WO2008012301A2 - Procédé pour exécuter un service dans un réseau de données décentralisé - Google Patents

Procédé pour exécuter un service dans un réseau de données décentralisé Download PDF

Info

Publication number
WO2008012301A2
WO2008012301A2 PCT/EP2007/057609 EP2007057609W WO2008012301A2 WO 2008012301 A2 WO2008012301 A2 WO 2008012301A2 EP 2007057609 W EP2007057609 W EP 2007057609W WO 2008012301 A2 WO2008012301 A2 WO 2008012301A2
Authority
WO
WIPO (PCT)
Prior art keywords
service
network node
network
plo
network nodes
Prior art date
Application number
PCT/EP2007/057609
Other languages
German (de)
English (en)
Other versions
WO2008012301A3 (fr
Inventor
Christoph Gerdes
Shankar N. Ratneshwaran
Felix Sprick
Original Assignee
Siemens Enterprise Communications Gmbh & Co. Kg
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from DE102006054090A external-priority patent/DE102006054090A1/de
Application filed by Siemens Enterprise Communications Gmbh & Co. Kg filed Critical Siemens Enterprise Communications Gmbh & Co. Kg
Publication of WO2008012301A2 publication Critical patent/WO2008012301A2/fr
Publication of WO2008012301A3 publication Critical patent/WO2008012301A3/fr

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/50—Network services
    • H04L67/51—Discovery or management thereof, e.g. service location protocol [SLP] or web services
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/50—Network services
    • H04L67/60—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources
    • H04L67/61—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources taking into account QoS or priority requirements

