete (e.g. peer N_i wants to send a service announcement, but peer N_i+1 already knows that the respective service is no longer available). In this case, peer N_i+1 provides peer N_i with the updated message M_i+1,k.

[0075] Within a wireless multi-hop ad-hoc network a number of services can be available at the same time. Some of them might be seen as a conglomerate service (e.g. the television control and the surround control services might be seen as one home entertainment service). In case a service user wants to find such a conglomerate service, it specifies such within a query. Decomposing the conglomerate service into particular services and discovering them within the network is then done by the service discovery protocol according to the present invention. For this purpose, a request for the conglomerate service is spread within the network, and each peer wishing to contribute to it will send back a corresponding response. These responses are collected at the requesting peer, and a check is done if the conglomerate service can be created or not.

[0076] As basis for the service description only key/value pairs are supported. They allow describing a service by assigning key (attribute) values. Since only the basic service discovery mechanism is implemented by the given protocol, this information will be sufficient. In contrast to other service description languages, e.g. the one in UPnP, information like event or state description is intentionally left out to minimize overhead. To still provide such a functionality towards higher layers, an additional component is introduced: the Service Discovery (SD) Metadata Handler 204. This component can mediate between more complex higher-layer service description means, as e.g. the one of UPnP, and the proposed service discovery protocol 208 according to the present invention (see FIG. 2). For this purpose, it extracts the core service description attributes to pass them to the service discovery protocol layer. Once the service discovery protocol has found some services, it correspondingly re-assembles the answer in the expected format. If required, such a meta-data handler 204 could be deployed to map various service discovery protocols to the given service discovery protocol 208.

[0077] Even though the protocol is especially designed for the purpose of service discovery, the basic communication scheme might also work well for other data such as e.g. network status information, device presence messages, event propagation or any kind of data that do not have real-time constraints in terms of data delivery. Such kind of data could be (among others): distributed routing information, distributed information to support/enable QoS provisioning, network status information, (instant) messaging data, sensor data, peer status information, etc.

[0078] The procedures of registering a service, handling a service message indicating the availability of said service according to the present invention and deregistering the service are illustrated in the UML message sequence charts 300, 400, 500, and 600 depicted in FIGS. 3 to 6. An overview of the proposed service discovery mechanism executed by the service discovery protocol 208 according to the present invention is given by the UML state chart 700 depicted in FIG. 7. Finally, a UML class diagram 800 showing the classes for an object-oriented implementation of the service discovery protocol according to the present invention is depicted in FIG. 8.

[0079] The flexibility of the proposed service discovery protocol provides some degrees of freedom, some of which will subsequently be briefly discussed:

• • [0080] 1. Any peer N_i can locally decide when and how many announcement messages its service announcement message pool P_i contains. Peers with power constraints such as mobile peers N_i might have some larger pools P_i to avoid sending too many messages M_ik while stationary peers N_j might have small pools P_j, thus forcing them to send more messages M_ik. Moreover, a threshold value R_th,i (R_th,j) for the cumulative relevance value R_i (R_j) can be set according to the current constraints of the respective mobile (N_i) or stationary peer (N_j). The value itself can be changed dynamically by the respective peer according to its current state.
  • [0081] 2. With the possibility to have service announcement degeneration, available services are quickly announced in the immediate neighborhood of a peer N_i (N_j) but become only slowly visible at peers farther away.
  • [0082] 3. Services which have to be made known to the entire network very quickly can be sent with a relatively high relevance value r_ik (r_jk); others can be sent with a lower one. In any case, the speed of spreading a service announcement message M_ik within the network can easily be controlled by peer N_i (N_j).
  • [0083] 4. Fluctuations in the number of the peers forming said mobile multi-hop ad-hoc network can be used to trigger message exchanges and thus contribute to overall information dissemination.

[0084] It should be noted that neither of the given items and especially the combination of them can already be found in conventional service discovery protocols according to the state of the art.

[0085] The proposed protocol is specially tailored to the requirements of wireless multi-hop ad-hoc networks designed for operation in e.g. home environments and personal usage scenarios and features the following characteristics:

• • [0086] No broadcasting: In wireless multi-hop ad-hoc networks broadcasting is an expensive and power-consuming task (if it is possible at all). Hence, the proposed service discovery protocol can be seen as an enabling technology for service discovery in wireless multi-hop ad-hoc networks.
  • [0087] Sending no unnecessary messages, thus avoiding traffic overhead and preserving energy: This is especially of interest for mobile, battery-powered peers. The proposed protocol allows building devices that are less power-consuming by offering the same (or even better) functionality compared to the state of the art.
  • [0088] Keeping a very up-to-date view of the immediate neighborhood of a wireless peer (e.g. a television set located in the same room) and thus an enhanced user experience when utilizing this technology.

