[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

WO2011144178A1 - 联合优化的方法、装置和系统 - Google Patents

联合优化的方法、装置和系统 Download PDF

Info

Publication number
WO2011144178A1
WO2011144178A1 PCT/CN2011/075388 CN2011075388W WO2011144178A1 WO 2011144178 A1 WO2011144178 A1 WO 2011144178A1 CN 2011075388 W CN2011075388 W CN 2011075388W WO 2011144178 A1 WO2011144178 A1 WO 2011144178A1
Authority
WO
WIPO (PCT)
Prior art keywords
bandwidth
subnet
server
information
external port
Prior art date
Application number
PCT/CN2011/075388
Other languages
English (en)
French (fr)
Inventor
张洪波
施广宇
文刘飞
徐向阳
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 EP11783095.0A priority Critical patent/EP2515478B1/en
Priority to RU2012137874/08A priority patent/RU2520354C2/ru
Priority to BR112012022286A priority patent/BR112012022286A2/pt
Publication of WO2011144178A1 publication Critical patent/WO2011144178A1/zh
Priority to US13/721,530 priority patent/US9003029B2/en

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L43/00Arrangements for monitoring or testing data switching networks
    • H04L43/08Monitoring or testing based on specific metrics, e.g. QoS, energy consumption or environmental parameters
    • H04L43/0876Network utilisation, e.g. volume of load or congestion level
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/08Configuration management of networks or network elements
    • H04L41/0896Bandwidth or capacity management, i.e. automatically increasing or decreasing capacities
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00Routing or path finding of packets in data switching networks
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00Network arrangements or protocols for supporting network services or applications
    • H04L67/50Network services
    • H04L67/60Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources
    • H04L67/63Routing a service request depending on the request content or context
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L41/00Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
    • H04L41/12Discovery or management of network topologies