Definitions

  • the invention relates to a method for executing a service in a distributed data network having a plurality of network nodes and a corresponding decentralized data ⁇ network.
  • Decentralized data networks in particular in the form of so-called peer-to-peer data networks, are increasingly being used today .
  • the network nodes of these data networks can communicate with one another without a central instance in the form of a server being present in the data network.
  • the resources present on the individual network nodes are published by appropriate mechanisms, so that the resources are known to at least some of the network computers or can be searched for these resources.
  • Resources here are any information in the data network to understand.
  • the resources may in particular relate to information about stored data contents in the network or they may relate to tasks to be performed in the network by services which are published by a network node. Examples of services are undesirables be ⁇ calculations, such as a prime factorization, or any data processing, such as processing a video data stream, in particular the coding and the utmost flexibility of the video stream.
  • Res ⁇ resources can be registered in a decentralized data network so that the resources can be searched for by the network nodes.
  • references [1] and [2] describe a Resource Management Framework (RMF) resource management system which represents a layer attached on a peer-to-peer data network.
  • RMF Resource Management Framework
  • This RMF Concept provides information in the form of RMF resources as metadata in the network.
  • the information may also contain information about a service to be executed in the data network.
  • the object of the invention is therefore to provide a method for executing a service in a decentralized data network, which ensures a stable and efficient execution of the service in this data network.
  • a task to be performed by a service to be executed is published in the data network, wherein the publication can be carried out using any method known from the prior art.
  • the task may be published in the form of an above-mentioned RMF resource.
  • state parameters are assigned to the network nodes of the data network, and after the publication of the task to be performed by the service, each network node of at least a part of the network nodes of the data network determines whether it can execute the service with its status parameters.
  • the network nodes that can run the service then publish their own parameters in the data network, forming a group associated with the service, the group comprising the network nodes that can run the service.
  • At least one network node of the group is determined as a coordinating network node, wherein any selection method, in particular also selection methods from the prior art, can be used to determine a network node.
  • a coordinating network node can optionally be selected according to a random ⁇ process.
  • the at least one coordinating network node is used for the Ver ⁇ administration of the execution of the service, wherein the coordinating node in response to the service and the horrorspa- rametern the network nodes of the group defines at least one network node of the group as executive node. This executing network node then executes the service.
  • Data network is initialized and carried out. This is because ⁇ achieved by that network node, which can cause a service from ⁇ , are combined into a group, then a suitable network node is selected in the group with the aid of a coordinating network node that can execute the service.
  • a service here can be described in any way in the data ⁇ net.
  • a service contains parameters for characterizing the service and / or quality parameters to be maintained when the service is executed.
  • Such parameters may be, for example, a minimum number and / or maximum number of network nodes in the group associated with the service.
  • the quality parameters may also comprise a time interval, in the event that a node does not report any activity to the coordi ⁇ alternating network node within the time interval, an error in the execution of the service is detected. In this way it is ensured ⁇ that an executed service is always done with a set by a user minimum quality.
  • the state parameters of a network node in the data network can affect any properties of the network node, in particular the processing power of the network node and / or the Prozes ⁇ sorauslastung and / or the available memory and / or stored on the network node content.
  • network nodes are informed of a new task to be performed by a service or of other network nodes using state parameters with the aid of a subscription service.
  • subscription services are well known in the art. It is a kind of subscription to certain content, with a new publication of the subscribed content of the corresponding network node that has subscribed to this content is informed about this.
  • the network node of a group according to an evaluation criterion which depends on the state parameters of the respective gene network node and the service evaluated as to how well it for the execution of the of the service to erledi ⁇ constricting task are are suitable, wherein the evaluation vorzugswei ⁇ se is performed by the coordinating network node.
  • one network node of the group is defined as an exporting network node, which is in accordance with the evaluation criterion, the most suitable for the Erledi ⁇ supply of tasks to be by the service object network nodes.
  • the network nodes used are those network nodes which, after the executing network node, are the network nodes which are best suited for the task to be performed by the service. It should also be ensured that the monitoring network nodes than the at least one executive Netzkno ⁇ th and / or coordinating network nodes are other network nodes.
  • the monitoring by the monitoring network nodes proceeds in such a way that the at least one executive network node sends messages to the network node (s) at intervals, whereby if the messages fail, an error occurs during the execution of the service the or the monitoring network node is determined.
  • monitoring network nodes are advantageous, but not mandatory. If monitoring network nodes are used, a maximum number of monitoring network nodes for a group is preferably determined with the aid of the quality parameters of the service.