[0089] Utilizing functionality at lower layers and keeping the protocol rather simple allows easily integrating (at least some) functionality into hardware if required.

[0090] A further embodiment of the present invention pertains to a peer N_i serving as a proxy server for providing service discovery information needed for determining the avail-ability of requested remote services in a peer-to-peer-based wireless multi-hop ad-hoc network based on a store-and-forward messaging principle. Thereby, said peer comprises a service discovery manager unit 204 for implementing a method as described above.

[0091] The invention finally pertains to a software program product designed to support a service discovery method as described above when running on this peer N_i.

[0092]

	Technical Term	Brief Explanation
	Service	Service is any component within a network
		that exposes some interface. The interface
		must be accessible externally, i.e. from
		other processes and possibly from other
		peers within the network. The process of
		exposing the services is given by the
		utilized service discovery protocol and its
		implementation.
	Service	Service Discovery describes the process of
	Discovery (SD)	finding suitable services within the
		network. A service is matching a request if
		it fulfills some specification as given by
		the instance requesting the service
		discovery. This service specification might
		only be a simple service type, but could
		consist of a more complex description of the
		desired service by e.g. a list of attributes
		that has to be met. Details on how services
		are looked up are defined by the according
		service discovery protocol.
	Target	Target environment is a wireless multi-hop
	Environment	ad-hoc network tailored for operation in
		e.g. home environments and for personal use.

[0093]

