CN109474696A - A kind of network service method, device, electronic equipment and readable storage medium storing program for executing - Google Patents

A kind of network service method, device, electronic equipment and readable storage medium storing program for executing Download PDF

Info

Publication number
CN109474696A
CN109474696A CN201811504958.9A CN201811504958A CN109474696A CN 109474696 A CN109474696 A CN 109474696A CN 201811504958 A CN201811504958 A CN 201811504958A CN 109474696 A CN109474696 A CN 109474696A
Authority
CN
China
Prior art keywords
isp
cache node
user
value
request
Prior art date
Legal status (The legal status 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 status listed.)
Granted
Application number
CN201811504958.9A
Other languages
Chinese (zh)
Other versions
CN109474696B (en
Inventor
张宇超
王文东
阙喜戎
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing University of Posts and Telecommunications
Original Assignee
Beijing University of Posts and Telecommunications
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 Beijing University of Posts and Telecommunications filed Critical Beijing University of Posts and Telecommunications
Priority to CN201811504958.9A priority Critical patent/CN109474696B/en
Publication of CN109474696A publication Critical patent/CN109474696A/en
Application granted granted Critical
Publication of CN109474696B publication Critical patent/CN109474696B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • 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
    • H04L67/51Discovery or management thereof, e.g. service location protocol [SLP] or web services
    • 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
    • H04L67/56Provisioning of proxy services
    • H04L67/568Storing data temporarily at an intermediate stage, e.g. caching
    • 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
    • H04L67/56Provisioning of proxy services
    • H04L67/568Storing data temporarily at an intermediate stage, e.g. caching
    • H04L67/5682Policies or rules for updating, deleting or replacing the stored data
    • 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

Landscapes

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

Abstract

The embodiment of the invention provides a kind of network service method, device, electronic equipment and readable storage medium storing program for executing, applied to Internet technical field, the described method includes: extracting the key feature in each user's request, according to key feature and the corresponding relationship of the ISP network group and key feature that pre-establish, determine that each user requests corresponding ISP network group;For each user request in each ISP network group, according to deeply learning algorithm, the step of choosing ISP cache node, calculating the income of the ISP network group and update policing parameter is executed by circuiting sequentially, it determines whether there is the maximum revenue for making ISP network group and meets the target ISP cache node of user's request condition, if it exists, user's request is responded by target ISP cache node and transmits request data to user;Otherwise user's request is responded by CDN and transmits request data to user.Network service quality can be improved in the present invention.

Description

A kind of network service method, device, electronic equipment and readable storage medium storing program for executing
Technical field
The present invention relates to Internet technical field, more particularly to a kind of network service method, device, electronic equipment and can Read storage medium.
Background technique
With the high speed development of internet, CDN (Content Deliver Network, content distributing network) gradually at For carrying video, live streaming, the internets innovation and application such as VR (Virtual Reality, virtual reality) Infrastructure support platform and Network speed-raising main force amount, importance increasingly improve.Nowadays internet data is with annual 50% rate rapid growth, according to Gartner (companies of first hand information technology research and analysis) prediction, arrives the year two thousand twenty, and global metadata amount of storage is up to 35Zettabyte (10,000,100,000,000 byte).Therefore, how efficiently to carry out content caching and distribution becomes academia and industry Focusing hot spot.
In the related technology, content and service provider bridges ISP by providing Intelligent hardware to ten million domestic consumer (Internet Service Provider, Internet Service Provider) network, thus by content caching to thousands of ISP In cache node, so as to which using the idle bandwidth of ISP cache node, data to be quickly and efficiently distributed to request source. This method is by being guided to ISP cache node for user's request, using ISP idle bandwidth, to alleviate the flow of data center Burst pressure saves the network bandwidth of content distributing network data center.Due to the cache contents distance request of ISP cache node The distance of source is closer, is transmitted compared to from content distributing network remote data center, and data transmission efficiency can obtain greatly Width is promoted, to request user the Service Promotion of response.In addition, the slack resources of ISP can effectively be used, by CDN Infrastructure cost be down to industry floor level, provide a newest skill for the big demand industry of the CDN such as live streaming, VR, 4K TV Art, least cost solution.
But this method of servicing has also caused new challenge, such as: (1) performance of ISP cache node is unstable.Compared to For data center server, in terms of available bandwidth, service success rate, the connection that can be carried, ISP minor node Performance reliability is greatly reduced, and service jitter phenomenon is fairly obvious;(2) buffer update of ISP cache node.Due to memory space Limited, the cache contents deficiency of ISP cache node will become another problem.It is outmoded that renewal speed will lead to data slowly excessively, utilizes Rate is low, and the too fast data transmission that can frequently cause from server to node of renewal speed, increases network flow, reduces node Data buffer storage utilization rate.As it can be seen that the service quality of existing network service method is lower.