  • a service according to the OF INVENTION ⁇ dung at least one performer and a coordinating node and optionally also monitored network node BETEI ⁇ ligt.
  • all other network nodes ie the network nodes, which are not executing, coordinating or monitoring network nodes, are classified as inactive network nodes, which are not currently actively involved in the performance of the service.
  • These network nodes are ne waiting list set, and may optionally be later than a ⁇ direct time, for example upon failure of a network node or the change of the status parameter of the network nodes of the group, are used.
  • a migration i.e., a transmission of a service running on an executing network node to another network node is then initiated if there is a corresponding migration criterion, the migration being initiated by the at least one coordinating network node.
  • the migration criterion is in particular at Eintre ⁇ th one or more of the following events: - an error in the execution of the service is determined by at least one monitoring network node; - The at least one coordinating network node indicates
  • this is a fault-related and therefore essential migration of the service to a new executing network node.
  • this is a migration according to other criteria or depending on external influences, the cause of which is not an error.
  • the stop signal which is outputted in the latter case, for example, can then be generated when Festge ⁇ provides is that a different network node than a performing network node is suitable due to its stand parameters ⁇ better by a predetermined level at least, by the service to complete task at wherein the suitability for execution of tasks to be by the service object is preferably determined by means of the above-described Bewer ⁇ tung measure.
  • a stop signal may be output also when it is determined that preparespa ⁇ parameters of the service performed no longer be observed.
  • the inventive method is preferably such ⁇ out forms that is transmitted when an error occurs during the execution of the service of the service on at least one of the monitored network node. This migration is particularly easy because the monitoring nodes are already active when the service is running.
  • intermediate results of a service rule being executed in zeitli ⁇ chen intervals cached are intermediate results of a service rule being executed in zeitli ⁇ chen intervals cached.
  • ⁇ resumption of a service is facilitated to another computer according to its migration.
  • a service which has not yet been completed can be resumed with the aid of the last buffered intermediate results.
  • the at least one executing network node and the at least one coordinating network node coincide. Furthermore, in a particularly preferred variant, a single network node is determined as a coordinating network node, whereby the coordination of the service is made particularly simple. Similarly, the at least one coordinating network node also specifies only a single network node as the executing network node, which simplifies the execution of the service.
  • the published state parameters are updated by the network nodes of the group belonging to the service at intervals during the performance of a service. If this serious changes occur, such as the migration described above can be triggered to another leading from ⁇ network nodes.
  • the method according to the invention is preferably used in a peer-to-peer data network, the network nodes being peers of the data network.
  • the data network is employed in an ⁇ be Sonder preferred embodiment, the initially mentioned RMF network.
  • the invention further relates to a decentralized data network with a plurality of network nodes, wherein the data network is designed such that the inventive method is feasible.
  • Fig. 1 is a schematic representation of a decentralized data network in which the method according to the invention is used;
  • FIG. 2 shows a flowchart which illustrates the steps carried out in the method according to the invention.
  • RMF Resource Management Framework
  • RMF resources which describe the meta-information about these services.
  • ServiceTaskRessource two general resources used to describe services. These two resources are on the one hand the so-called ServiceTaskRessource and on the other hand a state resource in the form of so-called checkpoints.
  • a service task resource in this case to the alder by a service-damaging ⁇ object within the meaning of the claims and the check represents ⁇ points represent the last state of a service and correspond substantially between the stored intermediate results of the service within the meaning of the claims.
  • Task Service Resources are adaptable to the different Anforde ⁇ approximations of the task of the corresponding service.
  • a service that charges a prime number decomposition has at ⁇ game, other input parameters as services, which deal with the processing of video streams.
  • the prime factorization of the input parameters is the number to be broken down, whereas the input parameters be ⁇ relate to the location of the video stream and further information pens of the video stream in a video stream.
  • QoS Quality of Service
  • QoS parameters can specify, for example, the minimum number of peers or the maximum number of peers which may be used in the formation of the group in the sense of the claims.
  • the parameters may relate to a so-called heartbeat interval, which determines at which intervals a coordinating network node within the meaning of the claims sends messages to monitoring network nodes in the sense of the claims.
  • Fig. 1 shows a peer-to-peer network, with 10 network nodes Pl to PlO are shown.
  • Service Task resource in the peer-to-peer network published are light.
  • Individual peers can be informed about the presence of a specific ServiceTask resource, for example by means of a subcription mechanism.
  • individual peers subscribe to a specific service ⁇ task resource.
  • ServiceTaskResource is pub- lished, the peers who have subscribed to this ServiceTaskRes ⁇ source are informed about the corresponding parameters of the ServiceTaskResource.
  • the CandidatePeerRessources entspre ⁇ chen the status parameters of the network nodes within the meaning of the claims.
  • the CandidatePeerRessources contain information about the CPU load and the available space of the corresponding peer as well as information about certain preferences of the user of the corresponding peer, about the binary codes of services available on the peer, about the last update of the CandidatePeerResource in the network and the services in which the peer participates.