Definitions

  • the present invention relates to the field of communications technologies, and in particular, to a method, apparatus, and system for joint optimization.
  • the role of the ISP Internet Service Provider
  • ISP Internet Service Provider
  • TE Traffic Engineering
  • CP Content Provider
  • SS Server Selection
  • the joint optimization mathematical model of TE and SS established by NBS is a constrained convex optimization problem.
  • the network-wide status information uses a centralized approach to obtain the optimal solution of the system, and applies the optimal solution to the network, thereby implementing a routing change strategy and a new server selection strategy.
  • the scalability of the solution is poor. As the network scale expands, the time spent collecting information and computing will increase dramatically. If joint optimization is to be applied to larger networks, such as the entire metropolitan area network or even the national backbone network, The real-time performance of the existing joint optimization method on the network will be difficult to guarantee.
  • the embodiments of the present invention provide a method, a device, and a system for jointly optimizing, which effectively reduce the computational complexity and complexity of the joint optimization, and improve the performance and efficiency of the entire network.
  • Joint optimization methods including: Obtaining link information, server and bandwidth information, and user requirement information of the subnet; the server and bandwidth information includes a virtual server bandwidth of each external port of the subnet, where the virtual server bandwidth is the subnet pass The bandwidth required by the external port to the server outside the subnet;
  • An embodiment of the present invention provides a joint optimization computing device, including:
  • a parameter collection module configured to obtain link information, server and bandwidth information, and user requirement information of the subnet;
  • the server and bandwidth information includes virtual server bandwidth of each external port of the subnet, and the virtual server The bandwidth is the bandwidth requirement of the subnet through the external port to the server outside the subnet;
  • the calculation module is configured to obtain an optimal routing parameter and a server selection parameter of the subnet according to link information, server and bandwidth information, and user requirement information, and obtain each subnet of the subnet according to an optimal routing parameter and a server selection parameter. Optimized ingress bandwidth of the port;
  • Output module used to compare the optimized ingress bandwidth and the virtual server bandwidth of each external port. If the comparison result of the optimized ingress bandwidth and the virtual server bandwidth of all the external ports is less than the setting error, the application is performed in the subnet. Optimal routing parameters and server selection parameters, and outputting the optimal routing parameters and server selection parameters.
  • the embodiment of the invention provides a communication network, including:
  • the joint optimization system obtains link information, server and bandwidth information, and user demand information of the subnet; and the server and bandwidth information includes a virtual server bandwidth of each external port of the subnet;
  • An embodiment of the present invention provides a joint optimization system, including:
  • the joint optimization computing device is configured to collect link information, server and bandwidth information, and user demand information; obtain routing parameters and server selection parameters, obtain optimized ingress bandwidth and optimize export bandwidth of each external port;
  • the content routing engine is configured to convert the routing parameters and server selection parameters obtained by the joint optimization computing device into routing parameters and server selection parameters of the network in which the application is located.
  • the embodiment of the present invention compares the servers outside the subnet to the virtual server on the device at the entrance of the subnet by performing joint optimization calculation in parallel in each subnet.
  • the bandwidth of this virtual server is the smaller of the optimized egress bandwidth calculated by the port to the terminal network joint optimization and the ingress bandwidth on the subnet minus the ingress background traffic, and the bandwidth through the virtual server is also between the subnets.
  • Bandwidth coupling, joint optimization calculations are performed on each subnet to obtain the routing policy and server selection policy of the network. Applying this routing policy and server selection strategy in the network makes the performance and utilization of the network the highest.
  • Figure 1 illustrates an embodiment in which a jointly optimized network is decomposed into two layers
  • Figure 2 illustrates an embodiment in which a jointly optimized network is decomposed into three layers
  • FIG. 5 is an embodiment of a network-wide joint optimization execution process
  • Figure 6 is an embodiment of a joint optimization control system.
  • Figure 1 shows the network that performs joint optimization into two layers.
  • the first layer is the access layer.
  • the subnet is divided into BAS (Broadband Access Server), that is, the BAS and the switch connected to it, DSLAM. ( , Digital Subscriber Line Access Multiplexer, etc.) form a subnet, and the BAS and the Border Router (BR) connected to it and the core router (CR, Core Router) form an upper layer.
  • BAS Broadband Access Server
  • DSLAM Digital Subscriber Line Access Multiplexer, etc.
  • BR Border Router
  • CR Core Router
  • Subnet At the access layer, subnetting in units of BAS is a method of subnetting. It can also be subnetted by an edge router or other device.
  • the second is an embodiment in which a network performing joint optimization is divided into three layers.
  • the first layer is an access layer, which is divided into multiple subnets by BAS and edge routers, and is divided into different areas by region in the middle layer.
  • the subnet, the uppermost layer is a subnet, including network nodes such as egress routers.
  • the decomposition of the network is not limited to two or three layers, and may be decomposed into multiple layers.
  • the bottom layer may be referred to as a first layer subnet and a second layer subnet.
  • the third layer subnet, ..., the M layer subnet, the uppermost layer network can also be called the upper layer subnet, and the first layer subnet can also be called the bottom layer subnet.
  • 3 is an embodiment of a joint optimization computing device including a parameter collection module, a calculation module, and an output module.
  • the parameter collection module is responsible for collecting network status information, server information, user demand information, and optimized output bandwidth of the adjacent subnet output, and providing the collected input to the calculation module; wherein the network status information includes link bandwidth and link
  • the user demand information includes a collection of users, a user's demand for content, etc.
  • the server information includes a server set, a server bandwidth, and the server set includes a server in the subnet and an entrance of the adjacent subnet.
  • the virtual server, the optimized egress bandwidth output by the neighboring subnet is the egress bandwidth that can be provided by the adjoining subnet for joint optimization calculation. This bandwidth is used as the bandwidth of the virtual server at the entrance of the adjacent subnet.
  • the calculation module receives the network state information collected by the parameter collection module, the server information, the user information, and the optimized output bandwidth of the neighboring subnet output, performs a joint optimization algorithm calculation, and calculates the egress bandwidth and the required ingress bandwidth that the subnet can provide. And the corresponding routing policy and server selection policy, the results are provided to the output module.
  • the output module receives the optimal result calculated by the calculation module and calculates the calculation result of the module, and compares Compared with the ingress bandwidth required by the subnet and the egress bandwidth that the adjacent subnet can provide, if the difference between the two is within the error range, a routing policy and a server selection policy are provided to the content routing engine. Also provide the egress bandwidth that this subnet can provide to the content routing engine.
  • Users in a subnet may download content from servers on this subnet, or may download content from servers on other subnets or upper subnets. Users in subnets request servers outside the subnet to pass through this subnet and other subnets.
  • the outlet of the network connection so in terms of a subnet, it can be considered that the server outside the subnet is the virtual server on the device at the entrance of the subnet, and the sum of the bandwidths that all servers in the subnet can provide.
  • the smallest of the maximum egress bandwidth of the neighboring subnet and the maximum ingress bandwidth of the subnet which is the bandwidth that the virtual server at the entrance of the subnet can provide.
  • the maximum egress bandwidth of the subnet refers to the value of the subnet's egress bandwidth minus the egress background traffic.
  • the maximum ingress bandwidth of the subnet refers to the subnet's ingress bandwidth minus the entry background. The value after the flow.
  • the user set T is a collection of all users in the subnet requesting CP content
  • the server set S is a collection of all servers in the subnet and virtual servers on the devices at the ingress.
  • the bandwidth of the server in the subnet is the bandwidth that the server can provide, and the bandwidth of the virtual server on the device at the ingress is not greater than the value of the ingress bandwidth of the subnet minus the background traffic of the portal.
  • the value is According to the statistical prediction of network information, when the neighboring subnet outputs optimized outlet bandwidth, the virtual server bandwidth on the device at the ingress is the minimum of the neighboring subnet output optimized outlet bandwidth, the subnet entry bandwidth minus the entrance background traffic.
  • the starting node set I of the background stream is a set of those nodes requesting the background stream in the subnet
  • the terminating node set J is the egress node of the metropolitan area network
  • the station is also assumed to have a device on the device at the entrance point.
  • the virtual background stream terminates the node.
  • the mathematical model of the joint optimization algorithm is established by using the convex optimization technique and the NBS (Nash Bargaining Solution) of the game theory, that is, the constraints such as link information, server and bandwidth information, and user demand information are considered.
  • NBS Network Bargaining Solution
  • the optimization problem is solved to obtain a set of optimal values, which reflect the optimal routing change strategy of the TE and the optimal server selection strategy of the SS.
  • the user-to-server band can be calculated.
  • Wide demand which also includes the bandwidth requirement for the server outside the subnet, that is, the bandwidth requirement of the virtual server on the device at the entrance, the bandwidth requirement and the maximum ingress bandwidth of the subnet on the port minus the background traffic.
  • the smaller value is used as the optimized ingress bandwidth for this subnet on this port. According to the set of optimal values, the remaining bandwidth that can be provided by the server in the subnet through the port can be calculated, and the bandwidth and the maximum egress bandwidth of the subnet on the port are subtracted from the background traffic. The small value is used as the optimized exit bandwidth for this subnet on this port.
  • the SS-drawn TE-NBS sub-problem after the first-stage decomposition of the joint optimization original problem is further advanced in the angle of the link, the server, and the background flow end point.
  • the level is decomposed to decompose it into smaller problem solvers, reducing the time it takes for the algorithm to execute.
  • the decomposition principle is as follows:
  • the joint optimization mathematical model of the network is decomposed into the first-level decomposition according to the dual decomposition method, which is decomposed into two sub-problems of SS-NBS and TE-NBS, and the main problem of DualTE-SS, and the main problem ⁇ > ⁇ /73 ⁇ 4 -
  • the price variables, and introduced by the decomposition method control the two sub-problems -A ⁇ and 73 ⁇ 4-A3 ⁇ 4. Since the two sub-problems - ⁇ ⁇ and 73 ⁇ 4-A3 ⁇ 4 are still constrained optimization problems, further decomposition is necessary.
  • the second-level decomposition is performed in units of link/server S, and is decomposed into two sub-problems of SS-NBS-L and SS-NBS-S, and the DimlSS-NBS master and
  • the main problem wa/ -A ⁇ controls the two sub-problems of SS-NBS-L and SS-NBS-S by the price variables introduced by the second-level decomposition, and . Since the two sub-problems SS-NBS-L and SS-NBS-S are the smallest solvable problems, it is not necessary to continue the decomposition. Otherwise, it is still necessary to continue the decomposition with a similar decomposition method until the minimum solvable problem is obtained.
