WO2012109946A1 - 一种大规模网络的数据采集方法和网络节点 - Google Patents

一种大规模网络的数据采集方法和网络节点 Download PDF

Info

Publication number
WO2012109946A1
WO2012109946A1 PCT/CN2011/085057 CN2011085057W WO2012109946A1 WO 2012109946 A1 WO2012109946 A1 WO 2012109946A1 CN 2011085057 W CN2011085057 W CN 2011085057W WO 2012109946 A1 WO2012109946 A1 WO 2012109946A1
Authority
WO
WIPO (PCT)
Prior art keywords
node
nodes
network
resource occupancy
occupancy rate
Prior art date
Application number
PCT/CN2011/085057
Other languages
English (en)
French (fr)
Inventor
胡斐然
赵伟
Original Assignee
华为技术有限公司
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
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Priority to CN201180004193.2A priority Critical patent/CN102652425B/zh
Priority to PCT/CN2011/085057 priority patent/WO2012109946A1/zh
Publication of WO2012109946A1 publication Critical patent/WO2012109946A1/zh

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L43/00Arrangements for monitoring or testing data switching networks
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00Routing or path finding of packets in data switching networks
    • H04L45/02Topology update or discovery

Definitions

  • Embodiments of the present invention relate to cloud computing technologies, and in particular, to a data collection method and a network node of a large-scale network.
  • BACKGROUND With the rapid development of cloud computing, the scale of network architectures is also increasing. Therefore, in large-scale networks, statistics on various information data of various network nodes face enormous challenges. In the traditional method, statistical indicators such as the global resource occupancy rate of the entire network are often obtained through post-analysis.
  • the post-analysis method is used to collect statistics on large-scale resource indicators, generally adopting a centralized management mode, and dividing a data warehouse (ie, a management node) and a resource node, wherein the resource nodes are responsible for collecting and transmitting respective resource occupancy rates, and managing
  • the node is responsible for the collection and storage of global resource-related data of the entire network;
  • the post-analysis system is a specialized data mining and analysis system connected with the management node, and performs data mining analysis on the global resource-related data collected and stored by the management node, and finally outputs the data.
  • the results of the analysis are passed to the decision system as the basis for decision making.
  • post-analysis based on this architecture takes a long time and often takes hours or even days to get the final analysis.
  • Embodiments of the present invention provide a data acquisition method and a network node for a large-scale network, which are used to solve The problem that the statistical indicator data such as the global resource occupancy rate cannot be obtained in real time in a large-scale network is determined.
  • the embodiment of the present invention provides a data collection method for a large-scale network, including: the management side arbitrarily selects a node in the network as a start node, and sends a network resource occupancy rate statistics request to the start node. The node range of the other nodes in the network, requesting the starting node to perform statistics on the network resource occupancy rate;
  • the originating node divides other nodes in the network into N groups according to the received network resource occupancy rate statistical request and the node range of other nodes in the network, and randomly selects one node from each group.
  • the network resource occupancy statistics request and the node range of other nodes in the group are separately sent to each of the split start nodes;
  • Each of the sub-starting nodes determines, according to the received network resource occupancy rate statistical request and the node range of other nodes in the group, whether the number of other nodes in the group is greater than N, and the number of other nodes in the group. If the value is less than or equal to N, the group is not further grouped, and the other nodes in the group are the final nodes, and the network resource occupancy statistics request is separately sent to the final node; if the number of other nodes in the group is greater than N, the other nodes in the group are further divided into N groups, and one node is randomly selected from each group as the next layer starting node, and the network resources are respectively sent to each of the next layer starting nodes. Occupancy statistics request and node range of other nodes in the group;
  • Each of the final nodes collects the network resource occupancy rate of the final node according to the received network resource occupancy rate statistics request, and reports the statistical result to the branch node to which the final node belongs;
  • each branching node After receiving the statistics of the network resource occupancy rate reported by the node under its jurisdiction, each branching node performs the calculation and statistics again according to the statistics of the network resource occupancy rate of the starting node, and reports the statistical result to the starting point.
  • the upper layer to which the node belongs is divided into the starting node, and is reported to the starting node layer by layer, and the final network resource occupancy rate is calculated by the starting node;
  • the initial node reports the calculated final network resource occupancy rate to the management terminal.
  • the embodiment of the invention further provides a network node, including:
  • a receiving unit configured to receive a network resource occupancy rate request and a node range of other nodes in the set to which the node belongs, where the network resource occupancy rate statistics request includes a source IP address of the request and a destination IP address of the request, The start time and end time of the statistics and the resource metric name of the request statistic, the node range includes a set consisting of node identifiers or a set consisting of IP addresses of nodes;
  • a determining a grouping unit configured to determine, according to the network resource occupancy rate statistical request and a node range of other nodes in the set to which the node belongs, whether the number of other nodes in the set is greater than N, if the number of other nodes in the set is less than or equal to N, the group is not further grouped, and the network resource occupancy statistics request is separately sent to other nodes in the set; if the number of other nodes in the set is greater than N, the other nodes in the set are continued Dividing into N groups, arbitrarily selecting one node from each group, and separately sending the network resource occupancy rate statistics request and the node range of other nodes in the group to the node;
  • the resource occupancy statistics unit is configured to perform statistical reporting on the network resource occupancy rate of the node, and after receiving the statistical result of the network resource occupancy rate reported by the ruled node, perform calculation and statistics again according to the following calculation formula, and obtain the node to the network.
  • the statistical result of the resource occupancy rate is reported, and the statistical result is reported: (the resource occupancy rate of the node + the sum of the resource occupancy rates of the nodes under the jurisdiction) I (1 + the number of nodes under its jurisdiction).
  • the data collection method of the large-scale network provided by the embodiment of the present invention organizes the nodes of the entire network according to the N-tree structure, and adopts a recursive manner for the information statistics of the nodes, so that the statistical algorithm does not need to be modified.
  • Other business logic can perform real-time statistics on large-scale network node monitoring data, which greatly improves the statistical speed of indicators such as global resource occupancy rate of the entire network, thereby meeting the requirements of timely response to application scenarios such as resource scheduling, and also The maintenance cost of large-scale network nodes is greatly reduced.
  • FIG. 1 is a schematic diagram of a large-scale cloud computing network applied according to an embodiment of the present invention
  • FIG. 2 is a schematic diagram of organizing resource nodes of a whole network according to a tree hierarchical structure according to an embodiment of the present invention
  • FIG. 3 is a schematic flowchart of recursive grouping of nodes in a whole network according to an embodiment of the present invention
  • FIG. 4 is a schematic flowchart of a method for collecting data on a large-scale network according to an embodiment of the present invention
  • FIG. 5 is a schematic diagram of message interaction of a data collection method for a large-scale network according to an embodiment of the present invention
  • FIG. 6 is a schematic structural diagram of a network node according to an embodiment of the present invention. detailed description