  • the candidate peer resource is used to determine if a peer is capable of performing a service to be performed and if a peer may be considered a potential candidate for migrating (ie, transmitting) a service from another peer ,
  • the parameters of the given task service resource with the parameters of CandidatePeerRessource be compared locally in a ⁇ individual peers.
  • the CandidatePeerRessources be periodically aktua ⁇ larra so used in the comparison whenever possible, the current status of the peer.
  • the comparison of its Candi- datePeerResource with the published ServiceTaskResource takes place locally. If the comparison shows that the peer is capable of executing the corresponding service, the corresponding peer registers by publishing its candidate peer resource in the data network, thereby informing it that it wants to be involved in the execution of the service. The registration takes place within one Regist ⁇ r istsperiode, and after this period, an initial service group is formed, all of which during the
  • Registration period includes registered peers.
  • a group is exemplified as G.
  • the group comprises here the peers Pl to P6.
  • Peers Pl to P6 are selected as so-called co-ordinator or coor ⁇ render peer. Any algorithms known from the prior art can be used to determine a peer as coordinator. In particular, known election methods can be used. Optionally, a coordinating peer may also be randomly selected. In the example of FIG. 1, the coordinator of the peer is Pl. The task of the coordinator is the administration of the group G. The coordinator determines which peers are provided for the execution of the service and which peers the execution of the
  • Monitor service S The peers which the services exporting ⁇ reindeer, here are the executive node within the meaning of the claims and are explained in more detail below.
  • the peers which monitor the execution are the monitoring network nodes in the sense of the claims. These network nodes are also referred to as so-called watchdogs and also explained in more detail below.
  • the invention exis- advantage common to each formed group G only a Coordina ⁇ gate, wherein a group can not exist without a coordinator.
  • coordinators may be the same peers, who You can also do a service.
  • the executing peer and the coordinator represent the same peer, this has the advantage that the coordinator is monitored and protected against failure by the supervising watchdogs. This is because the watchdogs monitor the entire peer, not just the role of the executing peer. Thus detect the watchdog Aussoci ⁇ le executive peers and a coordinating peers if both rollers are running on the same computer.
  • the group is analyzed by the coordinator based on the state parameters of the individual peers of the group. Each peer is evaluated using predefined procedures, these procedures being used in the service
  • the ServiceTaskResource contains an evaluation measure in the form of QoS definitions, which determine how individual states of the peer are to be classified in terms of how well the peer is suitable for performing the task to be performed by the service.
  • the most suitable for the completion of the task peer is automatic ⁇ table executive peer.
  • the role of executive and coordinating peer coincides. In the case that the initial coordinator is not rated as the peer, which is the most suitable peer for completing the task, Therefore, the peer immediately relinquishes his role as coordinator, and the coordinator's duties, together with the role of the executive peer, are taken over by the peer who has been deemed to be the most appropriate peer for the task.
  • Coordinating peers also take account of metrics, which are defined in the Service Task resource, in particular a Re ⁇ dundanz composition and the maximum tolerated period of time before it should be noted a failure of a peer in a group.
  • the redundancy factor specifies the number of watchdogs which monitor the executing peer and also serve as replacement peers in the event the executing peer fails.
  • the redundancy factor sets a minimum number and a maximum number of peers in a group.
  • a group does not necessarily include surveil ⁇ sponding peers here, but may on the other hand sev- eral this monitoring peers include.
  • the coordinating peer After the coordinating peer evaluated the other peers of the group according to pre-admit ⁇ ner, contained in the service criteria, it determines the executing peer and set in the redundancy factor number of watchdogs. All other network nodes that exceed the number of watchdogs are placed on a waiting list in the group. If the case occurring defects ⁇ th that fewer nodes than the minimum number of network nodes according to the redundancy factor is present, the group is considered incomplete.
  • the coordinators are also responsible for managing the groups. If a peer is no longer able to comply with a quality parameter that is specified, for example, in a service level agreement, the service is migrated to another peer.
  • the coordinator is thus able to Variegated ⁇ stakes in the provided performance of peers It can therefore trigger service migrations based on a service level agreement. It is thus possible that services will be migrated based on predetermined Krite ⁇ rien, result in an automatic load balancing between peers is achieved, so that services are always running on peers that provide the most resources.
  • peers are put on a waiting list when the set number of watchdogs is reached. These peers on the waiting list are also known as so-called idle ⁇ peers.
  • the state parameters of these peers are not good enough to qualify as executing peers or watchdogs. They are not actively participating in the service execution, but are registered as potential candidates who can become active participants in a group as they improve their state parameters or if the state parameters of other peers in the group deteriorate such that an idle peer becomes a peer having better state parameters than a peer currently in the role of watchdog or executive peer.
  • peers are defined by the coor ⁇ leaders Peer Peer an exporting and optionally monitored on.
  • Peer or as monitoring peer takes place depending on a corresponding evaluation measure, whereby the best rated peer as executive peer and then the peers are classified in descending order of their evaluation measure as watchdogs. .