Summary of the invention
The embodiment of the present invention is designed to provide a kind of network service method, device, electronic equipment and readable storage medium Matter, to improve network service quality.Specific technical solution is as follows:
The embodiment of the invention provides a kind of network service methods, which comprises
Extract the key feature in acquired each user request, according to the key feature in each user's request with And the corresponding relationship of the Internet Service Provider's ISP network group and key feature pre-established, determine each user's request pair The ISP network group answered includes at least one ISP cache node in each ISP network group;
For each ISP network group, each user in the ISP network group is requested, learns to calculate according to deeply Method chooses ISP cache node, calculates the income of the ISP network group and updates the step of policing parameter by circuiting sequentially to execute Suddenly, it is determined whether there is the maximum revenue for making the ISP network group and meet the target ISP cache node of user's request condition;
If it does, being responded by the target ISP cache node to user request and transmitting request data extremely The user;
If it does not, being responded by content distributing network CDN to user request and transmitting request data extremely The user.
Optionally, user request is responded and transmits number of request by the target ISP cache node described According to after the user, the method also includes:
Residue after the bandwidth resources of the target ISP cache node to be updated to transmit the request data can use band Width, and the available bandwidth in the network state of the target ISP cache node is updated to the residual available bandwidth, it is network-like Whether state includes: available bandwidth and includes the request data.
Optionally, described according to deeply learning algorithm, ISP cache node, calculating are chosen by circuiting sequentially to execute The step of income and update policing parameter of the ISP network group, it is determined whether there is the maximum revenue for making the ISP network group And meet the target ISP cache node of user's request condition, comprising:
The corresponding study of ISP network group is arranged in the state parameter for obtaining each ISP cache node in the ISP network group The parameter of model, the parameter of the learning model include: value threshold value and cycle-index;
Step A, according to initial policy parameter π (a | s0), choose ISP cache node, and the initial policy parameter π (a | s0) In original state s0Probability when each ISP cache node of lower selection is impartial;
Wherein, original state s0It indicates not yet to start the state to receive an assignment when requesting, a after the deployment of ISP cache node It indicates to match user request selection ISP cache node and data is transmitted;
Step B obtains the r value of the ISP network group, the r value according to the state parameter of selected ISP cache node Including pre-set big value and small value, the big value indicates that selected ISP cache node has the item of response user's request Part, the small value indicate that selected ISP cache node does not have the condition of response user's request;
According to cost function: V (s, a)=Eπ(a|s)[r+ γ W (s ')] determines the valence at state s after selection operation a Value V (s, a);
Wherein, (s a) indicates that the profit expectation of the selection operation a at state s, s indicate all ISP in current environment group to V The network state after operation a, policing parameter π (a | s) are taken in the state of cache node, s ' expression at current network state s Indicate that the probability of the selection operation a at state s, r indicate the reward value of the ISP network group at policing parameter π (a | s), γ table Show that discount rate, the value of γ are 0-1, W (s ') indicates the maximum value reached under next state s ', and r+ γ W (s ') is operation Value function;
Step C, according to advantage function: A (s, a)=Q (s, a)-V (s, a), it is determined whether selection operation a;
((s a) indicates to perform operation a at state s Q by s, a)=r+ γ W (s '), Q;
If (s a) is greater than 0, selection operation a to A, executes step D;
If (s a) is not more than 0, not selection operation a, return step A to A;
Step D, according to formula:It calculates from original state s0To state s institute it is stateful under ISP network group value average value, whether the average value judged be less than the value threshold value;
If obtained average value is less than the value threshold value, and the number for choosing ISP cache node is less than cycle-index, Execute step E;If obtained average value is less than the value threshold value, and the number for choosing ISP cache node is equal to circulation time Number determines that there is no the mesh of the maximum revenue for making the ISP network group and satisfaction response user's request condition in the ISP network group Mark ISP cache node;
If obtained average value is not less than the value threshold value, the ISP cache node of final choice is determined as target ISP cache node;
Step E, according to gradient function:
Determination makesChange most fast direction, more New strategy parameter π (a | s), return step B.
Optionally, it is described by CDN to the user request respond and transmit request data to the user after, The method also includes:
The request data is cached into the ISP cache node of the final choice, and by the ISP of the final choice Whether it is updated to by not including the request data comprising described comprising the request data in the network state of cache node Request data.
Optionally, after stepb, the method also includes:
Policing parameter π (a | s) is sent to parameter server by the ISP network group, and the parameter server is according to receiving Itself covering the policing parameter that sends of all ISP network groups, the policing parameter π of parameter server (a | s) is updated, And updated policing parameter π (a | s) is sent to all ISP network groups of itself covering.
The embodiment of the invention provides a kind of Web services apparatus, described device includes:
Grouping module, for extracting the key feature in acquired each user request, according in each user's request Key feature and the Internet Service Provider's ISP network group and key feature that pre-establish corresponding relationship, determine described in Each user requests corresponding ISP network group, includes at least one ISP cache node in each ISP network group;
Node determining module requests each user in the ISP network group, root for being directed to each ISP network group According to deeply learning algorithm, by circuit sequentially execute choose ISP cache node, calculate the ISP network group income and The step of updating policing parameter, it is determined whether there is the maximum revenue for making the ISP network group and meet user's request condition Target ISP cache node;
Node data transmission module passes through for determining in the node determining module there are when target ISP cache node The target ISP cache node responds user request and transmits request data to the user;
Network data transmission module is led to for determining in the node determining module there is no when target ISP cache node Content distributing network CDN is crossed to respond user request and transmit request data to the user.