  • FIG. 1 is a schematic diagram of a large-scale cloud computing network.
  • a node is randomly selected from the network as the entry point for the statistical operation, and the computing node reports the statistics of the resource occupancy rate of the entire network to the management layer.
  • the management uses this result to determine whether it is necessary to expand or reduce the capacity of the network.
  • the computing node in FIG. 1 may also be referred to as a resource node or a network node, which is not limited in this embodiment of the present invention.
  • the method logic of the embodiment of the present invention is as follows: 1.
  • the resource nodes ie, compute nodes
  • the status of each layer is equivalent, and the information statistics of the nodes are recursive, so that the network can be implemented.
  • the scale is getting bigger and bigger, but only need to increase the number of levels without modifying the statistical algorithm and other business logic; and each resource node is a computing node, which consists of two components.
  • the task of one component is to complete the relationship with other nodes.
  • Information synchronization, the task of another component is to perform recursive statistical reporting of indicators such as resource occupancy between nodes.
  • each compute node maintains all compute node information, including but not limited to: the heartbeat of the node, the state of the node (failure check / live / dead ), the node's current load.
  • the management terminal can initiate a statistical request for statistical indicators such as the resource occupancy rate of the entire network from any network resource node (decentralized computing node), then the selected starting node will group all other nodes, and the grouping method Grouping methods include, but are not limited to, binary trees; the starting node will first divide the other nodes into n groups, pick a distribution point from each group, and distribute statistical requests to these distribution points. The distribution point recursively divides the computing nodes under its jurisdiction into n components, until it is not subdivided.
  • the entire network node grouping logic is shown in Figure 3.
  • the computing node receiving the request starts to perform statistical calculation of indicators such as its own resource occupancy rate, and waits for the calculation result of the computing node under its jurisdiction. After the statistical results of all the computing nodes of the jurisdiction are reported, The computing node performs the summary statistics according to the following formula: (the resource occupancy rate of the node + the sum of the resource occupancy rates of the nodes under the jurisdiction) / (1 + the number of nodes under its jurisdiction); and then reports the statistical summary result to the distribution of the upper layer
  • the node, and so on, the statistical result of the statistical indicators such as the resource occupancy rate of the entire network will be summarized to the node originally requested by the management terminal, thereby completing the resource statistics operation of the entire network.
  • FIG. 4 is a schematic flowchart of a data collection method for a large-scale network according to an embodiment of the present invention.
  • the data collection method of the large-scale network in this embodiment may include the following steps:
  • the S10CK management end arbitrarily selects one node in the network as a starting node, and sends the network resource occupancy rate to the starting node.
  • the network is a decentralized node information synchronization network, and each node in the network maintains information of all nodes.
  • the network resource occupancy statistics request includes: a source IP address of the initiating request (such as an IP address of the management end) and a destination IP address of the receiving request (such as a node IP address randomly selected by the management terminal as a starting node), and statistics The start time and end time and the resource metric name for the request statistics.
  • the resource metric is a metric that measures resource usage, such as CPU utilization, memory usage, and network outflow.
  • the node range includes: a set consisting of node identifiers or a set consisting of node IP addresses. It should be noted that the range of the node may be carried in the network resource usage statistics request, or may be separately sent, which is not limited in this embodiment of the present invention.
  • the starting node divides other nodes in the network into N groups according to the received network resource occupancy rate statistics request and the node range of other nodes in the network, and arbitrarily selects each group.
  • a node is used as a split start node, and the network resource occupancy statistics request and the node range of other nodes in the group are respectively sent to each of the split start nodes;
  • the node range of other nodes in the network is a set of node identifiers or IP addresses of all other nodes except the start node in the network; and the node ranges of other nodes in the group are in the group. A collection of node identities or IP addresses of all other nodes except the starting node.
  • the other nodes in the network are divided into N groups, which may be halved or unequal, and preferably N is equal to 2, but in practical applications, it may be selected according to specific situations, and this embodiment of the present invention does not Make a limit.
  • nodes there are 99 nodes (identified as 1 -99) in the whole network, except for the starting node (for example, 1), and 98 nodes (2-99).
  • N 2, then it can be divided into 2 groups, and then select one starting node from each group (such as 2 and 51 respectively), then the range of the first group is (3-50), The range of the two groups is (52-99); but if N is 10, then it can be divided into 10 groups, and then one sub-start node is selected from each group (such as 2, 12, 22, 32, 42, 52 respectively).
  • the range of the first group is 3-11
  • the range of the second group is 13-21
  • the range of the third group to the tenth group is 23-31, 33-41 43-51, 53-61, 63-71, 73-81, and 93-99.
  • Each of the sub-starting nodes determines, according to the received network resource occupancy rate statistical request and the node range of other nodes in the group, whether the number of other nodes in the group is greater than N, if other nodes in the group If the number of the nodes is less than or equal to N, the group is not further grouped, and the other nodes in the group are the final nodes, and the network resource occupancy statistics request is separately sent to the final node; if other nodes in the group are If the number is greater than N, the other nodes in the group are further divided into N groups, and one node is randomly selected from each group as the next layer starting node, and the next node is sent to each of the next layer.
  • Network resource occupancy statistics request and node range of other nodes in the group
  • N For example, suppose the network has 99 nodes, N is 10, then 10 groups, it should be 9 groups of 10 nodes, a group of 8 nodes, because you want to select a root node (that is, the starting node).
  • the root node sends a statistical request to 10 sub-root nodes, the scope of the node to which each sub-root node belongs is attached.
  • the sub-root node finds that the number of its own nodes cannot be divided into statistical groups, it directly issues the information to the subordinate node.
  • the scope of the affiliation node is no longer attached to the statistic request. After each node receives the statistic request, if it finds that there is no affiliation node range in the request, then it knows that it is a leaf node (ie, the final node). Report the statistics of your own resource usage directly.
  • each final node performs statistics on the network resource occupancy rate of the final node according to the received network resource occupancy rate statistical request, and reports the statistical result to the branching node to which the final node belongs; S150. After receiving the statistics of the network resource occupancy rate reported by the node under the control, each branching node performs the calculation and statistics again according to the statistics of the network resource occupancy rate of the starting node, and reports the statistical result to the branch.
  • the upper layer to which the start node belongs is divided into the start node, and is reported to the start node layer by layer, and the final network resource occupancy rate is calculated by the start node;
  • the step is statistically reported in a recursive manner until all statistical results are summarized to the starting node.
  • the starting node and the starting node are all calculated according to the following formula:
  • the starting node reports the calculated final network resource occupancy rate to the management terminal.
  • the data collection method of the large-scale network provided by the embodiment of the present invention is performed according to the nodes of the entire network.