  • ge ⁇ Telss results in the following sequence evaluation criterion of the evaluation:
  • the peer Pl which is the coordinator and the executing peer, has the best evaluation criterion.
  • the second best rating is the peer P4, the third best the peer P2, the fourth best the peer P5, the fifth best the peer P6 and the sixth best the peer P3.
  • Peers then the P2 to P6 conces- their roles in response to the loading ⁇ êts ser grasslands.
  • the peers P4, P2 and P5 are successively classified as watchdogs.
  • the role assignment performed by the peer P1 is illustrated in FIG. 1 by corresponding arrows from the peer Pl to the other peers P2 to P6 in the group G.
  • the tasks of the executing peer and the watchdogs within the group G belonging to the service S are explained below.
  • the role of the executing peer is to actually execute the corresponding service set in the ServiceTaskResource.
  • a peer of the data network may only once assume the role of the executing peer for a service, but multiple services may be performed concurrently in the data network.
  • a peer is assigned the role of the executive peer by the coordinator. Any peer that is a member of a group can potentially be an executive peer, and the co-ordinator decides which of the most appropriate peers is for the role of the executing peer.
  • Executing peers play an important role in the migration process described below, in which a service is transferred from one executing peer to another peer.
  • the migration ensures that the state parameters of an executing peer are always such that the executing peer is suitable for carrying out the service. It is thus necessary for the executing peer to be constantly monitored to detect deviations within a period of time to thereby ensure seamless continuation of a service.
  • the monitoring of the executing peer is achieved here by the transmission of so-called heartbeat messages.
  • Heartbeat messages A simple way to tell other peers that a peer is still active and not down. However, sending heartbeat messages from each peer in the data network creates a large load, thereby consuming network bandwidth. To the
  • the watchdogs take over the task of receiving such messages. In this way, the watchdog obtain the responsibility for the detection of failures, a failure is detected when a predetermined time span ⁇ no heartbeat message is received from the exporting more peer. That is, in the absence of heartbeat messages longer than a predetermined threshold, an alarm is issued and a corresponding process for processing the failure initiated.
  • the associated with a service group must not necessarily ⁇ as watchdogs.
  • a watchdog (if any) may not be the same peer that is currently running the service. According to this restric ⁇ effect it is possible that more and detects a failure of executive peers in the presence of a watchdog advertising the can.
  • the watchdogs In addition to their monitoring task, the watchdogs also have the task of replacing a failed executing peer. According to the evaluation of the watchdogs made by the coordinator, the watchdogs are numbered, ie the peer with the highest score after the executing peer becomes the watchdog with the number 1, the peer with the second best rating measure the watchdog with the number 2, etc. Each peer in A group knows the ratings within the group and thus knows which peer is the top rated watchdog. In this way it is possible to react very quickly to the failure of an executing peer. Of the Best Rated Watchdog automatically migrated this service to itself, without that there is a loop for geeig ⁇ designated candidates. This feature allows for quick real-time migration of similar services, so the time is minimized between a failure and a resumption of a diene ⁇ tes.
  • the service migration is self-organizing and does not require a central instance to initialize the process.
  • the process of migration comprises five phases, namely an initialization phase, a service intake phase, a service execution phase, the actual migration phase and a termination phase.
  • the initialization phase is shown in FIG. 2 as IP gekennzeich ⁇ net and comprises in the previous explained Re ⁇ tration of a service in the RMF network via a service task resource that all the information about the service in ⁇ klusive of QoS requirements contains.
  • the ServiceTaskResource is thus published on the network, whereupon peers indicate their interest in participating in the service by updating their candidate resource.
  • the initial, associated with the service group is ge ⁇ forms, which is indicated in Fig. 2 by the step Sl.
  • an initial coordinating peer is determined, as previously explained.
  • the coordinating peer then evaluates, according to an evaluation measure, all peers of the group according to their state parameters in the candidate resource and in accordance with the service's QoS requirements (step S2 in Fig. 2).
  • the peer with the best score according to the score becomes the executing peer and the other peers become the watchdogs in descending order of their scores. If a fixed number of participating watchdogs reach ⁇ to, all subsequent peers than the idle peers are switched on stepped. Each peer will be notified by the coordinator about his / her role in the group.
  • the initialization phase is triggered according to the invention by registering a ServiceTaskResource in the RMF network, wherein the registering of resources can be performed by any peer connecting to the RMF network.
  • Peers are informed mechanisms through a newly registered in the network service in particular by Subkriptions-mechanics by the peers ptions corresponding RMF Subkri- for a particular type of disturbance in the network regist ⁇ Center. Upon the occurrence of a corresponding event of the registered type, the peer will be notified thereof by a message.
  • step S3 the steps S3 to S5 which correspond to the recording phase RP of the service follow.
  • step S3 it is checked in step S3 whether the service currently being considered has already been partially executed before. If this is the case, the state of the service is restored ⁇ (step S4). If this is not the case, the status of the service is reinitialized (step S5). Subsequently ⁇ wd the execution phase of the service, which is denoted by EP follows.
  • the actual EP exporting ⁇ tion of the service by the executing peer and the monitoring of the service by the watchdog is performed.