  • the second level decomposition is performed in the unit of link 1 and background flow end j, and is decomposed into two sub-problems of TE-NBS-L and TE-NBS-J, and the main problem of DiialTE-NBS.
  • DiialTE-NBS controls the two sub-problems TE-NBS-L and 73 ⁇ 4-A3 ⁇ 4 -J through the price variables introduced by the second-level decomposition.
  • the equation can be decomposed into SS-NBS, and two sub-problems SS-NBS, wherein the two sub-problems acceptor question D Ma / r £ _ ss ( 4, ⁇ , ⁇ ,) control.
  • G( ,v,) log(rE 0 +f )) + ⁇ ( l fl ,g ' -vj - )
  • V(s,t)eSxT; f t represents the set of users served by the server, s (o represents a collection of servers served by a user.
  • the last constraint indicates that for non-servers and users
  • the intermediate node V the flow rate into V is equal to the flow flow from V, that is, the flow conservation law is satisfied for the intermediate node V.
  • the establishment of this formula is obvious, so it is not considered when generating the Lagrangian expression.
  • / eJ L(w) represents the link from the server s to the user.
  • Layer 5 sub-problems are divided into two sub-problems based on link I and server S ⁇ -N ⁇ S- and
  • the TE-NBS subproblem in ( 1. 2 ) rE _ NBS(r; j , can be further expressed as follows
  • I e L ⁇ i,j represents the link from i to j.
  • the rE-Nas subproblem is divided into two sub-problems rE-N ⁇ s - and according to the link I and the background flow end point.
  • the Variable ⁇ , r t calculation module decomposes the original problem into two sub-problems with less computation according to the above decomposition process. After calculating the optimal result, it will feed back to the output module, and the output module outputs the calculation result to the content routing engine.
  • the result on the network will change the routing policy and server selection policy.
  • the routing change policy will allocate traffic to each path in an optimal manner to solve the traffic engineering problem; the server selection policy will be applied to the server (including the virtual server in the subnet), specifying the bandwidth that the server can allocate and the server to serve. New user collection.
  • Figure 5 is the execution flow of the joint optimization of the whole network.
  • the network to be jointly optimized is decomposed into multiple subnets, and the joint optimization is solved separately in each subnet. If the subnets are not solved from the adjacent network in the first solution,
  • the bandwidth value is set to an initial value of the virtual server bandwidth on the device at the entry. The initial value is not greater than the subnet entry bandwidth minus the entry background traffic.
  • the value can be predicted according to the network information. Alternatively, a value can be arbitrarily selected within this range, such as the subnet entry bandwidth minus one-half of the ingress background traffic. If an optimized egress bandwidth value is obtained from the neighboring network, the neighboring network is selected to optimize the egress bandwidth, the subnet. The minimum of the ingress bandwidth minus the ingress background traffic is used as the virtual server bandwidth on the device at the set entry.
  • the joint optimization described above is performed on each subnet in an iterative manner, and the optimization problems of all the subnets and upper subnets are finally converged to the optimal value, thereby achieving Pareto optimality.
  • the route change policy and the server selection policy after the system convergence are sent to the relevant network elements in the network.
  • Step 1 Set an initial hypothetical ingress bandwidth for each subnet in the lower layer network, and the parameter collection module collects network status information (such as link information, etc.), server and bandwidth information, and user requirement information of the subnet, and calculates The module solves the optimal routing parameters and server of this subnet according to the decomposition method of FIG. Parameters.
  • the optimal egress bandwidth that this subnet can provide is derived, where the optimized egress bandwidth is the smallest of the port's maximum egress bandwidth and the maximum bandwidth that the server within the subnet can provide.
  • the optimized ingress bandwidth of this subnet is obtained, that is, the bandwidth provided by the server outside the subnet is required;
  • Step 2 If there is a non-top layer intermediate layer subnet, the parameter collection module of the joint optimization system of the middle layer subnet collects the optimized exit bandwidth calculated by the adjacent lower layer subnet, and the intermediate layer subnet Network status information (such as link information, etc.), server and bandwidth information, user demand information, set initial intermediate layer subnet and hypothetical ingress bandwidth of the upper subnet interface, and the calculation module solves according to the decomposition method of FIG.
  • the optimal routing parameters and server parameters for this middle tier subnet The optimized egress bandwidth provided by the subnet in the upper layer network and the subnet in the lower layer network, where the optimized egress bandwidth refers to the smallest of the maximum egress bandwidth of the port and the maximum bandwidth that the server in the subnet can provide. .
  • the optimized ingress bandwidth of this subnet is obtained, that is, the bandwidth provided by the server outside the subnet is required;
  • Step 3 Optimize the egress bandwidth calculated by the parameter collection module of the joint optimization system of the uppermost subnet and the subnet, and the network state information (such as link information, etc.) calculated by the adjacent subnet. ), server and bandwidth information, the calculation module solves the optimal routing parameters and server parameters of the subnet according to the decomposition method of FIG. 4 .
  • the optimal egress bandwidth provided by the subnet in the lower layer network of this subnet is obtained, where the optimized egress bandwidth is the smallest of the maximum egress bandwidth of the port and the maximum bandwidth that the server within the subnet can provide.
  • the optimized ingress bandwidth of this subnet is obtained, that is, the bandwidth provided by the server outside the subnet is required;
  • Step 4 Replace the optimized assumed egress bandwidth calculated in step 3 with the lower subnet interface with the initial hypothetical ingress bandwidth set in step 2, and perform step 2, using step 2 to calculate the optimized egress bandwidth of the subnet interface.
  • the initial hypothetical ingress bandwidth is set in step 1
  • the first step is performed. If the calculated hypothetical ingress bandwidth and the optimized ingress bandwidth are within the error range, the optimal routing parameters and server parameters are output and act on the corresponding subnet. Otherwise, repeat step 1, step 2, step 3, and step 4 again.
  • the joint optimization system includes a content routing engine, a joint optimization computing device, and a composition of the joint optimization computing device.
  • FIG. 3 further includes a joint optimization system and a proxy module, and a name is also included.
  • Proxy As a proxy for the user, it establishes a bridge between the user terminal and the joint optimization system, and can also be used to aggregate the user's request. It can be deployed at the edge of the network, such as on a DSLAM.
  • Content Routing Engine It determines how to get the content you need. When it finds that the required content is not on the local cache device, it uses the Name Resolution Engine to get a list of nodes that own the content, which is input as server information to the parameter collection module.
  • the content routing engine obtains the routing change policy and the server selection policy from the output module of the joint optimization computing device, and converts it into a deployable implementation routing change policy and a server selection policy, the former being input to the transmission/retransmission engine to execute the underlying routing policy, and then The input is to the topology maintenance module to perform server selection at the content layer.
  • Name Resolution Engine Mainly returns a list of suitable nodes with the content the user wants.
  • Content GET/PUT Adapter It requests the content of the content routing engine from the local cache device and returns it to the content routing engine so that the content routing engine routes the content to the destination user.
  • Traffic Statistics Collector summarizes background traffic information, link information, etc., and reports them to the parameter collection module.
  • Requests Statistics Collector Aggregates user requirements and sends them to the parameter collection module.
  • For the terminal network joint optimization system obtain the optimized exit bandwidth calculated by the subnet joint optimization computing device, and use the optimized egress bandwidth for the terminal network to obtain the hypothetical ingress bandwidth required for performing joint optimization, which is virtualized on the device at the entrance.
  • the bandwidth of the server The bandwidth of the server.
  • Parameter collection module It mainly collects user requests, server list information, dynamic characteristics of traffic, and optimized export bandwidth information for the terminal network.
  • Calculation module Using the optimal decomposition theory to calculate the optimal (user, server) pair, and at the same time, derive the end-to-end path and the flow ratio on each path for each (user, server) pair.
  • Output module The output module is responsible for receiving the optimal calculation result of the calculation module and inputting it to the content routing engine.
  • Transmission/Forwarding Engine Receives the IP layer of the content routing engine, performs traffic transmission, and implements the optimal routing of the underlying layer.
  • Topology maintenance module Receives the application layer instructions of the content routing engine, performs server selection, and implements content layer routing.
  • the interaction process of the joint optimization system is as follows:
  • the proxy module completes the client user's request collection of content
  • the naming parsing engine performs content parsing according to the requested content name, and knows the user's demand information for the content and inputs it to the request generator;
  • the content routing engine obtains the server information, first obtains the server information of the local cache device through the content GET/PUT adapter, and when the local cache device does not have the required content, obtains the node list information with the content required by the user through the naming parsing engine. ;
  • the parameter collection module obtains the user's request information from the request generator, obtains the server information from the content routing engine, obtains network status information (such as background flow information, link information, etc.) from the traffic generator, and the slave terminal network.
  • the interface obtains the egress bandwidth information and the ingress bandwidth information of the terminal network;
  • the calculation module uses the information obtained by the parameter collection module to perform the joint optimization calculation using the optimization decomposition theory, and obtains the optimal (user, server) pair, and at the same time, the end-to-end formed by each (user, server) pair is obtained.
  • the output module inputs the optimal calculation result of the calculation module to the content routing engine
  • the content routing engine obtains the optimal solution from the output module and converts it into a deployable implementation of the route change policy and the server selection policy.
  • the former is input to the transport/forward engine to execute the underlying routing policy, and the latter is input to the topology maintenance module.
  • the relevant information required to update the egress bandwidth and the ingress bandwidth is input to the terminal network;
  • the transport/forward engine receives a routing change policy from the content routing engine to modify the underlying route and perform a routing change policy at the IP layer;
  • the topology maintenance module receives a server selection policy from the content routing engine to perform server selection at the application layer, thereby establishing a mapping relationship between the user and the server.
  • a global network optimization problem is equivalently converted into several subnet optimization problems.
  • the joint optimization model variables and constraints in the subnet are relatively small and the calculation amount is small.
  • the embodiment also discloses a method for solving the joint optimization model in the subnet, and decomposes the original mathematical model according to the link, the server and the background flow end point, and decomposes it into a smaller solvable problem, and reduces the execution time of the algorithm. Since the subnets are only coupled by bandwidth, the coupling relationship is loose. Solve the problem of optimization.
  • the foregoing program may be stored in a computer readable storage medium, and the program is executed when executed.
  • the foregoing steps include the steps of the foregoing method embodiments; and the foregoing storage medium includes: a medium that can store program codes, such as a ROM, a RAM, a magnetic disk, or an optical disk.