Optionally, the Web services apparatus of the embodiment of the present invention, further includes:
Bandwidth update module, for being updated to transmit the number of request by the bandwidth resources of the target ISP cache node Residual available bandwidth after, and the available bandwidth in the network state of the target ISP cache node is updated to described remain Whether remaining available bandwidth, network state include: available bandwidth and include the request data.
Optionally, the node determining module is used to obtain the state ginseng of each ISP cache node in the ISP network group Number, is arranged the parameter of the corresponding learning model of ISP network group, and the parameter of the learning model includes: value threshold value and circulation Number;
Step A, according to initial policy parameter π (a | s0), choose ISP cache node, and the initial policy parameter π (a | s0) In original state s0Probability when each ISP cache node of lower selection is impartial;
Wherein, original state s0It indicates not yet to start the state to receive an assignment when requesting, a after the deployment of ISP cache node It indicates to match user request selection ISP cache node and data is transmitted;
Step B obtains the r value of the ISP network group, the r value according to the state parameter of selected ISP cache node Including pre-set big value and small value, the big value indicates that selected ISP cache node has the item of response user's request Part, the small value indicate that selected ISP cache node does not have the condition of response user's request;
According to cost function: V (s, a)=Eπ(a|s)[r+ γ W (s ')] determines the valence at state s after selection operation a Value V (s, a);
Wherein, (s a) indicates that the profit expectation of the selection operation a at state s, s indicate all ISP in current environment group to V The network state after operation a, policing parameter π (a | s) are taken in the state of cache node, s ' expression at current network state s Indicate that the probability of the selection operation a at state s, r indicate the reward value of the ISP network group at policing parameter π (a | s), γ table Show that discount rate, the value of γ are 0-1, W (s ') indicates the maximum value reached under next state s ', and r+ γ W (s ') is operation Value function;
Step C, according to advantage function: A (s, a)=Q (s, a)-V (s, a), it is determined whether selection operation a;
((s a) indicates to perform operation a at state s Q by s, a)=r+ γ W (s '), Q;
If (s a) is greater than 0, selection operation a to A, executes step D;
If (s a) is not more than 0, not selection operation a, return step A to A;
Step D, according to formula:It calculates from original state s0To state s institute it is stateful under ISP network group value average value, whether the average value judged be less than the value threshold value;
If obtained average value is less than the value threshold value, and the number for choosing ISP cache node is less than cycle-index, Execute step E;If obtained average value is less than the value threshold value, and the number for choosing ISP cache node is equal to circulation time Number determines that there is no the mesh of the maximum revenue for making the ISP network group and satisfaction response user's request condition in the ISP network group Mark ISP cache node;
If obtained average value is not less than the value threshold value, the ISP cache node of final choice is determined as target ISP cache node;
Step E, according to gradient function:
Determination makesChange most fast direction, more New strategy parameter π (a | s), return step B.
Optionally, the Web services apparatus of the embodiment of the present invention, further includes:
Data cached update module, for caching the request data into the ISP cache node of the final choice, And by the network state of the ISP cache node of the final choice whether comprising the request data by not including described ask Data are asked to be updated to comprising the request data.
Optionally, the Web services apparatus of the embodiment of the present invention, further includes:
Policing parameter update module, for making the ISP network group that policing parameter π (a | s) is sent to parameter server, institute The policing parameter that parameter server is sent according to all ISP network groups of itself covering received is stated, to parameter server Policing parameter π (a | s) is updated, and updated policing parameter π (a | s) is sent to all ISP networks of itself covering Group.
The embodiment of the invention provides a kind of electronic equipment, comprising: processor, communication interface, memory and communication bus, Wherein, the processor, the communication interface, the memory complete mutual communication by the communication bus;
The memory, for storing computer program;
The processor when for executing the program stored on the memory, realizes any of the above-described network The step of method of servicing.
The embodiment of the invention provides a kind of computer readable storage medium, storage in the computer readable storage medium There is computer program, when the computer program is executed by processor, realizes the step of any of the above-described network service method Suddenly.
Network service method, device, electronic equipment and readable storage medium storing program for executing provided in an embodiment of the present invention, by extracting institute The key feature in each user request obtained, according to the key feature in each user request and the ISP network pre-established The corresponding relationship of group and key feature determines that each user requests corresponding ISP network group;For each ISP network group, for this Each user request in ISP network group chooses ISP caching section by circuiting sequentially to execute according to deeply learning algorithm The step of point, the income for calculating the ISP network group and update policing parameter, it is determined whether there are the receipts for making the ISP network group Benefit maximizes and meets the target ISP cache node of user's request condition;If it does, by target ISP cache node to this User's request responds and transmits request data to the user;If it does not, being responded simultaneously by CDN to user's request Request data is transmitted to the user.The embodiment of the present invention carries out intelligent networking and overall scheduling by deeply learning algorithm, To which using the idle bandwidth of ISP cache node, request data is distributed to user by ISP cache node.The present invention can lead to The method of inflow-rate of water turbine transfer, accelerates content distribution speed, shortens user's request response time, improves network service quality.Certainly, It implements any of the products of the present invention or method does not necessarily require achieving all the advantages described above at the same time.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with It obtains other drawings based on these drawings.