  • the watchdog check on ⁇ already mentioned heartbeat messages that are sent by the exporting peer, and she nalysis a- whether the time has exceeded a predetermined period of time since reaching the last heartbeat message.
  • the frequency of heartbeat messages depends on the type of service. This frequency is a QoS criterion that is set for the service along with other QoS criteria. Thus, a suitable Kri ⁇ criterion can be defined for each service.
  • Enable heartbeat messages but not the distinction between types of failure (eg power failure, computer failure, service failure). In the vorlie ⁇ constricting embodiment, these types of failures are generalized, and it is assumed that all types of failures have the same effect on the migration process.
  • the service execution phase In addition to monitoring the executing peer with regard to its failure, the service execution phase also detects whether there are signals for the "anticipated triggering" of migrations.
  • the coordinator can record such anticipated triggering signals, for example via a subcription mechanism, informing him of changes in the peer candidate candidates. For example, if a new peer is added to the network, or does the state change
  • the coordinating peer informed by RMF subscription mechanisms of this and can respond to these changes in that it re-evaluates the peers.
  • the CPU load is an example only and is not usually a suffi ⁇ and fair criterion for a service is migrated from one leading out ⁇ peer to another. If it were always integrated to the peer with the lowest CPU utilization, this would lead to jitter between the peers.
  • step S6 OWNER ⁇ Liche service is running.
  • step S7 is the PERIODIC ⁇ ge checking the state of the service, ie S7 advertising to update the state parameters of the individual peers and the current execution state of the service and be- censored.
  • step S8 relates to the continuous monitoring of the executing peer by the watchdogs.
  • step S7 If there are no trigger signals for stopping the service in step S7, the process returns to step S6 and the service continues to be executed. If the coordinating Peer depending ⁇ but due to a caused by the update of the service state release signal, the instruction being ⁇ ben that the service is to be stopped, the system goes in the migration phase, indicated with MP.
  • the migration phase a service is now transferred from one executing peer to another executing peer.
  • the current state of the service can be saved again in the form of intermediate results before the service is migrated. This is not possible with a migration to ⁇ ground of an error.
  • the migration phase is illustrated by steps S9 and SLO. in the
  • Step S9 the currently performed service is first stopped , the step S9 being triggered by a corresponding signal from the coordinating peer. It is so ⁇ with anticipated to triggering a migration. On the other hand, the service in the branch Bl becomes due to a
  • step S also includes the subsequent execution of the migration. Stopping the service and monitoring activities results in the service freezing when a service is transferred to another peer. The coordinator ensures this freezing. After the service has been migrated, the coordinator checks if the service has been previously completed or if it has not been completed (step Sil). If the service has not pot ⁇ closed, re-evaluate the composition of the belonging to the service group is performed. This is shown in Fig.
  • the termination phase CP follows.
  • the service being executed indicates the completion of the service to the coordinating peer, who then checks to see if the state of the service matches the service's QoS specifications. If this Be ⁇ conditions are satisfied, the coordinating peer freezes all participating peers in the group, then a message is output which indicates the end of the service execution. The peers then stop their activities and ignore any further messages related to this service.
  • all RMF resources that are no longer needed are removed. These resources are, in particular, the subscriptions that have been made in relation to the service as well as no longer relevant state resources. The resource which speaks ent ⁇ the final state of the service is not removed. In this way, the request is taken into account that the end result of the service can still be used even if the execution of the service has already ended.
  • the failure or interruption of the execution of a service is detected by the watchdogs.
  • the watchdogs analyze the frequency of the heartbeat messages that are sent out by the executing peer, whereby the migration of the service is triggered if certain criteria are met.
  • the concept of heartbeat messages is not the only concept that can detect failures. However, it is a very simple concept.
  • a disadvantage here is that not all kinds of errors with the help of Heartbeat messages can be identified.
  • no distinction is made between different types of errors network errors, service errors and computer errors. Consequently, the heartbeat messages are described in this execution ⁇ form simple and short messages which have almost no content. They are sent only for the purpose of informing other peers that the peer sending the message is still running.
  • the executing peer (s) periodically send these messages to predetermined watchdogs, the transmission frequency being specified as QoS parameters in the service.
  • the system may also deal with delays, that is, with messages that are delayed due to congestion in the network.
  • security periods are set, within which the messages to be sent can be received with a certain delay time. If these safety periods are exceeded, ie if no message is still received within the safety period, an alarm is finally issued by the corresponding watchdog and the migration of the service to another peer is triggered.
  • the heartbeat messages are always only from the executing peer to the
  • Watchdogs sent This has the advantage that the network is not overloaded with the heartbeat messages. However, it may also be possible that heartbeat messages are also sent by the watchdogs to other peers, so that the watchdogs can also be monitored with regard to their failures and appropriate measures taken in the event of failure of watchdogs.
  • the resource management framework A system for managing metadata in decentralized networks using peer-to-peer technology.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)
  • Computer And Data Communications (AREA)