  • the N-tree structure is organized, and the information statistics of the nodes are recursively, so that the statistical algorithms and other business logics can be modified, and real-time statistics of large-scale network node monitoring data can be performed, which greatly improves the global resources of the entire network.
  • the statistical speed of indicators such as occupancy rate, so as to meet the requirements of timely application scenarios such as resource scheduling, and greatly reduce the maintenance cost of large-scale network nodes.
  • FIG. 5 is a schematic diagram of message interaction of a data collection method for a large-scale network according to an embodiment of the present invention.
  • the management end is the management layer of the whole network; the starting node is a computing node in the entire network randomly selected by the management layer; the governing node is the starting node grouping all the nodes of the whole network, and after being divided into N groups, a computing node randomly selected from each group (ie, the starting node in the first embodiment); an n-th node under the jurisdiction, which is a node where the governing node performs n times of similar starting nodes for the nodes in the group (equivalent In the first embodiment, the nth layer is divided into starting nodes); and the final node is the node after the grouping is no longer possible.
  • the data collection method of the large-scale network in this embodiment may include the following steps:
  • the source IP address of the network-wide resource occupancy statistics request message is the IP address of the management end
  • the destination IP address is the IP address of the computing node randomly selected from the current network.
  • the statistical request message also includes statistics. The start time and the statistics end time, as well as the resource metric name of the request statistics, such as CPU utilization, memory usage, network outflow, and so on.
  • the starting node sends the quorum resource occupancy statistics request and the affiliation node range to the quorum node;
  • the source IP address in the quorum resource occupancy rate request message is the IP address of the starting node
  • the destination IP address is a group of nodes other than the starting node in the whole network, and each group is divided into N groups.
  • the request in the same is the same, but it also needs to be accompanied by the scope of the node to which each starting node belongs.
  • the affiliation node sends the quorum resource occupancy rate request and the ruling node range to the admin node n layer node;
  • the quorum resource occupancy rate request initiated by the affiliation node and the source IP address of the message of the affiliation node range are the IP address of each sub-start node, and the destination IP address is the starting node.
  • the nodes under the jurisdiction are divided into N groups again, and the IP address of one computing node is randomly selected from each group; the node will be the starting node of the next layer, and other computing nodes in the group will be used as the computing node.
  • the governing node; the statistical request part is the same as the request in 201, but it also needs to be accompanied by the scope of the node to which each starting node belongs.
  • the n-th node under the jurisdiction sends a statistical resource occupancy rate request to the final node;
  • the jurisdiction resource occupancy statistics request initiated by the subordinate n-th node is similar to 203, but does not need to be accompanied by the scope of the node to which each subordinate n-node belongs.
  • the final node returns a resource occupancy rate request response to the subordinate n-th node.
  • the source IP address in the resource occupancy statistics request response sent by the final node is the IP address of the final node
  • the destination IP address is the IP address of the starting node of the upper layer
  • the statistical request response includes the local The metric name of the resource, such as CPU utilization, memory usage, network outflow, etc., and the specific values corresponding to these metric names.
  • the n-layer node receives the sum of the resource occupancy rate of each node and the resource occupancy rate, and divides the total number of nodes under the jurisdiction by 1 to obtain the statistical result of the average resource occupancy rate of the node under the layer;
  • the address is the IP address of the starting node of the layer
  • the destination IP is the IP address of the starting node of the upper layer (that is, the source address from the statistical request);
  • the average resource occupancy rate of the governing node includes the resource.
  • Metric names such as CPU utilization, memory usage, network outflows, etc., and the specific statistical averages for these metric names.
  • the affiliation node reports the statistics of the resource occupancy rate of the layer to the starting node.
  • the average resource occupancy statistics result of the tier node of the tier node sent by the appointing node is similar to 207.
  • the starting node adds the total resource occupancy rate of each layer node to its own resource occupancy rate, and divides the total number of nodes under the jurisdiction by one to obtain the statistical result of the average network resource occupancy rate.
  • the starting node initiates
  • the source IP address of the network resource occupancy rate request response is the IP address of the initiating node
  • the destination IP address is the IP address of the management end.
  • the average resource occupancy rate of the entire network includes the metric name of the resource, such as CPU utilization. , memory usage, network outflows, etc., and the specific statistical averages for these metric names.
  • the nodes of the entire network are organized according to the N-tree structure, and the information statistics of the nodes are recursively, so that no repair is needed.
  • real-time statistics can be collected on the monitoring data of large-scale network nodes, which greatly improves the statistical speed of indicators such as the global resource occupancy rate of the entire network, so as to meet the requirements of timely application scenarios such as resource scheduling.
  • the maintenance cost of large-scale network nodes is greatly reduced.
  • FIG. 6 is a schematic structural diagram of a network node according to an embodiment of the present invention.
  • the network node 30 provided by the embodiment of the present invention may include a receiving unit 31, a determining grouping unit 32, and a resource occupancy statistics unit 33.
  • the receiving unit 31 is configured to receive a network resource occupancy rate statistics request and a node range of other nodes in the set to which the local node belongs, where the network resource occupancy rate statistics request includes a source IP address of the request and a destination IP address of the request.
  • the address, the start time and the end time of the statistics, and the resource metric name of the request statistic, the node range includes a set consisting of the node identifier or a set consisting of the IP addresses of the nodes; the determining grouping unit 32 is configured to use the network resource according to the network resource.
  • the occupancy statistics request and the node range of other nodes in the set to which the node belongs determine whether the number of other nodes in the set is greater than N. If the number of other nodes in the set is less than or equal to N, the group is not further grouped.
  • the resource occupancy rate statistics unit 33 is configured to perform statistical reporting on the network resource occupancy rate of the node, and receive the network resource occupancy rate statistical result reported by the ruled node.
  • network node provided by the embodiment of the present invention may further include:
  • the node information synchronization unit 34 is configured to synchronize the node information of the entire network, so that the node maintains the network. Information about all nodes in .
  • the network node organizes the nodes of the entire network according to the N-tree structure, and adopts a recursive manner for the information statistics of the nodes, so that the statistical algorithm and other business logics can be modified without requiring modification of the statistical algorithm and other business logics.
  • the network node monitors the data for real-time statistics, which greatly improves the statistical speed of the global resource occupancy rate and other indicators of the entire network, thereby meeting the requirements of timely response to application scenarios such as resource scheduling, and greatly reducing the maintenance of large-scale network nodes. cost.
  • the receiving unit 31, the determining grouping unit 32, the resource occupancy rate counting unit 33, and the node information synchronizing unit 34 in the third embodiment are all hardware.