Fig. 1 is the network service architectures figure of the embodiment of the present invention;
Fig. 2 is the flow chart of the network service method of the embodiment of the present invention;
Fig. 3 is the flow chart that the ISP cache node of the embodiment of the present invention is chosen;
Fig. 4 is the structure chart of the Web services apparatus of the embodiment of the present invention;
Fig. 5 is the structure chart of the electronic equipment of the embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other Embodiment shall fall within the protection scope of the present invention.
In order to solve the problems, such as that network service quality is low during ISP network node auxiliary CDN, the embodiment of the present invention is provided A kind of network service method, device, electronic equipment and readable storage medium storing program for executing, to improve network service quality.
Network service method is provided for the embodiments of the invention first below to describe in detail.
The embodiment of the present invention builds CDN using the method for ISP auxiliary, specifically, CDN is by disposing ISP to home network Cache node, bridge home router, thus by content caching into thousands of ISP cache node.ISP cache node is made For CDN endpoint node, to be collected into the idle bandwidth of domestic consumer with shared mode.It is that the present invention is real referring to Fig. 1, Fig. 1 Apply the network service architectures figure of example, it can be seen that, can be sufficiently sharp by expanding the endpoint node (i.e. ISP cache node) of CDN With the tip bandwidth of ISP cache node, the quick response of last one kilometer is realized.When client issues file request, the net Network can use cooperate throughout a ISP cache nodes up to a million throughout the country the backbone network of content supplier together into Row request response and content transmission.
Referring to fig. 2, Fig. 2 is the flow chart of the network service method of the embodiment of the present invention, comprising the following steps:
S201 extracts the key feature in acquired each user request, according to the key feature in each user request with And the corresponding relationship of the Internet Service Provider's ISP network group and key feature pre-established, determine that each user's request is corresponding ISP network group includes at least one ISP cache node in each ISP network group.
In the embodiment of the present invention, the number of ISP cache node can achieve up to a million or more, and correspondingly, user asks The number asked can also have very much, therefore ISP cache node can be divided into different ISP network groups in advance, make different ISP Group of networks handles different user's requests.It is, establishing the corresponding relationship of ISP network group and user's request, specially ISP net The corresponding relationship of network group and the key feature in user's request.For example, the key feature in user's request may include: source IP (Internet Protocol, Internet protocol) address, request data, the size of request data and request response cut-off when Between etc..So ISP cache node graduation can be divided into different ISP network groups by region different (different provinces).It extracts and uses Different user's requests can be mapped to different ISP network groups according to source IP address by the source IP address in the request of family In.
S202 requests each user in the ISP network group, for each ISP network group according to deeply Algorithm is practised, ISP cache node is chosen by circuiting sequentially to execute, calculate the income of the ISP network group and updates policing parameter The step of, it is determined whether there is the maximum revenue for making the ISP network group and meets the target ISP caching section of user's request condition Point.If it does, executing S203;If it does not, executing S204.
Specifically, being directed to each ISP network group, when receiving user's request, need to request for user specified corresponding ISP cache node is serviced, and different ISP cache node bring service effectivenesses is selected to be different, this step needs Request matching pair the ISP cache node that can maximize overall performance income-user can be being found in the convergent time, so as to While accelerating request response time, increase the maximum stream flow that ISP cache node cluster can carry.The embodiment of the present invention passes through Deeply learning algorithm determines whether there is the maximum revenue for making the ISP network group and meets the target of user's request condition ISP cache node will hereafter be described in detail deeply learning algorithm.
S203 responds user request by target ISP cache node and transmits request data to the user.
In the embodiment of the present invention, if there is target ISP cache node, show that target ISP cache node has user The data of request, remaining bandwidth are enough to complete data transmission at the appointed time, and in stable condition, data transmission success is higher. So, user request is responded by target ISP cache node and transmits request data to the user.
S204 responds user's request by CDN and transmits request data to the user.
It, can be in CDN and ISP in the case where ISP cache node is not able to satisfy user's request in the embodiment of the present invention Data are carried out between cache node and return source operation, i.e., are carried out data transmission from CDN to ISP cache node.Specifically, according to user Key feature in request carries out file search in the backbone network server of CDN, corresponding request data is transmitted to finally In the ISP cache node of selection, meanwhile, user request is responded by CDN, which can skip over the user and ask It asks, carries out subsequent response operation.
The network service method of the embodiment of the present invention, by extracting the key feature in acquired each user request, root According to pair of key feature and the Internet Service Provider's ISP network group and key feature that pre-establish in each user request It should be related to, determine that each user requests corresponding ISP network group;For each ISP network group, for every in the ISP network group A user's request chooses ISP cache node by circuiting sequentially execution, calculates the ISP network according to deeply learning algorithm The step of income and update policing parameter of group, it is determined whether there is the maximum revenue for making the ISP network group and meet and use The target ISP cache node of family request condition;If it does, being responded by target ISP cache node to user request And request data is transmitted to the user.Extremely should if it does not, being responded by CDN to user's request and transmitting request data User.The embodiment of the present invention carries out intelligent networking and overall scheduling by deeply learning algorithm, saves to be cached using ISP The idle bandwidth of point, is distributed to user by ISP cache node for request data.The present invention can by the method for load transfer, Accelerate content distribution speed, shorten user's request response time, improves network service quality.