Abstract

L'invention concerne un procédé pour exécuter un service (S) dans un réseau de données décentralisé comprenant une pluralité de noeuds de réseau (P1,..., P10). Selon ce procédé : a) une tâche à effectuer par un service (S) à exécuter est publiée dans le réseau de données; b) des paramètres d'état sont respectivement associés aux noeuds de réseau (P1,..., P10), et chaque noeud de réseau d'au moins une partie de la pluralité de noeuds de réseau (P1,..., P10) détermine si ses paramètres d'état lui permettent d'exécuter ledit service (S); c) les noeuds de réseau (P1,..., P10) pouvant exécuter le service (S) publient leurs paramètres d'état dans le réseau de données, ce qui crée un groupe (G) lié audit service, ce groupe comprenant les noeuds de réseau (P1,..., P10) pouvant exécuter le service (S); d) au moins un noeud de réseau (P1,..., P10) du groupe (G) est défini comme étant un noeud de réseau de coordination (P1); e) le(s) noeud(s) de réseau (P1) de coordination désigne(nt) au moins un noeud de réseau (P1,..., P10) du groupe (G) comme étant le noeud de réseau (P1) exécutant, en fonction du service (S) et des paramètres d'état des noeuds de réseau (P1,..., P10) du groupe (G); f) le(s) noeud(s) de réseau (P1) exécutants exécute(nt) ledit service (S).
PCT/EP2007/057609 2006-07-25 2007-07-24 Procédé pour exécuter un service dans un réseau de données décentralisé WO2008012301A2 (fr)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
EP06015487.9 2006-07-25
EP06015487 2006-07-25
DE102006054090.5 2006-11-16
DE102006054090A DE102006054090A1 (de) 2006-07-25 2006-11-16 Verfahren zum Ausführen eines Dienstes in einem dezentralen Datennetz

Publications (2)

Publication Number Publication Date
WO2008012301A2 true WO2008012301A2 (fr) 2008-01-31
WO2008012301A3 WO2008012301A3 (fr) 2008-05-22

Family

ID=38981822

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2007/057609 WO2008012301A2 (fr) 2006-07-25 2007-07-24 Procédé pour exécuter un service dans un réseau de données décentralisé

Country Status (1)

Country Link
WO (1) WO2008012301A2 (fr)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2178267A1 (fr) * 2008-10-20 2010-04-21 Siemens Aktiengesellschaft Procédé d'exécution de services dans un réseau de données décentralisé
RU2510156C2 (ru) * 2008-06-04 2014-03-20 Конинклейке Филипс Электроникс Н.В. Способ создания беспроводной многоскачковой сети

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020007468A1 (en) * 2000-05-02 2002-01-17 Sun Microsystems, Inc. Method and system for achieving high availability in a networked computer system
US20050114448A1 (en) * 2003-11-03 2005-05-26 Apacheta Corporation System and method for delegation of data processing tasks based on device physical attributes and spatial behavior
EP1536612A1 (fr) * 2003-11-27 2005-06-01 Samsung Electronics Co., Ltd. Procédé et dispositifs pour le partage d'applications utilisant un protocole peer-to-peer