Landscapes

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

Description

一种大规模网络的数据釆集方法和网络节点
技术领域
本发明实施例涉及云计算技术, 特别涉及一种大规模网络的数据采集方 法和网络节点。 背景技术 随着云计算的蓬勃发展, 网络架构的规模也越来越大, 因此, 在大规模 网络中, 对于各个网络节点的各种信息数据的统计面临着巨大的挑战。 在传 统的方法中, 对于全网的全局资源占用率等统计指标往往都是通过后分析得 到。 现有技术中采用后分析方法对大规模的资源指标进行统计一般采取集中 管理模式, 划分数据仓库(即, 管理节点)和资源节点, 其中, 资源节点负 责各自资源占用率的采集和传送, 管理节点负责全网全局资源相关数据的收 集和存储; 后分析***是和管理节点连接的、 专门的数据挖掘分析***, 对 管理节点收集和存储的全网全局资源相关数据进行数据挖掘分析, 输出最终 的分析结果, 传递给决策***作为决策依据。 但是, 基于这种架构的后分析 方法耗时很长, 往往需要几小时甚至数天才能得到最终的分析结果。
而在诸如资源调度等需要实时得到统计数据的场景下, 得到的数据越实 时, 效果才越好。 因此, 这种后分析方法在资源调度等场景下, 就显得没有 意义了, 并且可能还是不可用的。 这就亟需一种方法, 来解决大规模网络中, 对全局资源占用率等统计指标数据能尽快获取, 最好是能实时获取。 发明内容
本发明实施例提供一种大规模网络的数据采集方法和网络节点, 用以解 决大规模网络中对于全局资源占用率等统计指标数据无法实时获取的问题。 有鉴于此, 本发明实施例提供了一种大规模网络的数据采集方法, 包括: 管理端任意选取网络中的一个节点作为起始节点, 向所述起始节点发送 网络资源占用率统计请求和所述网络中其他节点的节点范围 , 请求所述起始 节点对网络资源占用率进行统计;
所述起始节点根据接收到的所述网络资源占用率统计请求和所述网络中 其他节点的节点范围, 将所述网络中的其他节点分为 N组, 从每一组中任意 选取一个节点作为分起始节点, 向每一个分起始节点分别发送所述网络资源 占用率统计请求和该组中其他节点的节点范围;
所述每一个分起始节点根据接收到的所述网络资源占用率统计请求和该 组中其他节点的节点范围, 判断该组中其他节点的数目是否大于 N , 如果该 组中其他节点的数目小于或者等于 N , 则不再对该组继续进行分组, 该组中 的其他节点为最终节点, 向所述最终节点分别发送所述网络资源占用率统计 请求; 如果该组中其他节点的数目大于 N , 则将该组中的其他节点继续分为 N 组, 从每一组中任意选取一个节点作为下一层分起始节点, 向每一个下一 层分起始节点分别发送所述网络资源占用率统计请求和该组中其他节点的节 点范围;
重复上述判断步骤 , 直至所有分组中除分起始节点外的其他节点的数目 小于或者等于 N;
每一个最终节点根据收到的网络资源占用率统计请求, 对该最终节点的 网络资源占用率进行统计, 并将统计结果上报给该最终节点所属的分起始节 点;
每一个分起始节点收到所辖节点上报的网络资源占用率统计结果后, 结 合该分起始节点的网络资源占用率统计结果, 再次进行计算统计, 并将统计 结果上报给该分起始节点所属的上一层分起始节点, 直至逐层上报给所述起 始节点, 由所述起始节点计算得出最终的网络资源占用率; 所述起始节点将计算得出的最终网络资源占用率上报给所述管理端。 本发明实施例还提供了一种网络节点, 包括:
接收单元, 用于接收网络资源占用率统计请求和本节点所属集合中其他 节点的节点范围, 其中, 所述网络资源占用率统计请求中包括发起请求的源 IP地址和接收请求的目的 IP地址、统计的起始时间和结束时间以及请求统计 的资源度量名称, 所述节点范围包括由节点标识组成的集合或者由节点的 IP 地址组成的集合;
判断分组单元, 用于根据所述网络资源占用率统计请求和本节点所属集 合中其他节点的节点范围, 判断该集合中其他节点的数目是否大于 N , 如果 该集合中其他节点的数目小于或者等于 N , 则不再对该集合继续进行分组, 向该集合中的其他节点分别发送所述网络资源占用率统计请求; 如果该集合 中其他节点的数目大于 N , 则将该集合中的其他节点继续分为 N组, 从每一 组中任意选取一个节点, 向所述节点分别发送所述网络资源占用率统计请求 和该组中其他节点的节点范围;
资源占用率统计单元, 用于对本节点的网络资源占用率进行统计上报, 以及当收到所辖节点上报的网络资源占用率统计结果后, 按照如下计算公式 再次进行计算统计, 得到本节点对网络资源占用率的统计结果, 并上报该统 计结果: (本节点资源占用率 +所辖节点资源占用率之和 ) I ( 1 +所辖节点数)。
由上述技术方案可知, 本发明实施例提供的大规模网络的数据采集方法 通过对全网的节点按照 N叉树结构进行组织, 并对节点的信息统计采用递归 的方式, 使得无需修改统计算法和其他业务逻辑, 即可对大规模网络节点监 控数据进行实时统计, 极大提高了整个网络的全局资源占用率等指标的统计 速度, 从而达到及时响应诸如资源调度等应用场景的要求, 同时, 也大大降 低了大规模网络节点的维护成本。 附图说明 为了更清楚地说明本发明实施例或现有技术中的技术方案, 下面将对实 施例或现有技术描述中所需要使用的附图作简单地介绍, 显而易见地, 下面 描述中的附图仅仅是本发明的一些实施例, 对于本领域普通技术人员来讲, 在不付出创造性劳动的前提下, 还可以根据这些附图获得其他的附图。
图 1为本发明实施例应用的大规摸云计算网络示意图;
图 2为依据本发明实施例对全网的资源节点按照树状层次结构进行组织的 示意图;
图 3为依据本发明实施例对全网节点进行递归分组的流程示意图; 图 4 为本发明实施例提供的一种大规模网络的数据采集方法的流程示意 图;
图 5为本发明实施例提供的一种大规模网络的数据采集方法的消息交互示 意图;
图 6为本发明实施例提供的一种网络节点的结构示意图。 具体实施方式
下面将结合本发明实施例中的附图, 对本发明实施例中的技术方案进行 清楚、 完整地描述, 显然, 所描述的实施例仅仅是本发明一部分实施例, 而 不是全部的实施例。 基于本发明中的实施例, 本领域普通技术人员在没有做 出创造性劳动前提下所获得的所有其他实施例, 都属于本发明保护的范围。
本发明实施例应用的***环境如图 1所示。 图 1是一个大规模云计算网 络的示意图。 在图 1 中, 管理层面需要统计整个网络的资源占用率情况时, 从网络中任意选取一个节点作为统计操作的切入点, 并由此计算节点向管理 层面汇报整个网络的资源占用率情况的统计结果, 管理层以此结果来决定是 否需要对网络进行扩容或者减容的操作。 其中, 图 1 中的计算节点也可称之 为资源节点或者网络节点, 本发明实施例对此不做限定。
基于图 1所示的***环境, 本发明实施例的方法逻辑如下: 1、 如图 2所示, 将全网的资源节点(即计算节点)按照树状层次结构进 行组织, 但是每层的地位等同, 对节点的信息统计采用递归的方式, 这样可 以做到, 网络规模越来越大, 但只需要增加层级数量, 而无需修改统计算法 和其他业务逻辑; 并且每个资源节点即计算节点, 有两个组件组成, 一个组 件的任务是完成同其他节点之间的信息同步, 另外一个组件的任务是完成节 点间资源占用率等指标的递归统计上报。