In a kind of implementation of the invention, after Fig. 2 embodiment S203, the bandwidth of target ISP cache node is provided Source is updated to the residual available bandwidth after transmission request data, and by the available band in the network state of target ISP cache node Width is updated to residual available bandwidth, and whether network state includes: available bandwidth and include request data.
More specifically, each ISP cache node can have certain network state, and network state includes: available band It is wide and whether include request data.After ISP cache node requests response to user, request data is transmitted to user, at this point, Since the available bandwidth of ISP cache node changes, the bandwidth resources of ISP cache node can be updated, i.e., by ISP The bandwidth resources of cache node are updated to initial bandwidth and subtract bandwidth used in transmission request data.Wherein, number of request is transmitted According to used bandwidth can be requested according to user in key feature calculate.Correspondingly, the network state of ISP cache node It will change, it is possible to which the available bandwidth in network state is updated to residual available bandwidth.Certainly, if ISP is cached Node does not respond any user request, and the network state of the ISP cache node is constant.In this way, by right in time The state of ISP cache node is updated, and when determining target ISP cache node by deeply learning algorithm, improves mesh The accuracy that ISP cache node is chosen, and then the accuracy of improve data transfer are marked, to improve network service quality.
In a kind of implementation of the invention, Fig. 3, including following step can be found in the detailed process of Fig. 2 embodiment S202 It is rapid:
S301 obtains the state parameter of each ISP cache node in the ISP network group, and it is corresponding that the ISP network group is arranged The parameter of learning model, the parameter of learning model include: value threshold value and cycle-index.
In the embodiment of the present invention, during being learnt by deeply learning algorithm to model, include at least with Lower parameter:
State s, i.e., the state parameter of all ISP cache nodes in current ISP network group, comprising: ISP cache node Whether ID, memory capacity size are connected to as available mode, stored file, residual available bandwidth, historical data transmission success Rate etc..
A is operated, i.e., the user's request one ISP cache node of selection being currently received is matched and data is transmitted. ISP cache node K are shared in ISP network group, then the selection that operation has K kind different each time.
Policing parameter π (a | s) indicate the probability that operation a is had chosen at state s.
Income r, i.e. reward mean the financial value of the system at tactful π.
Therefore, in this step, the state parameter of available each ISP cache node further includes certainly in ISP network group The number of ISP cache node.Later, model is learnt according to the state parameter of each ISP cache node.Furthermore it is also possible to The parameter of learning model is set, comprising: value threshold value and cycle-index, value threshold value refer to making the income of ISP network group most When bigization, in learning process it is stateful under ISP network group value average value minimum value, cycle-index refers to The number of ISP cache node can be chosen in learning process.
S302, according to initial policy parameter π (a | s0), selection ISP cache node, and initial policy parameter π (a | s0) initial State s0Probability when each ISP cache node of lower selection is impartial.
Wherein, original state s0It indicates not yet to start the state to receive an assignment when requesting after the deployment of ISP cache node, it is deep Spend nitrification enhancement initial period, strategy be it is random, that is, choose ISP cache node be random.
S303 obtains the r value of the ISP network group according to the state parameter of selected ISP cache node, and r value includes pre- The big value being first arranged and small value, big value indicate that selected ISP cache node has the condition of response user's request, and small value indicates Selected ISP cache node does not have the condition of response user's request.
In the embodiment of the present invention, if having chosen inappropriate ISP cache node, i.e., ISP cache node does not have response and uses The condition of family request is not (for example, having request data, having request data but bandwidth is insufficient or has request data bandwidth fills Foot but ISP cache node state labile failure rate is very high), then r value is lower;If having chosen suitable ISP cache node, i.e., ISP cache node has the condition of response user's request, then the r value is higher.Therefore, r value can be set in advance as to two values: Big value and small value determine that r value is in this way, whether can have the condition of response user's request according to the ISP cache node of selection Big value or small value.
Later, according to cost function: V (s, a)=Eπ(a|s)[r+ γ W (s ')] is determined at state s after selection operation a Value V (s, a);
Wherein, (s a) indicates that the profit expectation of the selection operation a at state s, s indicate all ISP in current environment group to V The network state after operation a, policing parameter π (a | s) are taken in the state of cache node, s ' expression at current network state s Indicate that the probability of the selection operation a at state s, r indicate the reward value of the ISP network group at policing parameter π (a | s), γ table Show that discount rate, the value of γ are 0-1, W (s ') indicates the maximum value reached under next state s ', and r+ γ W (s ') is operation Value function.
By V, (s a) is defined as iteration form, indicates that current state s can be obtained being worth, is next state s ' institute energy The sum of the value of acquisition and the income obtained in transfer process, the profit expectation for enabling NextState s' be fed back is multiplied by discount Rate indicates the maximum value W (s ') that current r value is more important than the r value of future feedback, reaches under next state s ', refers to It is that under next state s ', inside all selection operations, selection operation a ' can make W (s ') maximum, but in practical selection In, it might not selection operation a '.