Landscapes

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

Abstract

本发明实施例公开了一种联合优化的方法、装置和系统,其中,联合优化方法包括:将全网的联合优化分解成在各个子网中执行联合优化,对子网外服务器的带宽需求看成对所述端口上的虚拟服务器的带宽需求,各子网迭代执行联合优化,将执行联合优化的结果应用到网络中。本发明实施例通过对子网外服务器的带宽需求看成对所述端口上的虚拟服务器的带宽需求,在各子网迭代执行联合优化,并将执行联合优化的结果应用到网络中,从而实现并行执行全网联合优化,同时提高网络的性能和用户使用网络时的体验。

Description

联合优化的方法、 装置和系统
本申请要求于 2010 年 10 月 30 日提交中国专利局、 申请号为 201010525399.7、 发明名称为 "联合优化的方法、 装置和系统" 的中国专利申 请的优先权, 其全部内容通过引用结合在本申请中。
技术领域
本发明涉及通信技术领域, 尤其涉及一种联合优化的方法、 装置和系统。
背景技术
ISP ( Internet Service Provider,互联网服务提供商 )的角色主要是提供互联 网接入, 需要解决流量工程(TE, Traffic Engineering )的问题, 即为流量指定 最优的路径以最小化网络的拥塞, CP ( Content Provider, 内容提供商)的角色 是向用户提供所需的内容, 需要解决服务器选择(SS, Server Selection ) 的问 题, 即为不同的用户指定最优的服务器以最小化端到端的时延。 当 ISP和 CP 分别独立完成 TE和 SS时, 从全网的角度看并不是一种全局最优的状态。
为了获得全网的最优性能,现在人们开始尝试 TE和 SS的联合优化( Joint Optimization ), 利用博弈论中的纳什议价解( NB S , Nash Bargain Solution )及 凸优化技术对给定的 TE和 SS这两个最优化问题实现合作博弈, 以期达到全 局最优的平衡。
利用 NBS建立的 TE和 SS联合优化数学模型为一个带约束的凸优化问题, 现有技术存在一种基于集中式方式的联合优化求解方法,该方案假定某个计算 设备收集优化问题求解所需的全网状态信息,进而釆用集中式方式获取系统的 最优解, 并将最优解作用到网络,从而执行路由变更策略和新的服务器选择策 略。 但该方案扩展性较差, 随着网络规模的扩大, 收集信息和计算所耗费的时 间将急剧增加,如果要将联合优化作用到规模较大的网络,如整个城域网甚至 全国骨干网, 则现有的联合优化方法对网络产生作用的实时性将很难得到保 证。
发明内容
本发明实施例提供一种联合优化的方法、装置和系统,有效降低联合优化 的计算量和复杂度, 提高全网的性能和效率。
联合优化方法, 包括: 获得所述子网的链路信息、 服务器及带宽信息、 用户需求信息; 所述服务 器及带宽信息包括所述子网每个对外端口的虚拟服务器带宽,所述虚拟服务器 带宽为所述子网通过所述对外端口对子网外服务器需求的带宽;
根据链路信息、服务器及带宽信息、用户需求信息得到所述子网的最优路 由参数和服务器选择参数;
根据最优路由参数和服务器选择参数得到所述子网每个对外端口的优化 入口带宽;
比较每个对外端口的优化入口带宽和虚拟服务器带宽,若所有对外端口的 优化入口带宽和虚拟服务器带宽的比较结果均小于设定误差,则在所述子网中 应用所述最优路由参数和服务器选择参数。
本发明实施例提供一种联合优化计算装置, 包括:
参数釆集模块: 用于获得所述子网的链路信息、 服务器及带宽信息、 用户 需求信息;所述服务器及带宽信息包括所述子网每个对外端口的虚拟服务器带 宽,所述虚拟服务器带宽为所述子网通过所述对外端口对子网外服务器的带宽 需求;
计算模块: 用于根据链路信息、 服务器及带宽信息、 用户需求信息得到所 述子网的最优路由参数和服务器选择参数,根据最优路由参数和服务器选择参 数得到所述子网每个对外端口的优化入口带宽;
输出模块: 用于比较每个对外端口的优化入口带宽和虚拟服务器带宽, 若 所有对外端口的优化入口带宽和虚拟服务器带宽的比较结果均小于设定误差, 则在所述子网中应用所述最优路由参数和服务器选择参数,并输出所述最优路 由参数和服务器选择参数。
本发明实施例提供一种通信网络, 包括:
联合优化系统获得子网的链路信息、 服务器及带宽信息、 用户需求信息; 所述服务器及带宽信息包括所述子网每个对外端口的虚拟服务器带宽;
根据链路信息、服务器及带宽信息、用户需求信息计算出所述子网的最优 路由参数和服务器选择参数, 得到此子网每个对外端口的优化入口带宽; 比较每个对外端口的优化入口带宽和此对外端口的虚拟服务器带宽,若比 较结果 d、于设定误差, 则在网络中应用所述最优路由参数和服务器选择参数。 本发明实施例提供一种联合优化系统, 包括:
联合优化计算装置, 用于收集链路信息、 服务器及带宽信息、 用户需求信 息; 获得路由参数和服务器选择参数, 获得每个对外端口的优化入口带宽和优 化出口带宽;
内容路由引擎,用于将联合优化计算装置获得的路由参数和服务器选择参 数转化为应用在所在网络的路由参数和服务器选择参数。
由上述本发明实施例提供的技术方案可以看出,本发明实施例通过在各个 子网中并行进行联合优化计算,将子网外的服务器看成本子网的入口处的设备 上的虚拟服务器,此虚拟服务器的带宽为此端口对端子网联合优化计算出的优 化出口带宽和子网此端口上的入口带宽减去入口背景流量中的较小者,通过虚 拟服务器的带宽也就是子网之间的带宽耦合, 在各子网迭代进行联合优化计 算,从而获得网络的路由选择策略和服务器选择策略,在网络中应用此路由选 择策略和服务器选择策略使得网络的性能和利用率最高。
附图说明
为了更清楚地说明本发明实施例中的技术方案,下面将对实施例描述中所 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性的前提 下, 还可以根据这些附图获得其他的附图。
图 1执行联合优化的网络分解为两层的实施例;
图 2执行联合优化的网络分解为三层的实施例;
图 3是联合优化计算装置的实施例;
图 4是联合优化数学模型进行两级分解的实施例;
图 5是全网联合优化执行流程的实施例;
图 6是联合优化控制系统的实施例。
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。 下面对本发明实施例应用的通信网络进行举例说明。
图 1 为将执行联合优化的网络分解为两层, 第一层为接入层, 以 BAS ( Broadband Access Server, 宽带接入服务器)为单位划分子网, 即 BAS及其 下面连接的交换机、 DSLAM ( , Digital Subscriber Line Access Multiplexer, 数 字用户线路接入复用器)等组成一个子网, BAS 及其上面连接的边缘路由器 ( BR, Border Router )、核心路由器( CR, Core Router )等组成一个上层子网。 在接入层, 以 BAS为单位划分子网是一种划分子网的方法, 也可以以边缘路 由器或其它设备为范围划分子网。
图 2是将执行联合优化的网络划分为三层的实施例, 第一层为接入层, 以 BAS、 边缘路由器为单位划分为多个子网, 在中间层, 以地域为单位划分为不 同的子网, 最上层是一个子网, 包括出口路由器等网络节点。
需要说明的是, 网络的分解也不限于两层或三层, 可以分解为多层, 当分 解为多层时,从下往上可以依次称作第一层子网,第二层子网,第三层子网, ... , 第 M层子网, 最上层的网络也可以称作上层子网, 第一层子网也可以称作底 层子网。
图 3 是联合优化计算装置的实施例, 联合优化计算装置包括参数釆集模 块、 计算模块和输出模块。
参数釆集模块负责收集网络状态信息、 服务器信息、 用户需求信息、 以及 邻接子网输出的优化输出带宽,将釆集到的输入提供给计算模块; 其中的网络 状态信息包括链路带宽、 链路的服务质量 QoS、链路的成本等, 用户需求信息 包括用户的集合、 用户对内容的需求等, 服务器信息包括服务器集合, 服务器 带宽,服务器集合包括子网内的服务器和邻接子网入口处的虚拟服务器,邻接 子网输出的优化出口带宽是邻接子网进行联合优化计算后可以提供的出口带 宽, 此带宽作为邻接子网入口处的虚拟服务器的带宽。
计算模块接收参数釆集模块收集的网络状态信息、服务器信息、用户信息、 以及邻接子网输出的优化输出带宽,执行联合优化算法计算,计算出此子网可 以提供的出口带宽、需要的入口带宽、以及相应的路由策略和服务器选策策略, 将结果提供给输出模块。
输出模块接收计算模块计算出的最优结果并将计算模块的计算结果,并比 较本子网需要的入口带宽和邻接子网所能提供的出口带宽,如果两者之差在误 差范围内, 则提供路由策略和服务器选择策略给内容路由引擎。 同时提供此子 网可以提供的出口带宽给内容路由引擎。
为了进一步理解上述联合优化装置的实现原理,下面进一步阐述邻接子网 入口处的虚拟服务器的原理, 以及建立联合优化数学模型的原理:
子网内的用户可能从本子网内的服务器上下载内容,也可能从其它子网或 上层子网的服务器那里下载内容,子网内的用户请求子网外的服务器需经过本 子网和其它子网相连接的出口, 因此以一个子网为单位来看, 可以认为子网外 的服务器就是位于本子网的入口处的设备上的虚拟服务器,邻接子网内所有服 务器能提供的带宽之和、邻接子网的最大出口带宽、本子网的最大入口带宽中 最小者,作为本子网入口处的虚拟服务器能提供的带宽。邻接子网内所有服务 器能提供的带宽之和、邻接子网的最大出口带宽中的最 d、者也称作邻接子网的 优化出口带宽。本文中如果没有特别说明, 所说的子网的最大出口带宽是指本 子网出口带宽减去出口背景流量后的值,所说的子网的最大入口带宽是指本子 网入口带宽减去入口背景流量后的值。