2、 为了简化树状层级资源的维护, 所有资源节点被组成为一个去中心化 的计算节点信息同步的集体(如图 1所示), 其中, 信息同步算法包括但不限 于 Gossip等方法。每个计算节点都保持有所有计算节点信息, 这些信息包括 但不限于: 节点的心跳、 节点的状态 (失效检查 / live / dead )、 节点当前负 载。
3、 管理端可以从任意一个网络资源节点(去中心化计算节点)发起对全 网资源占用率等统计指标的统计请求, 那么这个被选中的起始节点将对其他 所有节点进行分组, 分组方法包括但不限于二叉树等分组方法; 起始节点首 先将把其他节点分为 n组, 从每组中挑选一个分发点, 将统计请求分发给这 些分发点。 分发点再递归的将自己所辖计算节点继续分为 n组分发, 直到不 够再细分为止, 整个网络节点分组划分逻辑如图 3所示。
4、分发的同时,接收到请求的计算节点就开始执行自身资源占用率等指 标的统计计算, 并等待其辖属计算节点的计算结果, 等所有辖属计算节点的 统计结果都汇报完成之后, 该计算节点按照如下公式再做一次汇总统计: (本 节点资源占用率 +所辖节点资源占用率之和) / (1 + 所辖节点数); 然后将统计 汇总结果上报到上一层的分发节点, 依此类推, 全网的资源占用率等统计指 标的统计最终结果将汇总到管理端最初请求的节点, 从而完成了整个网络的 资源统计操作。
实施例一
图 4 为本发明实施例提供的一种大规模网络的数据采集方法的流程示意 图, 如图 4所示, 本实施例的大规模网络的数据采集方法可以包括以下步骤: S10CK 管理端任意选取网络中的一个节点作为起始节点, 向所述起始节 点发送网络资源占用率统计请求和所述网络中其他节点的节点范围, 请求所 述起始节点对网络资源占用率进行统计;
具体地, 所述网络为去中心化的节点信息同步的网络, 该网络中的每个 节点都保持有所有节点的信息。 所述网络资源占用率统计请求中包括: 发起 请求的源 IP地址 (如管理端的 IP地址 )和接收请求的目的 IP地址 (如管理 端随机选取作为起始节点的节点 IP地址)、 统计的起始时间和结束时间以及 请求统计的资源度量名称。 其中, 所述的资源度量, 即衡量资源使用情况的 指标, 如 CPU利用率、 内存使用率、 网络流出量等。 所述节点范围, 包括: 由节点标识组成的集合或者由节点的 IP地址组成的集合。 需要说明的是, 所 述节点范围既可以携带在网络资源占用率统计请求中下发,也可以单独下发, 本发明实施例对此不做限定。
S110、 所述起始节点根据接收到的所述网络资源占用率统计请求和所述 网络中其他节点的节点范围, 将所述网络中的其他节点分为 N组, 从每一组 中任意选取一个节点作为分起始节点, 向每一个分起始节点分别发送所述网 络资源占用率统计请求和该组中其他节点的节点范围;
需要说明的是, 所述网络中其他节点的节点范围就是该网络中除了起始 节点外, 由所有其他节点的节点标识或者 IP地址构成的集合; 而分组中其他 节点的节点范围就是该组中除了分起始节点外, 由所有其他节点的节点标识 或者 IP地址构成的集合。 另外, 将所述网络中的其他节点分为 N组, 既可以 等分, 也可以不等分, 优选 N等于 2, 但是在实际应用中, 可根据具体情况 选择, 本发明实施例对此不做限定。
例如, 全网有 99个节点(标识为 1 -99 ), 除起始节点(例如选为 1 )夕卜, 还有 98个节点 (2-99 ), 如果 N为 2, 那么就可以划分为 2组, 再从每组中 选一个分起始节点(如分别选 2和 51 ), 那么第一组的范围就是(3-50 ), 第 二组的范围就是(52-99 ); 但是如果 N为 10, 那么可以划分为 10组, 再从 每组中选一个分起始节点(如分别选 2、 12、 22、 32、 42、 52、 62、 72、 82、 92 ), 那么第一组的范围就为 3-11、 第二组的范围为 13-21、 第三组至第十组 的范围分别为 23-31、 33-41、 43-51、 53-61、 63-71、 73-81和 93-99。
S120、 所述每一个分起始节点根据接收到的所述网络资源占用率统计请 求和该组中其他节点的节点范围, 判断该组中其他节点的数目是否大于 N , 如果该组中其他节点的数目小于或者等于 N , 则不再对该组继续进行分组, 该组中的其他节点为最终节点, 向所述最终节点分别发送所述网络资源占用 率统计请求; 如果该组中其他节点的数目大于 N , 则将该组中的其他节点继 续分为 N组, 从每一组中任意选取一个节点作为下一层分起始节点, 向每一 个下一层分起始节点分别发送所述网络资源占用率统计请求和该组中其他节 点的节点范围;
例如, 假设全网有 99个节点, N为 10, 那么就是 10组, 应该是 9组 10个节点组成, 一组 8个节点, 因为要选择一个根节点 (即, 起始节点)。 另外, 10组中, 每组中还有一个子根节点 (即, 分起始节点), 其中 9个子 根节点各自辖属 9个节点, 1个子根节点辖属 7个节点。根节点给 10个子根 节点下发统计请求时, 附带每个子根节点所辖属节点的范围, 当子根节点发 现自己辖属节点数无法再划分统计组时, 就直接给辖属节点下发统计请求, 该统计请求中不再附带辖属节点的范围, 每个节点收到统计请求后, 如果发 现请求中没有辖属节点范围, 那么就知道自己为叶子节点 (即, 最终节点) 了, 直接上报自己的资源占用率统计情况。
S130、 重复上述判断步骤, 直至所有分组中除分起始节点外的其他节点 的数目小于或者等于 N;
S140、 每一个最终节点根据收到的网络资源占用率统计请求, 对该最终 节点的网络资源占用率进行统计, 并将统计结果上报给该最终节点所属的分 起始节点; S150、 每一个分起始节点收到所辖节点上报的网络资源占用率统计结果 后, 结合该分起始节点的网络资源占用率统计结果, 再次进行计算统计, 并 将统计结果上报给该分起始节点所属的上一层分起始节点, 直至逐层上报给 所述起始节点 , 由所述起始节点计算得出最终的网络资源占用率;
具体地, 该步骤采用递归的方式统计上报, 直至所有的统计结果汇总到 起始节点为止。 其中, 所述起始节点和分起始节点都是按照以下公式进行计 算的:
(本节点资源占用率 +所辖节点资源占用率之和) / (1 + 所辖节点数) S160、 所述起始节点将计算得出的最终网络资源占用率上报给所述管理 端。
本发明实施例提供的大规模网络的数据采集方法通过对全网的节点按照
N 叉树结构进行组织, 并对节点的信息统计采用递归的方式, 使得无需修改 统计算法和其他业务逻辑, 即可对大规模网络节点监控数据进行实时统计, 极大提高了整个网络的全局资源占用率等指标的统计速度, 从而达到及时响 应诸如资源调度等应用场景的要求, 同时, 也大大降低了大规模网络节点的 维护成本。 实施例二