S304, according to advantage function: A (s, a)=Q (s, a)-V (s, a), it is determined whether selection operation a;
((s a) indicates to perform operation a at state s Q by s, a)=r+ γ W (s '), Q;
If (s a) is greater than 0, selection operation a to A, executes S305;
If (s a) is not more than 0, not selection operation a to A, returns to S302;
Specifically, can be determined by advantage function to indicate whether operation a better than current average case, if A (s, A) it is greater than 0, indicates that selection operation a is better than current average case, then selection operation a;Otherwise, not selection operation a.In addition, After having chosen operation a, policing parameter π (a | s) is sent to parameter server by ISP network group, and parameter server is according to reception The policing parameter that all ISP network groups for itself covering arrived are sent carries out more the policing parameter π of parameter server (a | s) Newly, and by updated policing parameter π (a | s) all ISP network groups of itself covering are sent to.For example, 1 He of ISP network group ISP network group 2 respectively corresponds different policing parameters, i.e., removes the different shapes for being transferred to next step at state s with much probability State.So, the policing parameter of parameter server can be averaging according to ISP network group 1 and the respective probability of ISP network group 2, be come Update ISP network group 1 and ISP network group 2.
S305, according to formula:It calculates from original state s0To state s it is stateful under Whether the average value of the value of ISP network group, the average value judged are less than value threshold value.If obtained average value is less than It is worth threshold value, and the number for choosing ISP cache node is less than cycle-index, shows that the ISP cache node chosen is unsatisfactory for matching Condition, then executing S306.
S306, according to gradient function:
Determination makesChange most fast direction, Update policing parameter π (a | s), return to S303.
S307, if obtained average value is less than value threshold value, and the number for choosing ISP cache node is equal to circulation time Number determines that there is no the mesh of the maximum revenue for making the ISP network group and satisfaction response user's request condition in the ISP network group Mark ISP cache node.
In the embodiment of the present invention, if the number for choosing ISP cache node is equal to cycle-index, show to choose ISP node The number average value that has reached the maximum value, and obtain at this time be less than value threshold value, determine that there is no make the ISP network group Maximum revenue and the target ISP cache node for meeting response user's request condition.
S308, if the ISP cache node of final choice is determined as target not less than value threshold value by obtained average value ISP cache node.
As it can be seen that by the Matching Model of building ISP cache node and user's request, according to deeply learning algorithm intelligence Optimal service path is dynamically chosen, to improve the data transmission bauds of CDN, improves network service quality.
In a kind of implementation of the invention, after Fig. 3 embodiment S307, request data can also be cached to final In the ISP cache node of selection, and by the network state of the ISP cache node of final choice whether comprising request data by It is updated to not comprising request data comprising request data.
In the embodiment of the present invention, after CDN carries out data transmission to ISP cache node, by updating ISP cache node The hit rate of follow-up data access can be improved in interior cache file.
Embodiment one
If in ISP network group including 5 ISP cache nodes, data cached file situation is as shown in table 1 respectively.
Table 1
ISP cache node Cache file
1 A,B
2 A,C
3 B,D
4 C,E
5 D,E
Wherein, value threshold value is set as 5, and cycle-index is set as the big value in 3, r value and is set as 10, and small value is set as 0.1。
If receiving user request<132.147.34.68, D, 10G, 100s>, indicate that IP address is 132.147.34.68's It is the file D of 10G that user, which requests size, needs to complete transmission in 100s.
At this point, carrying out ISP cache node pre-selection, it is assumed that the first round has chosen No. 2 ISP cache nodes with 0.2 probability, π (2 | s)=0.2, it is contemplated that demand file D is not present in No. 2 ISP cache nodes, then r value is set as 0.1, at this time J=0.1, and Since No. 3 and No. 5 ISP cache nodes have demand file D, i.e. r value is higher, so that gradient functionIt is slow at No. 3 and No. 5 ISP Acquirement higher gradient at node is deposited, therefore when next round is chosen, improves the probability for choosing No. 3 and No. 5 ISP cache nodes.
Assuming that No. 3 ISP cache nodes are had chosen with 0.4 probability, π (3 | s)=0.4, read ISP cache node remnant information Bandwidth resources are 150M/s, can satisfy the deadline requirement of current request, then r value are set as 10, at this time J=5.05 > value Threshold value, and the number for choosing ISP cache node is 2, less than 3, therefore meets matching condition, No. 3 ISP cache nodes is returned to, to this Service request is responded, while the remaining bandwidth of No. 3 ISP cache nodes being adjusted are as follows:
Embodiment two
Receive user request<132.147.34.68, F, 2G, 100s>, indicate that IP address is the user of 132.147.34.68 Requesting size is the file F of 2G, needs to complete transmission in 100s.
At this point, carrying out ISP cache node pre-selection, it is assumed that the first round has chosen No. 2 ISP cache nodes with 0.2 probability, π (2 | s)=0.2, it is contemplated that demand file F is not present in No. 2 ISP cache nodes, then r value is set as 0.1, at this time J=0.1, and Since any ISP cache node in current environment group does not have demand file F, i.e. r value is lower, fails to obtain higher ladder Degree, therefore ISP cache node is still randomly selected in next round.