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020007468A1 (en) * 2000-05-02 2002-01-17 Sun Microsystems, Inc. Method and system for achieving high availability in a networked computer system
US20050114448A1 (en) * 2003-11-03 2005-05-26 Apacheta Corporation System and method for delegation of data processing tasks based on device physical attributes and spatial behavior
EP1536612A1 (fr) * 2003-11-27 2005-06-01 Samsung Electronics Co., Ltd. Procédé et dispositifs pour le partage d'applications utilisant un protocole peer-to-peer

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
RU2510156C2 (ru) * 2008-06-04 2014-03-20 Конинклейке Филипс Электроникс Н.В. Способ создания беспроводной многоскачковой сети
EP2178267A1 (fr) * 2008-10-20 2010-04-21 Siemens Aktiengesellschaft Procédé d'exécution de services dans un réseau de données décentralisé

Also Published As

Publication number Publication date
WO2008012301A3 (fr) 2008-05-22

Similar Documents

Publication Publication Date Title
EP0635784B1 (fr) Système multiprocesseur
DE60207251T2 (de) Verfahren zur sicherstellung des betriebs eines gruppierten nachrichtenvergebenden servers während knotenausfällen und netzwerkzuteilungen
DE60215002T2 (de) Verfahren und system für effiziente verteilung von netzwerk-ereignisdaten
DE10134492B4 (de) Ausfallübernahme des Dateimanagementsystems in einem Rechnercluster
DE69811148T2 (de) Mitgliedschaft in einem unzuverlässigen verteilten Rechnersystem
DE60220263T2 (de) Server-duplexverfahren und geduplextes serversystem
DE60220418T2 (de) Verfahren und Anbieter zur Systemsynchronisation
DE112012002631T5 (de) Stream-Verarbeitung unter Verwendung einer Client-Server-Architektur
DE102006054090A1 (de) Verfahren zum Ausführen eines Dienstes in einem dezentralen Datennetz
WO2008071448A1 (fr) Système informatisé d'analyse, de gestion, de maîtrise, d'exploitation et de surveillance d'une architecture complexe matérielle/logicielle
DE10123067A1 (de) Synchrone Vervielfältigung von Transaktionen in einem verteilten System
DE102011018878B3 (de) Verfahren zum Synchronisieren der Datenbestände von Datenbanken
DE112005001995B4 (de) Computeranordnung und Verfahren zum Anbieten von Diensten für Benutzer über ein Netzwerk
DE112010004530T5 (de) Transaktionsaktualisierung bei Dynamischen Verteilten Arbeitslasten
DE102016219854A1 (de) Computersystem und Verfahren zum dynamischen Anpassen eines software-definierten Netzwerks
WO2008012301A2 (fr) Procédé pour exécuter un service dans un réseau de données décentralisé
EP1798892A1 (fr) Procédé de chargement d'une liste d'alarmes par une application de gestion d'alarmes
EP0433350A1 (fr) Programme d'exploitation d'un ordinateur
EP1619849B1 (fr) Procédé de synchronisation d'un système distribué
EP1647898A1 (fr) Système de replication de base de données sans serveur
EP1195678B1 (fr) Méthode d'opération d'un système de traitement de données avec une unité de traitement redondante
EP1790120A1 (fr) Procede, programme informatique comprenant des moyens de codage de programme et produit de programme informatique pour surveiller l'etat d'un systeme comprenant des composantes reparties, en particulier l'etat d'un reseau comprenant des composantes reparties, et reseau a composantes reparties
WO2009087056A1 (fr) Procédé de gestion de processus de calcul dans un réseau de données décentralisé
DE10123822A1 (de) Einrichtung zur Verwaltung einer Dienstverbindung zwischen einem Clientprozess mit einer Single-Thread-Bibliothek und einem Serverprozess
EP2178267B1 (fr) Procédé d'exécution de services dans un réseau de données décentralisé

Legal Events

Date Code Title Description
NENP Non-entry into the national phase

Ref country code: DE

NENP Non-entry into the national phase

Ref country code: RU

122 Ep: pct application non-entry in european phase

Ref document number: 07787849

Country of ref document: EP

Kind code of ref document: A2