图 5为本发明实施例提供的一种大规模网络的数据采集方法的消息交互示 意图。 其中, 管理端是全网的管理层面; 起始节点, 是管理层面任意挑选的 全网络中的一个计算节点; 辖属节点是起始节点对全网所有节点进行分组 , 划分为 N组后, 从每组中随意挑选的一个计算节点 (即实施例一中的分起始 节点); 辖属 n层节点,是辖属节点对组内节点进行 n次类似起始节点的分组 的节点(相当于实施例一中第 n层的分起始节点); 而最终节点, 则是无法再 进行分组后的节点。 如图 5所示, 本实施例的大规模网络的数据采集方法可 以包括以下步骤: 该步骤中, 该全网资源占用率统计请求消息中源 IP地址是管理端的 IP地 址, 目的 IP地址为从当前网络中随机选出的计算节点的 IP地址, 统计请求的 消息中, 还包含统计的起始时间和统计结束时间, 以及请求统计的资源度量 名称, 如 CPU利用率、 内存使用率、 网络流出量等。
202、起始节点向辖属节点发送辖属资源占用率统计请求以及辖属节点范 围;
该步骤中, 该辖属资源占用率统计请求消息中的源 IP地址为起始节点的 IP地址, 而目的 IP地址为将全网除起始节点之外的节点划分 N组后,从每组中 随机挑选出的一个计算节点的 I P地址; 该节点将作为下一层的分起始节点, 而该组中的其他计算节点将作为该计算节点的辖属节点; 其中的统计请求部 分同 201中的请求相同, 不过还需要附带有每个分起始节点所辖属节点的范 围。
203、 辖属节点向辖属 n层节点发送辖属资源占用率统计请求以及辖属节 点范围;
该步骤中 , 辖属节点所发起的辖属资源占用率统计请求以及辖属节点范 围的消息的源 IP地址为每个分起始节点的 IP地址, 而目的 IP地址则是将分起 始节点所辖属的节点再次划分为 N组,从每组中随机挑选出一个计算节点的 I P 地址; 该节点将作为下一层的分起始节点, 而该组中其他计算节点将作为该 计算节点的辖属节点; 其中的统计请求部分同 201中的请求相同, 不过还需要 附带有每个分起始节点所辖属节点的范围。
204、 辖属 n层节点向最终节点发送辖属资源占用率统计请求;
该步骤中, 辖属 n层节点所发起的辖属资源占用率统计请求同 203类似, 不过不需要附带有每个辖属 n层节点所辖属节点的范围。
205、 最终节点向辖属 n层节点返回资源占用率统计请求响应; 该步骤中, 最终节点所发的资源占用率统计请求响应中的源 IP地址为最 终节点的 IP地址, 而目的 IP地址为自己上一层分起始节点的 IP地址, 统计请 求响应中包括本地资源的度量名称, 如 CPU利用率、 内存使用率、 网络流出 量等, 以及这些度量名称所对应的具体数值。
206、 辖属 n层节点将收到的各个节点资源占用率总和加上自身资源占用 率, 除以辖属总节点数加 1 , 得到该层辖属节点平均的资源占用率统计结果;
207、 辖属 n层节点向辖属节点上 该层辖属资源占用率统计结果; 该步骤中,辖属 n层节点发送的该层辖属节点平均的资源占用率统计结果 响应中的源 IP地址就是该层分起始节点的 IP地址, 目的 IP为上一层分起始节 点的 IP地址(即, 来自统计请求中的源地址); 辖属节点平均的资源占用率统 计结果中包括资源的度量名称, 如 CPU利用率、 内存使用率、 网络流出量等, 以及这些度量名称所对应的具体统计平均值。
208、 辖属节点将收到的各层节点资源占用率总和加上自身资源占用率, 除以辖属总节点数加 1, 得到该层辖属节点平均的资源占用率统计结果;
209、 辖属节点向起始节点上报该层辖属资源占用率统计结果; 该步骤中, 辖属节点发送的该层辖属节点平均的资源占用率统计结果同 207相似。
210、起始节点将各层节点资源占用率总和加上自身资源占用率, 除以辖 属总节点数加 1 , 得到全网平均的资源占用率统计结果; 该步骤中,起始节点发起的全网资源占用率统计请求响应中的源 IP地址 就是起始节点的 IP地址, 目的 IP地址则是管理端的 IP地址, 全网平均的资 源占用率统计结果包括资源的度量名称, 如 CPU利用率、 内存使用率、 网络 流出量等, 以及这些度量名称所对应的具体统计平均值。
根据本实施例提供的大规模网络的数据采集方法, 通过对全网的节点按 照 N叉树结构进行组织, 并对节点的信息统计采用递归的方式, 使得无需修 改统计算法和其他业务逻辑,即可对大规模网络节点监控数据进行实时统计, 极大提高了整个网络的全局资源占用率等指标的统计速度, 从而达到及时响 应诸如资源调度等应用场景的要求, 同时, 也大大降低了大规模网络节点的 维护成本。 实施例三
本发明实施例还提供了一种应用于上述方法实施例中的网络节点设备。 图 6为本发明实施例提供的一种网络节点的结构示意图, 如图 6所示, 本发 明实施例提供的网络节点 30可以包括接收单元 31、判断分组单元 32和资源 占用率统计单元 33。 其中, 接收单元 31用于接收网络资源占用率统计请求 和本节点所属集合中其他节点的节点范围, 其中, 所述网络资源占用率统计 请求中包括发起请求的源 IP地址和接收请求的目的 IP地址、 统计的起始时 间和结束时间以及请求统计的资源度量名称, 所述节点范围包括由节点标识 组成的集合或者由节点的 IP地址组成的集合; 判断分组单元 32用于根据所 述网络资源占用率统计请求和本节点所属集合中其他节点的节点范围, 判断 该集合中其他节点的数目是否大于 N , 如果该集合中其他节点的数目小于或 者等于 N, 则不再对该集合继续进行分组, 向该集合中的其他节点分别发送 所述网络资源占用率统计请求; 如果该集合中其他节点的数目大于 N , 则将 该集合中的其他节点继续分为 N组, 从每一组中任意选取一个节点, 向所述 节点分别发送所述网络资源占用率统计请求和该组中其他节点的节点范围; 资源占用率统计单元 33, 用于对本节点的网络资源占用率进行统计上报, 以 及当收到所辖节点上报的网络资源占用率统计结果后, 按照如下计算公式再 次进行计算统计, 得到本节点对网络资源占用率的统计结果, 并上报该统计 结果: (本节点资源占用率 +所辖节点资源占用率之和) I ( 1 +所辖节点数)。
进一步, 本发明实施例提供的网络节点还可以包括:
节点信息同步单元 34, 用于全网节点信息的同步, 使本节点保持有网络 中所有节点的信息。
根据本发明实施例提供的网络节点, 通过对全网的节点按照 N叉树结构 进行组织, 并对节点的信息统计采用递归的方式, 使得无需修改统计算法和 其他业务逻辑, 即可对大规模网络节点监控数据进行实时统计, 极大提高了 整个网络的全局资源占用率等指标的统计速度, 从而达到及时响应诸如资源 调度等应用场景的要求, 同时, 也大大降低了大规模网络节点的维护成本。
需要说明的是: 实施例三中的接收单元 31、 判断分组单元 32、 资源占 用率统计单元 33和节点信息同步单元 34均为硬件。
本领域普通技术人员可以理解: 实现上述方法实施例的全部或部分步骤 可以通过程序指令相关的硬件来完成, 前述的程序可以存储于一计算机可读 取存储介质中, 该程序在执行时, 执行包括上述方法实施例的步骤; 而前述 的存储介质包括: ROM、 RAM , 磁碟或者光盘等各种可以存储程序代码的介 最后应说明的是: 以上实施例仅用以说明本发明的技术方案, 而非对其 限制; 尽管参照前述实施例对本发明进行了详细的说明, 本领域的普通技术 人员应当理解: 其依然可以对前述各实施例所记载的技术方案进行修改, 或 者对其中部分技术特征进行等同替换; 而这些修改或者替换, 并不使相应技 术方案的本质脱离本发明各实施例技术方案的精神和范围。