Assuming that actor has chosen No. 3 ISP cache nodes with 0.2 probability, π (3 | s)=0.2, it is contemplated that No. 3 ISP cachings Demand file F is also not present in node, then reward is set as 0.1, at this time J=0.1 < TH=5, then carries out third time selection.
Assuming that No. 4 ISP cache nodes are had chosen with 0.2 probability, π (4 | s)=0.2, it is contemplated that No. 4 ISP cache nodes Still demand file F is not present, then r value is set as 0.1, at this time J=0.1 < value threshold value, but chooses ISP cache node at this time Number be 3, it is therefore desirable to carry out data Hui Yuan.At this point, data file F is searched in raw content distribution network backbone network, and It is cached on No. 4 ISP cache nodes currently chosen, if No. 4 ISP cache node storage ISP cache nodes are insufficient, according to slow It deposits replacement policy and carries out file replacement (such as first in first out dispatching algorithm, minimum dispatching algorithm, recently least common scheduling recently Algorithm etc.).The nodal cache data cases of table 1 are updated simultaneously, and updated table 1 is table 2.
Table 2
ISP cache node Cache file
1 A,B
2 A,C
3 B,D
4 E,F
5 D,E
Corresponding to above method embodiment, the embodiment of the invention provides a kind of Web services apparatus, and referring to fig. 4, Fig. 4 is The structure chart of the Web services apparatus of the embodiment of the present invention, comprising:
Grouping module 401, for being requested according to each user for extracting the key feature in acquired each user request In key feature and the Internet Service Provider's ISP network group and key feature that pre-establish corresponding relationship, determine each User requests corresponding ISP network group, includes at least one ISP cache node in each ISP network group;
Node determining module 402 asks each user in the ISP network group for being directed to each ISP network group It asks, according to deeply learning algorithm, by circuiting sequentially the income for executing and choosing ISP cache node, calculating the ISP network group And the step of updating policing parameter, it is determined whether there is the maximum revenue for making the ISP network group and meet user and request item The target ISP cache node of part;
Node data transmission module 403 passes through for determining in node determining module there are when target ISP cache node Target ISP cache node responds user request and transmits request data to the user;
Network data transmission module 404 is led to for determining in node determining module there is no when target ISP cache node CDN is crossed to respond user's request and transmit request data to the user.
The Web services apparatus of the embodiment of the present invention, by extracting the key feature in acquired each user request, root According to pair of key feature and the Internet Service Provider's ISP network group and key feature that pre-establish in each user request It should be related to, determine that each user requests corresponding ISP network group;For each ISP network group, for every in the ISP network group A user's request chooses ISP cache node by circuiting sequentially execution, calculates the ISP network according to deeply learning algorithm The step of income and update policing parameter of group, it is determined whether there is the maximum revenue for making the ISP network group and meet and use The target ISP cache node of family request condition;If it does, being responded by target ISP cache node to user request And request data is transmitted to the user.Extremely should if it does not, being responded by CDN to user's request and transmitting request data User.The embodiment of the present invention carries out intelligent networking and overall scheduling by deeply learning algorithm, saves to be cached using ISP The idle bandwidth of point, is distributed to user by ISP cache node for request data.The present invention can by the method for load transfer, Accelerate content distribution speed, shorten user's request response time, improves network service quality.
Optionally, the Web services apparatus of the embodiment of the present invention, further includes:
Bandwidth update module, it is surplus after transmission request data for being updated to the bandwidth resources of target ISP cache node Remaining available bandwidth, and the available bandwidth in the network state of target ISP cache node is updated to residual available bandwidth, it is network-like Whether state includes: available bandwidth and includes request data.
Optionally, node determining module is used to obtain the state parameter of each ISP cache node in the ISP network group, if The parameter of the corresponding learning model of ISP network group is set, the parameter of learning model includes: value threshold value and cycle-index;
Step A, according to initial policy parameter π (a | s0), selection ISP cache node, and initial policy parameter π (a | s0) first Beginning state s0Probability when each ISP cache node of lower selection is impartial;
Wherein, original state s0It indicates not yet to start the state to receive an assignment when requesting, a after the deployment of ISP cache node It indicates to match user request selection ISP cache node and data is transmitted;
Step B obtains the r value of the ISP network group according to the state parameter of selected ISP cache node, and r value includes Pre-set big value and small value, big value indicate that selected ISP cache node has the condition of response user's request, small value table Show that selected ISP cache node does not have the condition of response user's request;
According to cost function: V (s, a)=Eπ(a|s)[r+ γ W (s ')] determines the valence at state s after selection operation a Value V (s, a);
Wherein, (s a) indicates that the profit expectation of the selection operation a at state s, s indicate all ISP in current environment group to V The network state after operation a, policing parameter π (a | s) are taken in the state of cache node, s ' expression at current network state s Indicate that the probability of the selection operation a at state s, r indicate the reward value of the ISP network group at policing parameter π (a | s), γ table Show that discount rate, the value of γ are 0-1, W (s ') indicates the maximum value reached under next state s ', and r+ γ W (s ') is operation Value function;
Step C, according to advantage function: A (s, a)=Q (s, a)-V (s, a), it is determined whether selection operation a;
((s a) indicates to perform operation a at state s Q by s, a)=r+ γ W (s '), Q;
If (s a) is greater than 0, selection operation a to A, executes step D;
If (s a) is not more than 0, not selection operation a, return step A to A;
Step D, according to formula:It calculates from original state s0To state s it is stateful under Whether the average value of the value of ISP network group, the average value judged are less than value threshold value;
If obtained average value is less than value threshold value, and the number for choosing ISP cache node is less than cycle-index, executes Step E;If obtained average value is less than value threshold value, and the number for choosing ISP cache node is equal to cycle-index, and determining should There is no the maximum revenue for making the ISP network group and the target ISP cachings for meeting response user's request condition in ISP network group Node;
If the ISP cache node of final choice is determined as target ISP and delayed by obtained average value not less than value threshold value Deposit node;
Step E, according to gradient function:
Determination makesChange most fast direction, more New strategy parameter π (a | s), return step B.