对于该子网,假定用户集合 T为该子网内所有请求 CP内容的那些用户的 集合,服务器集合 S为子网内所有服务器以及入口处的设备上虚拟的服务器组 成的集合。 子网内的服务器带宽为所述服务器能提供的带宽, 而入口处的设备 上虚拟的服务器带宽为不大于该子网入口带宽减去入口背景流量后的某个值, 初始计算时, 该值可根据网络信息统计预测得出, 当邻接子网输出优化出口带 宽后,入口处的设备上虚拟的服务器带宽为邻接子网输出优化出口带宽、本子 网入口带宽减去入口背景流量之中的最小者。 背景流的起始节点集合 I为子网 内请求背景流的那些节点组成的集合, 终止节点集合 J为城域网的出口节点, 站在子网的角度也即假定入口处的设备上有个虚拟的背景流终止节点。
在此子网内,利用凸优化技术和博弈论的 NBS ( Nash Bargaining Solution, 纳什议价解)建立联合优化算法的数学模型, 即在考虑链路信息、 服务器及带 宽信息和用户需求信息等约束条件下建立一个带约束的联合优化问题。对该最 优化问题进行求解, 得到一组最优值, 这组值反映了 TE的最优路由变更策略 和 SS的最优服务器选择策略。 根据这组最优值可以计算出用户对服务器的带 宽需求, 其中也包括了对子网外服务器的带宽需求,也即入口处的设备上虚拟 服务器的带宽需求,该带宽需求和所述子网在此端口上的最大入口带宽减去背 景流量后的较小值作为此子网在此端口上的优化入口带宽。根据这组最优值还 可以计算出所述子网内服务器可以通过所述端口向外提供的剩余带宽,该带宽 和所述子网在此端口上的最大出口带宽减去背景流量后的较小值作为此子网 在此端口上的优化出口带宽。
图 4是联合优化求解的一种实施例,对联合优化原问题进行第一级分解后 的 SS-画 TE-NBS两个子问题站在链路、 服务器和背景流终点的角度作进 一步的第二级分解, 使其分解为更小的可解子问题, 减少算法执行的时间。
分解原理如下:
首先对所述网络的联合优化数学模型按照对偶分解法做第一级分解,将其 分解为 SS-NBS和 TE-NBS两个子问题, 以及 DualTE-SS主问题, 且主问题 Γ>Μα/7¾- 通过分解法引入的价格变量 、 和 来控制 -A^ 和 7¾-A¾ 这两 个子问题。 由于 -ΛΦ 和 7¾-A¾ 两个子问题仍为带约束的最优化问题, 因 此有必要做进一步的分解。 此时, 针对 SS-A^ 子问题, 以链路 /和服务器 S 为单位做第二级分解, 分解为 SS-NBS-L 和 SS-NBS-S 两个子问题, 以及 DimlSS-NBS主 ^且主问题 wa/ -A^ 通过第二级分解引入的价格变量 、 和 来控制 SS-NBS-L和 SS-NBS-S这两个子问题。 由于 SS-NBS-L和 SS-NBS-S 这两个子问题已是最小的可解子问题, 因而不必继续分解, 否则仍需釆用类似 分解法继续分解, 直到得到最小的可解子问题。 同理, 针对 7¾-A¾ 子问题, 以链路 1和背景流终点 j为单位做第二级分解, 分解为 TE-NBS-L和 TE-NBS-J 两个子问题, 以及 DiialTE-NBS主问题 , 且主问题 DiialTE-NBS通过第二级分 解引入的价格变量 和 来控制 TE-NBS-L 和 7¾-A¾ -J这两个子问题。 由于 TE-NBS-L和 7¾-A¾ -J这两个子问题已是最小的可解子问题,因而不必继续分 解, 否则仍需釆用类似分解法继续分解, 直到得到最小的可解子问题。
上述的两级分解过程进一步详细描述如下:
一、 联合优化的原始问题
利用凸优化技术和博弈论中的 NBS , 建立子网中联合优化问题的数学模 型如下所示: maximize log(rE0 -∑ g, (/ + ) + l0g(1¾ - (/ + )) Subject to ∑ V/
Figure imgf000009_0001
∑ r - ∑ r卜 I , {i,j)^SxT, V V\ i}
Figure imgf000009_0002
Variables xf >0; 0≤ rf < 1, V( ,y) iS T; ft cp; ft bg 式中, (rE。, 5¾)表示现状, 即 TE和 SS单独优化时获得的最优值; &(·)和 (·)分别为联合优化的 TE和 SS取定的代价函数,均为链路 /的函数; / 和^^ 分别表示链路 /的背景流量和从 isp角度看链路 /上首选的 CP流量; 和 分别表示链路 I的 CP流量和从 CP角度看链路 I上首选的背景流量; 表示 (s, 对在链路 /上产生的 CP流量; ¾和 分别表示( , )对的背景流量 和( , ·)对在链路 /上产生的背景流比重; ς、 和 A分别表示链路 /的容 量、 用户 的需求和服务器 S的带宽; 为指示函数, 表示当节点 V为节点 时该函数取 1, 为除节点 和 ·外的其它节点时该函数取 0; 和 的含义与 I 相同; / e In(v)和 I e Out(v)分别表示流量流进节点 V的链路和流量流出节点 V 的链路。 变量 、 r' . 和 即为联合优化问题所要求解的最优化变量。
对上式作部分拉格朗日变换, 可得:
+i。g(SS。 -∑ +W)+Y f -in
ι ι
+∑ucl-f -fl bs)
= log(7E0 -∑g/(^ +^)) +∑( ¾ -ν,/Γ ~Ubg)
+iog(SS0 - j f + g)+∑Vl f - 利用对偶分解法, 上式可分解为 SS-NBS和 SS-NBS两个子问题, 其中这 两个子问题受主问题 DMa/_ss(4, ^,ν,)的控制。
(1.1)、
Figure imgf000010_0001
Subject to ft cp = jX, V/
Figure imgf000010_0002
Variable ≥ 0, (s,t) SxT; (1.2)、 TE-NBS子问题 rE _ NBSif , p ) maximize log(7E0
Figure imgf000010_0003
Subject to j g = ^ V/
(i Sx-T
∑ r卜 ∑ =1 , (i,j)^SxT, v V\{i}
Variable 0 < r < 1, V (, j)^Sx T;
(1.3)、 主问题 D / ss(/l,,〃,,v,)
令 SS-A^ 子问题目标函数的最大值为 Hd^A) , 7¾-A¾ 子问题目标函 数的最大值为 , 其中 ' ' ;
G( ,v,) = log(rE0 +f )) +∑( lfl,g' -vj - )
/ / 。 则主问题可表示为:
minimize DualTE_ss (Az ,μ ν}) = Η{λ} , , ) + 0{λ} ,μ νι) + ^λιϋι
ι
Variable λ} > 0; μ ν, (5) 通过分析 SS-NBS子问题和 TE-NBS千 m 发现这两个子问题可根据在 Link, Server和背景流量终点的角度进一步分解, 使其分解为更小的可并行处 理的子问题。
二、 -ΛΦ 子问题的进一步分解
(1.1)中的 SS-NBS子问题 - NBSix J,"8)可进一步表述如下:
maximize log(^0 + -
I
Subject t。 jcp , V/
Figure imgf000011_0001
∑ ∑ VseS
∑ = ∑ , Vv^{ ^}
Variable x > 0, V(s,t)eSxT; ft 这里, 表示服务器 s服务的用户的集合, s(o表示为某个用户 服务的 服务器的集合。 最后一个约束条件表示对于非服务器和用户的中间节点 V, 流 进 V的流量等于从 V流出的流量, 即对中间节点 V而言, 满足流量守恒定律。 该式的成立显而易见, 故在生成拉格朗日表达式时未予考虑。
对上式作拉格朗日变换, 可得:
Figure imgf000011_0002
+ ∑ ∑ χι ~Μ, iog(SS0 - f +Jn f - j^ -\f +P n
Figure imgf000011_0003
+ ∑rA-∑vtMt 这里, /eJL(w)表示从服务器 s到用户 所经过的链路。
層《5子问题依据链路 I 和服务器 S 分为两类子问题^-N^S- 和
SS-NBS-S , 其中这两个子问题受主问题 Z¾/a/ss_爾 (ρ,,^,τ^)的控制。
(2.1)、 SS-A^ - 子问题: SS-NBS-Lif^J^) 链路 /:
maximize logiSS^^h^ +fl bg)) +∑(v r ~ '8 ~ °P + Ρ,/Π
1 1
Variable ft cp; J""
(2.2)、 -A^S ^子问题: SS-NBSK 服务器 maximize -∑ Ρ,Ο ∑ xf
Figure imgf000012_0001
Variable x > 0
(2.3)、 -A^ 主问题:
Figure imgf000012_0002
令 SS-NBS-L子问题目标函数的最大值为 LSSMS— 、 , SS-NBS-S子问题目 标函数的最大值为 SSS—NBS JS,";) , 其中 爾—》 0¾ -∑w/ ' +7l bs'))+∑(vlf
Figure imgf000012_0003
-Mr +p r')
Figure imgf000012_0004
则主问题可表示为:
minimize DualssNBS (p, ,γ,,η,) = LSS_NBS_L (p, ) +∑ SSS_NBS_S (p,,Ys, t) + ^ YSBS - ]tMt
(10) Variable pt; ys; r\t 三、 7¾-A¾ 子问题的进一步分解
( 1.2)中的 TE-NBS子问题 rE _ NBS(r;j , 可进一步表述如下
maximize log(rE0 (/ + / - - A,/ "
Subject to f ∑ = W
∑ r卜 \, (i,j)^SxT (11)
∑ r卜 ∑ r/, (i,j)^SxT, v V\{i,j} Variable 0 < r < 1, V ( , j)^S T; ft cp 最后一个约束条件表示对于非 i节点和 j节点的中间节点 V, 流进 V的背 景流量等于从 V流出的背景流量,即满足流量守恒定律。该式的成立显而易见, 故在生成拉格朗日表达式时未予考虑。