Claims

权 利 要求
1、 一种大规模网络的数据采集方法, 其特征在于, 包括:
管理端任意选取网络中的一个节点作为起始节点, 向所述起始节点发送 网络资源占用率统计请求和所述网络中其他节点的节点范围 , 请求所述起始 节点对网络资源占用率进行统计;
所述起始节点根据接收到的所述网络资源占用率统计请求和所述网络中 其他节点的节点范围, 将所述网络中的其他节点分为 N组, 从每一组中任意 选取一个节点作为分起始节点, 向每一个分起始节点分别发送所述网络资源 占用率统计请求和该组中其他节点的节点范围;
所述每一个分起始节点根据接收到的所述网络资源占用率统计请求和该 组中其他节点的节点范围, 判断该组中其他节点的数目是否大于 N , 如果该 组中其他节点的数目小于或者等于 N , 则不再对该组继续进行分组, 该组中 的其他节点为最终节点, 向所述最终节点分别发送所述网络资源占用率统计 请求; 如果该组中其他节点的数目大于 N , 则将该组中的其他节点继续分为 N 组, 从每一组中任意选取一个节点作为下一层分起始节点, 向每一个下一 层分起始节点分别发送所述网络资源占用率统计请求和该组中其他节点的节 点范围;
重复上述判断步骤 , 直至所有分组中除分起始节点外的其他节点的数目 小于或者等于 N;
每一个最终节点根据收到的网络资源占用率统计请求, 对该最终节点的 网络资源占用率进行统计, 并将统计结果上报给该最终节点所属的分起始节 点;
每一个分起始节点收到所辖节点上报的网络资源占用率统计结果后, 结 合该分起始节点的网络资源占用率统计结果, 再次进行计算统计, 并将统计 结果上报给该分起始节点所属的上一层分起始节点, 直至逐层上报给所述起 始节点, 由所述起始节点计算得出最终的网络资源占用率; 所述起始节点将计算得出的最终网络资源占用率上报给所述管理端。
2、 根据权利要求 1所述的方法, 其特征在于, 所述网络为去中心化的节 点信息同步的网络, 该网络中的每个节点都保持有所有节点的信息。
3、根据权利要求 1或 2所述的方法, 其特征在于, 所述网络资源占用率 统计请求中包括: 发起请求的源 IP地址和接收请求的目的 IP地址、 统计的 起始时间和结束时间以及请求统计的资源度量名称。
4、 根据权利要求 1或 2或 3所述的方法, 其特征在于, 所述节点范围, 包括: 由节点标识组成的集合或者由节点的 IP地址组成的集合。
5、根据权利要求 1至 4任一权利要求所述的方法, 其特征在于, 所述每 一层的分起始节点收到所辖节点上报的网络资源占用率统计结果后, 结合该 分起始节点的网络资源占用率统计结果, 再次进行计算统计, 包括:
所述每一层的分起始节点在下发所述网络资源占用率统计请求时, 同时 采集和计算本节点的网络资源占用率;
当收到所辖节点上报的网络资源占用率统计结果后, 按照如下计算公式 再次进行计算统计, 得到该分起始节点对网络资源占用率的统计结果:
(本节点资源占用率 +所辖节点资源占用率之和) I ( 1 +所辖节点数)。
6、根据权利要求 1至 5任一权利要求所述的方法, 其特征在于, 所述由 起始节点计算得出最终的网络资源占用率, 包括:
所述起始节点在下发所述网络资源占用率统计请求时, 同时采集和计算 本节点的网络资源占用率;
当收到所辖节点上报的网络资源占用率统计结果后, 按照如下计算公式 进行计算, 得到最终的网络资源占用率:
(本节点资源占用率 +所辖节点资源占用率之和) I ( 1 +所辖节点数)。
7、 一种网络节点, 其特征在于, 包括:
接收单元, 用于接收网络资源占用率统计请求和本节点所属集合中其他 节点的节点范围, 其中, 所述网络资源占用率统计请求中包括发起请求的源 IP地址和接收请求的目的 IP地址、统计的起始时间和结束时间以及请求统计 的资源度量名称, 所述节点范围包括由节点标识组成的集合或者由节点的 IP 地址组成的集合;
判断分组单元, 用于根据所述网络资源占用率统计请求和本节点所属集 合中其他节点的节点范围, 判断该集合中其他节点的数目是否大于 N , 如果 该集合中其他节点的数目小于或者等于 N , 则不再对该集合继续进行分组, 向该集合中的其他节点分别发送所述网络资源占用率统计请求; 如果该集合 中其他节点的数目大于 N , 则将该集合中的其他节点继续分为 N组, 从每一 组中任意选取一个节点, 向所述节点分别发送所述网络资源占用率统计请求 和该组中其他节点的节点范围;
资源占用率统计单元, 用于对本节点的网络资源占用率进行统计上报, 以及当收到所辖节点上报的网络资源占用率统计结果后, 按照如下计算公式 再次进行计算统计, 得到本节点对网络资源占用率的统计结果, 并上报该统 计结果: (本节点资源占用率 +所辖节点资源占用率之和 ) I ( 1 +所辖节点数)。
8、根据权利要求 7所述的网络节点,其特征在于,所述网络节点还包括: 节点信息同步单元, 用于全网节点信息的同步, 使本节点保持有网络中 所有节点的信息。
PCT/CN2011/085057 2011-12-30 2011-12-30 一种大规模网络的数据采集方法和网络节点 WO2012109946A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN201180004193.2A CN102652425B (zh) 2011-12-30 2011-12-30 一种大规模网络的数据采集方法和网络节点
PCT/CN2011/085057 WO2012109946A1 (zh) 2011-12-30 2011-12-30 一种大规模网络的数据采集方法和网络节点