No.	Technical Feature (System Component or Procedure Step)
100	table comparing different characteristic features,
	advantages and disadvantages of conventional service
	discovery mechanisms and protocols
200	diagram showing different peer protocol layers used for
	the handling of a metadata-based service description
	according to the present invention
202	application layer of said peer protocol for running client
	applications
204	intermediate protocol layer for a service discovery meta-
	data handler
206	interface between the service discovery protocol layer 208
	and the application layer 202 as well as the intermediate
	protocol layer 204 for said service discovery metadata
	handler
208	service discovery protocol layer for determining the
	availability of requested remote services
300	UML message sequence chart showing the interactions for
	registering a remote service according to the present
	invention
302	element aService
304	element aBoundary, derived from the class ServiceDiscovery
	802
304R	element aRemoteSDBoundary, derived from the class
	RemoteInterfaceSender 816
306	element aController, derived from the class
	ServiceDiscoveryHandler 806
306R	element aRemoteController, derived from the class
	ServiceDiscoveryHandler 806
308	element aServiceHandleFactory, derived from the class
	ServiceHandleFactory 818
310	element aServiceTable, derived from the class ServiceTable
	(not shown)
312	element aDLCBoundary
314	element aMessagePoolHandler, derived from the class
	MessagePoolHandler 808
314R	element aRemoteMessagePoolHandler, derived from the class
	MessagePoolHandler 808
316	element aMessagePoolImpl, derived from the class
	SimpleMessagePool 810
400	UML message sequence chart showing the interactions for
	registering a local service according to the present
	invention
500	UML message sequence chart showing the interactions for
	handling a service message indicating the availability of
	a remote service according to the present invention
600	UML message sequence chart showing the interactions for
	deregistering a local or remote service according to the
	present invention
700	UML state chart illustrating the service discovery
	mechanism executed by the service discovery protocol 208
	according to the present invention
702a	a first peer (peer A) in a wireless multi-hop ad-hoc
	network
702b	a second peer (peer B) in said wireless multi-hop ad-hoc
	network
704a	state ,,Create Message,, of said first peer 702a (peer A)
704b	state ,,Negotiate Message to Receive,, of the first peer
	702a (peer A)
704c	state ,,Awaiting Message,, of the first peer 702a (peer A)
704d	state ,,Check Message,, of the first peer 702a (peer A)
704e	state ,,Process Messages,, of the first peer 702a (peer A)
704f	state ,,Check Pool State,, of the first peer 702a (peer A)
704g	state ,,Recalculate Pool State,, of the first peer 702a
	(peer A)
704h	state ,,Flush Message Pool,, of the first peer 702a (peer A)
706a	message pool of said second peer 702b (peer B)
706b	state ,,Negotiate Messages to Send,, of the second peer 702b
	(peer B)
706c	state ,,Send Message,, of the second peer 702b (peer B)
800	UML class diagram showing the classes for an object-ori-
	ented implementation of the service discovery protocol
	according to the present invention
802	class ServiceDiscovery
802′	component ServiceDiscoveryEventHandler, derived from the
	class ServiceDiscovery 802
804	class MessagePool, obtained by a generalization of the
	NeighborhoodChangedListener component 808 and the
	SimpleMessagePool class 810
806	class ServiceDiscoveryHandler
806′	component NeighborhoodChangedListener, derived from the
	class ServiceDiscoveryHandler 806
808	class MessagePoolHandler
808′	component NeighborhoodChangedListener, derived from the
	class MessagePoolHandler 808
810	class SimpleMessagePool
812	class ServiceHandle
814	class RemoteInterfaceListener
816	class RemoteInterfaceSender
818	class ServiceHandleFactory
S0	step #0: controlling the delivery of received service
	announcement messages Mik referring to remote services
	offered by service providers within said network to its
	neighboring peers Nl by sorting (S2′) said messages Mik
	according to their potential relevance rik before being
	propagated (S4b) to said neighboring peers Nl and
	discarding (S1b) outdated and irrelevant messages Mik
S1a	step #1a: receiving said service announcement messages Mik
S1b	step #1b: discarding messages Mik which are identical with
	old service announcement messages Mik that have already
	been received by this peer Ni
S1c	step #1c: accumulating new service announcement messages
	Mik in a local message pool Pi assigned to said peer Ni
S2	step #2: tagging each service announcement message Mik
	received by this peer Ni with a relevance value rik
S3	step #3: adding the relevance values rik of all service
	announcement messages Mik stored in said local message
	pool Pi, thus yielding a cumulative relevance value Ri
S4a	step #4a: once this cumulative relevance value Ri exceeds
	a predefined relevance threshold value Rth,i, aggregating
	all service announcement messages Mik stored in said local
	message pool Pi
S4b	step #4b: sending the aggregated service announcement
	messages Mik to all neighboring peers Nl of said peer Ni
S4b′	step #4b′: sending received and aggregated service
	announcement messages Mik to all neighboring peers Nl in
	case the number Ki of stored service announcement messages
	Mik in the local message pool Pi assigned to said first
	peer Ni exceeds a certain number threshold Kth,i
S4b″	step #4b″: sending received and aggregated service
	announcement messages Mik to all neighboring peers Nl in
	case the time Δtik passed since the reception of the last
	service announcement message Mik exceeds a predefined time
	period threshold Δtth,i
S5	step #5: setting the relevance threshold values Rth,i of
	local message pools Pi assigned to mobile peers Ni to a
	higher value than the threshold values Rth,j of local
	message pools Pj assigned to stationary peers Nj
S6	step #6: dynamically changing said relevance threshold
	value Rth,i according to external conditions
S6a	step #6a: determining whether a neighboring second peer
	Nl is running low on energy and/or other information
	concerning varying resource availability within said net-
	work and/or changing link characteristics between said
	peers Ni and Nl
S6b	step #6b: if this is the case, increasing the relevance
	threshold value Rth,i of the according local message pool
	Pi of said first peer Ni to avoid sending unnecessary
	service announcement messages Mik to said second peer Nl
S7	step #7: recalculating the relevance values rik of
	received service announcement messages Mik by subtracting
	a percentage of degeneration given by a relevance
	degeneration rate dik from the respective relevance value
	rik, said degeneration rate dik being the greater the
	greater the number of hops a service announcement messages
	Mik has been propagated such that service announcement
	messages Mik are quickly spread to near-by peers Nl but
	only slowly propagated to peers Nl farther away from the
	first peer Ni
S8a	step #8a: monitoring the availability of neighboring peers
	Nl within said network
S8b	step #8b: in case a peer Nl is no longer present in the
	network and thus services hosted on that peer Nl are
	lost, propagating the loss of these service throughout the
	entire network
S9	step #9: in case the content of the local message pool Pi
	assigned to said first peer Ni is sent to anyone of its
	neighboring peers Nl or to a multiplicity of them,
	supplying said neighboring peers Nl with a short outline
	of service announcement messages Mik to be propagated,
	said outline including at least a service identifier, the
	address of a peer NH hosting the service and a message
	identifier, such that a neighboring peer Nl receiving
	these data can decide whether said message Mik is of
	particular interest to it or not
S10a	step #10a: proactively querying a local service table
	containing information on the availability of specific
	services of interest from a peer NH hosting these services
S10b	step #10b: in case locally none is available, executing a
	service discovery protocol for providing service discovery
	information needed for determining the availability of the
	requested services within said multi-hop ad-hoc network
S10c	step #10c: once some services of interest are known,
	retrieving more detailed information on these services
S10d	step #10d: forwarding this information to all neighboring
	peers Nl