对上式作拉格朗日变换, 可得:
LTE w , 7 , ^)= iog(¾ (f^ + j ))+∑ ( ^― v ― x ^ )
Figure imgf000013_0001
( s + 7r ))+∑ (P ^ - v - ^ + d )
Figure imgf000013_0002
ll I I≡L J 这里, I e L{i,j)表示从 i到 j所经过的链路。 rE-Nas子问题依据链路 I和背景流终点 分为两类子问题 rE-N^s - 和
TE-NBS-J, 其中这两个子问题受主问题 DualTE sι , τ.. )的控制。
(3.1)、 TE-NBS-L
Figure imgf000013_0003
链路 /:
maximize Ιο&(ΤΕ0-∑§Ι(^ +/^) +∑(^ -v ~ ^bg +ζ^)
(13)
Variable / ; f,cp (3.2), TE-NBS-J子 ^. 背景流终点 j': maximize
V l:l In(j) leL(iJ) (14)
Variable 0 < r < 1
(3.3)、 ΤΈ-ΛΦ 主问题: DuU ^ij) 令 TE-NBS-L子问题目标函数的最大值为 LTENBSL ^ , 7¾-A¾-J子问题目 标函数的最大值为 ^ , 其中
Figure imgf000014_0001
则主问题可表示为:
minimize DualTE_NBS (ζ, ,τ..) = LTE_NBS_L +
Figure imgf000014_0002
Variable ζ{, rt 计算模块按照上述的分解过程将原问题分解为计算量较小的两个子问题, 计算出最优结果后将回馈给输出模块,输出模块将计算结果输出给内容路由引 擎, 该结果作用于网络上将会改变路由策略和服务器选择策略。 其中, 路由变 更策略将按最优方式为每条路径分配流量,从而解决流量工程问题; 服务器选 择策略将作用到服务器(包括子网内虚拟的服务器),指定服务器可分配的带宽 及服务器所服务的新用户集合。
图 5是全网联合优化的执行流程,待联合优化的网络分解成多个子网,在 每个子网中分别进行联合优化求解,各子网在第一次求解时如果没有从邻接网 络获得优化出口带宽值, 设定入口处的设备上虚拟的服务器带宽为一初始值, 该初始值为不大于该子网入口带宽减去入口背景流量后的某个值,该值可根据 网络信息统计预测得出,也可以在此范围内任意选择一个值, 比如该子网入口 带宽减去入口背景流量的二分之一,如果从邻接网络获得优化出口带宽值,选 择邻接网络优化出口带宽、该子网入口带宽减去入口背景流量中的最小值作为 设定入口处的设备上虚拟的服务器带宽。
在各子网迭代执行上述的联合优化,当最终所有子网和上层子网的优化问 题均收敛到最优值,进而达到 Pareto最优。 系统收敛后的路由变更策略和服务 器选择策略发送给网络中的相关网元。
整个执行过程描述如下:
步骤一, 对下层网络中的每个子网, 设定初始的假设入口带宽, 参数釆集 模块釆集子网的网络状态信息 (如链路信息等)、 服务器及带宽信息、 用户需 求信息,计算模块按照图 4的分解方法求解出此子网的最优路由参数和服务器 参数。得出此子网可以提供的优化出口带宽, 此处优化出口带宽是指端口的最 大出口带宽和子网内服务器能提供的最大带宽中的最小者。得出此子网的优化 入口带宽, 也就是需要子网外服务器提供的带宽;
步骤二,如果存在非最上层的中间层子网, 中间层子网的联合优化系统的 参数釆集模块釆集相邻的下层子网计算出的优化出口带宽、以及所述中间层子 网的网络状态信息 (如链路信息等)、 服务器及带宽信息、 用户需求信息, 设 定的初始的中间层子网和更上层子网接口的假设入口带宽, 计算模块按照图 4 的分解方法求解出此中间层子网的最优路由参数和服务器参数。得出此子网向 上层网络中的子网和下层网络中的子网提供的优化出口带宽,此处优化出口带 宽是指端口的最大出口带宽和子网内服务器能提供的最大带宽中的最小者。得 出此子网的优化入口带宽, 也就是需要子网外服务器提供的带宽;
步骤三,对最上层子网, 子网的联合优化系统的参数釆集模块釆集相邻的 下层子网计算出的优化出口带宽、以及所述子网的网络状态信息 (如链路信息 等)、 服务器及带宽信息, 计算模块按照图 4的分解方法求解出此子网的最优 路由参数和服务器参数。 得出此子网向下层网络中的子网提供的优化出口带 宽,此处优化出口带宽是指端口的最大出口带宽和子网内服务器能提供的最大 带宽中的最小者。得出此子网的优化入口带宽,也就是需要子网外服务器提供 的带宽;
步骤四,用步骤三计算出的和下层子网接口的优化出口带宽替换步骤二中 设定的初始的假设入口带宽,执行步骤二, 用步骤二计算出的和下层子网接口 的优化出口带宽替换步骤一中的设定初始的假设入口带宽,执行步骤一,如果 计算出的假设入口带宽和优化入口带宽在误差范围内,输出最优路由参数和服 务器参数并作用于相应子网。 否则再次重复执行步骤一、 步骤二、 步骤三、 步 骤四。
图 6是联合优化系统的一种实施例, 联合优化系统包括了内容路由引擎、 联合优化计算装置, 联合优化计算装置的组成参考图 3 , 图 6中还包括了联合 优化系统与代理模块、 命名解析引擎、 内容 GET/PUT适配器、 流量统计汇总 器、 请求统计汇总器、 对端子网联合优化系统、 传输 /转发引擎、 拓朴维护模 块之间的连接关系。 代理模块( Proxy ): 作为用户的代理, 它建立起用户终端和联合优化系统 的桥梁, 亦可用来聚集用户的请求。 它可部署在网络边缘, 如部署在 DSLAM 上。
内容路由引擎( Content Routing Engine ): 它决定了如何获得所需的内容。 当它发现所需的内容不在本地緩存设备上时, 它会使用命名解析引擎(Name Resolution Engine )获得拥有该内容的节点列表, 作为服务器信息输入给参数 釆集模块。内容路由引擎从联合优化计算装置的输出模块获得路由变更策略和 服务器选择策略, 并转化为可部署实施的路由变更策略和服务器选择策略, 前 者输入给传输 /转发引擎以执行底层的路由策略, 后者输入给拓朴维护模块以 执行内容层的服务器选择。
命名解析引擎( Name Resolution Engine ): 主要返回具有用户所需内容的 合适的节点列表。
内容 GET/PUT适配器: 它从本地緩存设备中索取内容路由引擎所需的内 容并返回给内容路由引擎, 以便内容路由引擎将内容路由给目的用户。
流量统计汇总器( Traffic Statistics Collector ): 汇总背景流量信息、链路信 息等, 并报告给参数釆集模块。
请求统计汇总器( Requests Statistics Collector ): 汇总用户需求并发送给参 数釆集模块。
对端子网联合优化系统:获取本子网联合优化计算装置计算出的优化出口 带宽, 对端子网利用此优化出口带宽得到执行联合优化所需的假设入口带宽, 此带宽作为入口处的设备上虚拟的服务器的带宽。
参数釆集模块: 它主要釆集用户请求,服务器列表信息,流量的动态特性, 以及对端子网的优化出口带宽信息。
计算模块: 利用优化分解理论, 计算出最优的(用户, 服务器)对, 同时 得出每个(用户, 服务器) 对所形成的端到端路径及每条路径上的流量配比。
输出模块:输出模块负责接收计算模块的最优计算结果并输入给内容路由 引擎。
传输 /转发引擎: 接收内容路由引擎的 IP层的指令, 执行流量的传输, 实 现底层的最优路由。 拓朴维护模块: 接收内容路由引擎的应用层的指令, 执行服务器选择, 实 现内容层路由。
联合优化系统的交互过程如下:
1. 代理模块完成客户端用户对内容的请求收集;
2. 命名解析引擎根据请求的内容名执行内容解析, 得知用户对内容的需 求信息并输入给请求发生器;
3. 内容路由引擎获得服务器信息, 首先通过内容 GET/PUT适配器获得 本地緩存设备的服务器信息, 当本地緩存设备没有所需的内容时,再通过命名 解析引擎获得具有用户所需内容的节点列表信息;
4. 参数釆集模块从请求发生器获得用户的请求信息, 从内容路由引擎获 得服务器信息, 从流量发生器获得网络的状态信息(如背景流信息、 链路信息 等), 以及从对端子网接口获得对端子网的出口带宽信息和入口带宽信息;
5. 计算模块利用参数釆集模块获得的信息利用优化分解理论执行联合优 化计算, 得出最优的(用户, 服务器)对, 同时得出每个(用户, 服务器)对 所形成的端到端路径及每条路径上的流量配比;
6. 输出模块将计算模块的最优计算结果输入给内容路由引擎;
7. 内容路由引擎从输出模块获得最优解, 并转化为可部署实施的路由变 更策略和服务器选择策略, 前者输入给传输 /转发引擎以执行底层的路由策略, 后者输入给拓朴维护模块以执行内容层的服务器选择。此外还向对端子网输入 更新出口带宽和入口带宽时所需的相关信息;
8.传输 /转发引擎从内容路由引擎接收路由变更策略,以便修改底层路由, 执行 IP层的路由变更策略;
9. 拓朴维护模块从内容路由引擎接收服务器选择策略, 以便执行应用层 的服务器选择, 从而建立用户和服务器之间的映射关系。
本实施例将一个全局网络优化问题等价转化为若干个子网优化问题,相对 于全网的联合优化问题, 子网内的联合优化模型变量和约束条件比较小,计算 量小。 本实施例还公开了求解子网内的联合优化模型的方法, 根据链路、 服务 器和背景流终点分解原数学模型,使其分解为更小的可解子问题, 减少算法执 行的时间。 由于各子网之间仅通过带宽进行耦合, 因此耦合关系比较松, 这利 于优化问题的求解。
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质。
最后所应说明的是: 以上实施例仅用以说明本发明的技术方案, 而非对本 域的普通技术人员应当理解:其依然可以对本发明的技术方案进行修改或者等 同替换, 而这种修改或者等同替换并不脱离本发明技术方案的精神和范围。