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2011/085057 WO2012109946A1 (zh) 2011-12-30 2011-12-30 一种大规模网络的数据采集方法和网络节点

Publications (1)

Publication Number Publication Date
WO2012109946A1 true WO2012109946A1 (zh) 2012-08-23

Family

ID=46671944

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2011/085057 WO2012109946A1 (zh) 2011-12-30 2011-12-30 一种大规模网络的数据采集方法和网络节点

Country Status (2)

Country Link
CN (1) CN102652425B (zh)
WO (1) WO2012109946A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102882740A (zh) * 2012-09-13 2013-01-16 曙光信息产业(北京)有限公司 云环境下集中式配置监测方法
CN111600748B (zh) * 2020-04-28 2023-05-30 江苏方天电力技术有限公司 一种基于无线自组网网络的增码拓扑识别***及方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101667200A (zh) * 2009-09-18 2010-03-10 浙江大学 一种p2p环境中的窗口查询方法
TW201025919A (en) * 2008-12-18 2010-07-01 Univ Nat Taiwan Centralized balanced-tree algorithm and dynamic planning data transmission method for wireless sensor
CN101958805A (zh) * 2010-09-26 2011-01-26 中兴通讯股份有限公司 一种云计算中终端接入和管理的方法及***
CN102073695A (zh) * 2010-12-28 2011-05-25 汉柏科技有限公司 递进式统计方法

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8453156B2 (en) * 2009-03-30 2013-05-28 Intel Corporation Method and system to perform load balancing of a task-based multi-threaded application

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TW201025919A (en) * 2008-12-18 2010-07-01 Univ Nat Taiwan Centralized balanced-tree algorithm and dynamic planning data transmission method for wireless sensor
CN101667200A (zh) * 2009-09-18 2010-03-10 浙江大学 一种p2p环境中的窗口查询方法
CN101958805A (zh) * 2010-09-26 2011-01-26 中兴通讯股份有限公司 一种云计算中终端接入和管理的方法及***
CN102073695A (zh) * 2010-12-28 2011-05-25 汉柏科技有限公司 递进式统计方法

Also Published As

Publication number Publication date
CN102652425B (zh) 2014-06-25
CN102652425A (zh) 2012-08-29

Similar Documents

Publication Publication Date Title
US10757037B2 (en) Method and system for dynamic scheduling and allocation of network traffic
CN110191148B (zh) 一种面向边缘计算的统计函数分布式执行方法及***
WO2020258920A1 (zh) 一种网络切片资源管理方法及设备
Ioannidis et al. Optimal and scalable distribution of content updates over a mobile social network
CN101378409B (zh) 分层对等网络的最佳操作
EP3528440A1 (en) Path selection method and system, network acceleration node, and network acceleration system
CN107038071B (zh) 一种基于数据流预测的Storm任务伸缩调度算法
US20170339030A1 (en) Traffic Statistic Generation For Datacenters
US20030120666A1 (en) Real-time monitoring of service performance through the use of relational database calculation clusters
CN111966289B (zh) 基于Kafka集群的分区优化方法和***
CN107317707B (zh) 一种基于点覆盖集的sdn网络拓扑管理方法
CN104427625A (zh) 一种基于用户体验的网络资源调度方法及***
CN108021447B (zh) 一种基于分布式数据确定最优资源策略的方法及***
Kang et al. Application of adaptive load balancing algorithm based on minimum traffic in cloud computing architecture
CN114465945B (zh) 一种基于sdn的标识解析网络构建方法
WO2012109946A1 (zh) 一种大规模网络的数据采集方法和网络节点
CN107295103B (zh) 基于集中式p2p***的节点管理方法
Xu et al. Distributed decentralized collaborative monitoring architecture for cloud infrastructures
CN106790354B (zh) 一种防数据拥堵的通信方法及其装置
JP2013235515A (ja) データの分散管理システム及び装置及び方法及びプログラム
CN101989918A (zh) 对等网络管理***及其实现管理的方法
CN110099116B (zh) 一种基于大数据的子网安全性评估方法
CN107493333B (zh) 一种基于SDN的云计算环境下web***资源管理方法及***
CN112003900B (zh) 实现分布式***中高负载场景下服务高可用的方法、***
Li et al. Location-aware multi-user resource allocation in distributed clouds

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 201180004193.2

Country of ref document: CN

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 11858990

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 11858990

Country of ref document: EP

Kind code of ref document: A1