Claims

权 利 要 求
1、 一种网络联合优化方法, 所述网络包括多个子网, 其特征在于: 获得所述子网的链路信息、 服务器及带宽信息、 用户需求信息; 所述服务 器及带宽信息包括所述子网每个对外端口的虚拟服务器带宽,所述虚拟服务器 带宽为所述子网通过所述对外端口对子网外服务器需求的带宽;
根据链路信息、服务器及带宽信息、用户需求信息得到所述子网的最优路 由参数和服务器选择参数;
根据最优路由参数和服务器选择参数得到所述子网每个对外端口的优化 入口带宽;
比较每个对外端口的优化入口带宽和虚拟服务器带宽,若所有对外端口的 优化入口带宽和虚拟服务器带宽的比较结果均小于设定误差,则在所述子网中 应用所述最优路由参数和服务器选择参数。
2、 根据权利要求 1所述的方法, 其特征在于:
若并非所有对外端口的优化入口带宽和虚拟服务器带宽的比较结果均小 于设定误差, 则重复获得优化入口带宽, 并比较优化入口带宽和虚拟服务器带 宽的过程, 直到比较结果小于设定误差。
3、 根据权利要求 1所述的方法, 获得对外端口的虚拟服务器带宽的方法 包括:
如果获得所述对外端口对应的邻接子网的优化出口带宽,则所述对外端口 处的虚拟服务器带宽为所述优化出口带宽和所述对外端口可利用入口带宽的 较小者;
如果未获得所述对外端口对应的邻接子网的优化出口带宽,则所述对外端 口处的虚拟服务器带宽为小于所述对外端口可利用入口带宽的某个值。
4、 根据权利要求 1所述的方法, 所述方法还包括:
根据计算出的所述子网的最优路由参数和服务器选择参数,得到所述子网 每个对外端口的优化出口带宽。
5、 根据权利要求 1所述的方法, 其特征在于, 所述根据链路信息、 服务 器及带宽信息、用户需求信息得到所述子网的最优路由参数和服务器选择参数 包括: 建立所述子网的联合优化数学模型;
对所述联合优化数学模型进行第一级分解, 得到 SS-NBS和 TE-NBS两 个子问题;
根据链路、 服务器和背景流终点对所述 SS-NBS和 TE-NBS两个子问题 作第二级分解; SS-NBS子问题的第二级分解包括, 以链路 1和服务器 s为单 位做第二级分解, 分解为 SS-NBS-L 和 SS-NBS-S 两个子问题, 以及 DualSS-NBS主问题,且主问题 DualSS-NBS通过引入的价格变量 A、 ^和 控 制 SS-NBS-L和 SS-NBS-S两个子问题; TE-NBS子问题的分解包括, 以链路 1 和背景流终点 j为单位做第二级分解,分解为 TE-NBS-L和 TE-NBS-J两个子 问题, 以及 DualTE-NBS主问题, 且主问题 DualTE-NBS通过引入的价格变 量 和 控制 TE-NBS-L和 TE-NBS-J两个子问题;
根据分解后的联合优化数学模型计算所述子网的最优路由参数和服务器 选择参数。
6、 根据权利要求 1所述的方法, 其特征在于, 根据最优路由参数和服务 器选择参数得到所述子网每个对外端口的优化入口带宽包括:
应用所述子网的最优路由参数和服务器选择参数,计算出用户通过所述对 外端口的虚拟服务器带宽需求,以所述虚拟服务器带宽需求和此对外端口上的 可利用入口带宽的较小值作为所述子网在所述端口上的优化入口带宽。
7、 根据权利要求 4所述的方法, 其特征在于, 根据计算出的所述子网的 最优路由参数和服务器选择参数,得到所述子网每个对外端口的优化出口带宽 的步骤包括:
应用所述子网的最优路由参数和服务器选择参数,计算出所述子网内服务 器可以通过所述对外端口向外提供的剩余带宽,所述剩余带宽和此对外端口上 的可利用出口带宽的较小值作为所述子网在所述端口上的优化出口带宽。
8、 根据权利要求 1所述的方法, 其特征在于, 所述链路信息包括: 链路带宽、 链路的服务质量 QoS、 链路的成本之中的至少一个。
9、 根据权利要求 1所述的方法, 其特征在于, 所述服务器及带宽信息包 括:
子网内服务器的集合、 每个服务器所能提供的带宽之中的至少一个。
10、 根据权利要求 1所述的方法, 其特征在于, 所述用户需求信息包括: 用户的集合、 用户对内容的需求之中的至少一个。
11、 一种联合优化计算装置, 包括:
参数釆集模块: 用于获得所述子网的链路信息、 服务器及带宽信息、 用户 需求信息;所述服务器及带宽信息包括所述子网每个对外端口的虚拟服务器带 宽,所述虚拟服务器带宽为所述子网通过所述对外端口对子网外服务器的带宽 需求;
计算模块: 用于根据链路信息、 服务器及带宽信息、 用户需求信息得到所 述子网的最优路由参数和服务器选择参数,根据最优路由参数和服务器选择参 数得到所述子网每个对外端口的优化入口带宽;
输出模块: 用于比较每个对外端口的优化入口带宽和虚拟服务器带宽, 若 所有对外端口的优化入口带宽和虚拟服务器带宽的比较结果均小于设定误差, 则在所述子网中应用所述最优路由参数和服务器选择参数,并输出所述最优路 由参数和服务器选择参数。
12、一种通信网络,所述网络包括多个子网,每个子网包括联合优化系统, 其特征在于:
联合优化系统获得子网的链路信息、 服务器及带宽信息、 用户需求信息; 所述服务器及带宽信息包括所述子网每个对外端口的虚拟服务器带宽;
根据链路信息、服务器及带宽信息、用户需求信息计算出所述子网的最优 路由参数和服务器选择参数, 得到此子网每个对外端口的优化入口带宽;
比较每个对外端口的优化入口带宽和此对外端口的虚拟服务器带宽,若比 较结果 d、于设定误差, 则在网络中应用所述最优路由参数和服务器选择参数。
13、 根据权利要求 12所述的通信网络, 其特征在于:
所述子网包括第一层子网和一个或多个上层子网,第一层子网包括至少一 个和上一层子网的端口, 上一层子网包括和其下层子网端口, 和更上层子网的 端口;
从第一层子网开始计算所述子网的最优路由参数和服务器选择参数,得到 此子网每个对外端口的优化入口带宽,最后最上层子网计算最优路由参数和服 务器选择参数, 得到此子网每个对外端口的优化入口带宽; 比较每个子网的每个对外端口的优化入口带宽和此对外端口的虚拟服务 器带宽, 若比较结果小于设定误差, 则在网络中应用所述最优路由参数和服务 器选择参数。
14、 一种联合优化系统, 包括:
联合优化计算装置, 用于收集链路信息、 服务器及带宽信息、 用户需求信 息; 获得路由参数和服务器选择参数, 获得每个对外端口的优化入口带宽和优 化出口带宽;
内容路由引擎,用于将联合优化计算装置获得的路由参数和服务器选择参 数转化为应用在所在网络的路由参数和服务器选择参数。
PCT/CN2011/075388 2010-10-30 2011-06-07 联合优化的方法、装置和系统 WO2011144178A1 (zh)

Priority Applications (4)

Application Number Priority Date Filing Date Title
EP11783095.0A EP2515478B1 (en) 2010-10-30 2011-06-07 Method, apparatus and system for joint optimizations
RU2012137874/08A RU2520354C2 (ru) 2010-10-30 2011-06-07 Способ, устройство и система для совместной оптимизации
BR112012022286A BR112012022286A2 (pt) 2010-10-30 2011-06-07 método, aparelho e sistema para otimização de ligação.
US13/721,530 US9003029B2 (en) 2010-10-30 2012-12-20 Method, apparatus and system for joint optimization

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201010525399.7 2010-10-30
CN201010525399.7A CN102130824B (zh) 2010-10-30 2010-10-30 网络联合优化方法和装置

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US13/721,530 Continuation US9003029B2 (en) 2010-10-30 2012-12-20 Method, apparatus and system for joint optimization

Publications (1)

Publication Number Publication Date
WO2011144178A1 true WO2011144178A1 (zh) 2011-11-24

Family

ID=44268718

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2011/075388 WO2011144178A1 (zh) 2010-10-30 2011-06-07 联合优化的方法、装置和系统

Country Status (6)

Country Link
US (1) US9003029B2 (zh)
EP (1) EP2515478B1 (zh)
CN (1) CN102130824B (zh)
BR (1) BR112012022286A2 (zh)
RU (1) RU2520354C2 (zh)
WO (1) WO2011144178A1 (zh)

Families Citing this family (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9686232B2 (en) * 2012-06-25 2017-06-20 Connectify, Inc. Network address translating router for mobile networking
CN103974082B (zh) * 2013-01-25 2018-09-21 华为技术有限公司 子节点、父节点以及用于多层次视频网络的缓存方法及系统
US9386353B2 (en) 2013-01-25 2016-07-05 Huawei Technologies Co., Ltd. Child node, parent node, and caching method and system for multi-layer video network
EP2963970B1 (en) 2013-03-28 2022-01-05 Huawei Technologies Co., Ltd. Method and device for allocating bandwidth, user equipment and base station
CN103200113B (zh) * 2013-04-02 2016-04-06 北京邮电大学 兼顾运营成本和传输性能双优化的域间流量工程的实现方法
US10313232B2 (en) * 2015-03-06 2019-06-04 Nec Corporation Network control device, network control method, and recording medium for program
CN106293897B (zh) * 2015-05-15 2021-11-30 株式会社日立制作所 组件自动化调度系统
CN105515831A (zh) * 2015-11-27 2016-04-20 小米科技有限责任公司 网络状态信息展示方法及装置
CN107438013B (zh) * 2016-05-27 2022-05-17 中兴通讯股份有限公司 一种端口优化方法、装置及系统
CN106302221B (zh) * 2016-09-12 2019-09-10 中国联合网络通信集团有限公司 基于端局云化的流量调度方法和系统
RU2680764C1 (ru) * 2018-03-06 2019-02-26 Федеральное государственное казенное военное образовательное учреждение высшего образования Академия Федеральной службы охраны Российской Федерации Способ, устройство и система для оптимизации транспортной сети связи
CN111884854B (zh) * 2020-07-29 2022-09-02 中国人民解放军空军工程大学 基于多模式混合预测的虚拟网络流量迁移方法
CN113630456B (zh) * 2020-08-05 2022-07-05 北京航空航天大学 基于拉格朗日对偶分解和纳什议价博弈的域间合作缓存方法

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1688133A (zh) * 2005-04-11 2005-10-26 西安电子科技大学 传送网的资源利用优化方法

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7403482B2 (en) * 2000-10-30 2008-07-22 Nec Corporation Path provisioning for service level agreements in differentiated service networks
JP5138847B2 (ja) * 2001-08-31 2013-02-06 富士通株式会社 ネットワークシステム、ネットワーク中継装置、ネットワーク中継監視装置およびネットワーク運用方法
SE523714C2 (sv) 2002-07-05 2004-05-11 Packetfront Sweden Ab Ett filter i ett gränssnitt inom ett öppet system av typ skikt2 för trafikseparation i minst en router för åtkomstomkoppling inom ett nät, och ett förfarande för detta
US7292542B2 (en) * 2003-03-05 2007-11-06 At&T Bls Intellectual Property, Inc. Method for traffic engineering of connectionless virtual private network services
CN101541025B (zh) * 2009-04-28 2011-01-19 北京邮电大学 具有认知功能的无线通信网络系统
CN101707788B (zh) * 2009-10-27 2014-04-02 北京邮电大学 基于差异化定价策略的多层网络业务动态规划方法
CN101753450B (zh) * 2009-12-21 2012-09-05 西安电子科技大学 三层网络联合资源优化方法
US8688775B2 (en) * 2010-05-28 2014-04-01 Juniper Network, Inc. Application-layer traffic optimization service spanning multiple networks

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1688133A (zh) * 2005-04-11 2005-10-26 西安电子科技大学 传送网的资源利用优化方法

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
D. DIPALANTINO ET AL.: "Traffic Engineering vs. Content Distribution: A Game Theoretic Perspective", PROC. IEEE INFOCOM 2009, 30 December 2009 (2009-12-30) *
See also references of EP2515478A4 *
W. JIANG ET AL.: "Cooperative Content Distribution and Traffic Engineering in an ISP network", SIGMETRICS/PERFORMANCE'09, JUNE 15-19,2009, 19 June 2009 (2009-06-19), SEATTLE, WA, USA *

Also Published As

Publication number Publication date
EP2515478A4 (en) 2012-11-28
US20130117443A1 (en) 2013-05-09
RU2520354C2 (ru) 2014-06-20
EP2515478A1 (en) 2012-10-24
CN102130824A (zh) 2011-07-20
BR112012022286A2 (pt) 2018-05-15
US9003029B2 (en) 2015-04-07
RU2012137874A (ru) 2014-03-10
EP2515478B1 (en) 2016-04-27
CN102130824B (zh) 2014-09-17

Similar Documents

Publication Publication Date Title
WO2011144178A1 (zh) 联合优化的方法、装置和系统
Xu et al. Incremental deployment and throughput maximization routing for a hybrid SDN
EP3005620B1 (en) A method and system of bandwidth-aware service placement for service chaining
JP5975083B2 (ja) 通信システム、制御装置、パケット転送経路の制御方法およびプログラム
He et al. Achieving near-optimal traffic engineering in hybrid software defined networks
WO2019218812A1 (zh) 物理光网络虚拟化映射方法、装置、控制器及存储介质
JP2019122040A (ja) ネットワークソースリユース及びソフトウェアによりマルチソースを定義するルーティングメカニズム
Nguyen et al. Environment-aware virtual slice provisioning in green cloud environment
Attia et al. QoS-aware software-defined routing in smart community network
He et al. Performance of multipath in fiber-wireless (FiWi) access network with network virtualization
Ndikumana et al. Federated learning assisted deep Q-learning for joint task offloading and fronthaul segment routing in open RAN
Shirmarz et al. A novel flow routing algorithm based on non-dominated ranking and crowd distance sorting to improve the performance in SDN
CN109286563B (zh) 一种数据传输的控制方法和装置
Huang et al. Utility-optimized flow-level bandwidth allocation in hybrid SDNs
CN103200113B (zh) 兼顾运营成本和传输性能双优化的域间流量工程的实现方法
Elzain et al. QoS-aware topology discovery in decentralized software defined wireless mesh network (D-SDWMN) architecture
Oki et al. F-TPR: Fine two-phase IP routing scheme over shortest paths for hose model
García et al. Multicast routing and virtual network function placement in NFV-SDN networks: A genetic algorithms approach
CN115102831A (zh) 一种分布式bgp服务的部署方法和系统
Zhang et al. Q-SR: An extensible optimization framework for segment routing
ELseuofi Quality of service using PSO algorithm
Zhao et al. Hybrid routing by joint optimization of per-flow routing and tag-based routing in software-defined networks
Li et al. Self-adaptive genetic algorithm for LTE Backhaul network
Xu et al. An implementation of an intelligent PCE-agent-based multi-domain optical network architecture
Zhang et al. Multi-AS cooperative incoming traffic engineering in a transit-edge separate internet

Legal Events

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

Ref document number: 11783095

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 1761/KOLNP/2012

Country of ref document: IN

Ref document number: 2011783095

Country of ref document: EP

WWE Wipo information: entry into national phase

Ref document number: 2012137874

Country of ref document: RU

NENP Non-entry into the national phase

Ref country code: DE

REG Reference to national code

Ref country code: BR

Ref legal event code: B01A

Ref document number: 112012022286

Country of ref document: BR

ENP Entry into the national phase

Ref document number: 112012022286

Country of ref document: BR

Kind code of ref document: A2

Effective date: 20120904