Optionally, the Web services apparatus of the embodiment of the present invention, further includes:
Data cached update module, for caching request data into the ISP cache node of final choice, and will be final Whether it is updated to by not including request data comprising request comprising request data in the network state of the ISP cache node of selection Data.
Optionally, the Web services apparatus of the embodiment of the present invention, further includes:
Policing parameter update module is joined for making the ISP network group that policing parameter π (a | s) is sent to parameter server The policing parameter that number server is sent according to all ISP network groups of itself covering received, to the strategy of parameter server Parameter π (a | s) is updated, and updated policing parameter π (a | s) is sent to all ISP network groups of itself covering.
The embodiment of the invention also provides a kind of electronic equipment, are the electronic equipment of the embodiment of the present invention referring to Fig. 5, Fig. 5 Structure chart, comprising: processor 501, communication interface 502, memory 503 and communication bus 504, wherein processor 501, logical Letter interface 502, memory 503 complete mutual communication by communication bus 504;
Memory 503, for storing computer program;
Processor 501 when for executing the program stored on memory 503, realizes any of the above-described network service method The step of.
It should be noted that the communication bus 504 that above-mentioned electronic equipment is mentioned can be PCI (Peripheral Component Interconnect, Peripheral Component Interconnect standard) bus or EISA (Extended Industry Standard Architecture, expanding the industrial standard structure) bus etc..The communication bus 504 can be divided into address bus, data/address bus, Control bus etc..Only to be indicated with a thick line in Fig. 5, it is not intended that an only bus or a seed type convenient for indicating Bus.
Communication interface 502 is for the communication between above-mentioned electronic equipment and other equipment.
Memory 503 may include RAM (Random Access Memory, random access memory), also may include Nonvolatile memory (non-volatile memory), for example, at least a magnetic disk storage.Optionally, memory may be used also To be storage device that at least one is located remotely from aforementioned processor.
Above-mentioned processor 501 can be general processor, comprising: CPU (Central Processing Unit, center Processor), NP (Network Processor, network processing unit) etc.;It can also be DSP (Digital Signal Processing, digital signal processor), ASIC (Application Specific Integrated Circuit, it is dedicated Integrated circuit), FPGA (Field-Programmable Gate Array, field programmable gate array) or other are programmable Logical device, discrete gate or transistor logic, discrete hardware components.
In the electronic equipment of the embodiment of the present invention, processor is obtained by executing the program stored on memory, extraction The key feature in each user request taken, according to the key feature in each user request and the network service pre-established The corresponding relationship of provider ISP network group and key feature determines that each user requests corresponding ISP network group;For each ISP Group of networks requests each user in the ISP network group, according to deeply learning algorithm, is executed by circuiting sequentially The step of choosing ISP cache node, calculating the income of the ISP network group and update policing parameter, it is determined whether existing makes this The maximum revenue of ISP network group and the target ISP cache node for meeting user's request condition;If it does, passing through target ISP Cache node responds user request and transmits request data to the user.If it does not, being asked by CDN to user It asks and is responded and transmitted request data to the user.The embodiment of the present invention carries out intelligent networking by deeply learning algorithm With overall scheduling, so that request data is distributed to user by ISP cache node using the idle bandwidth of ISP cache node.This Invention can accelerate content distribution speed by the method for load transfer, shorten user's request response time, improve network service Quality.
The embodiment of the invention also provides a kind of computer readable storage medium, it is stored in computer readable storage medium Computer program, when computer program is executed by processor, the step of realizing any of the above-described network service method.
When the instruction stored in the computer readable storage medium of the embodiment of the present invention is run on computers, extraction is obtained The key feature in each user request taken, according to the key feature in each user request and the network service pre-established The corresponding relationship of provider ISP network group and key feature determines that each user requests corresponding ISP network group;For each ISP Group of networks requests each user in the ISP network group, according to deeply learning algorithm, is executed by circuiting sequentially The step of choosing ISP cache node, calculating the income of the ISP network group and update policing parameter, it is determined whether existing makes this The maximum revenue of ISP network group and the target ISP cache node for meeting user's request condition;If it does, passing through target ISP Cache node responds user request and transmits request data to the user.If it does not, being asked by CDN to user It asks and is responded and transmitted request data to the user.The embodiment of the present invention carries out intelligent networking by deeply learning algorithm With overall scheduling, so that request data is distributed to user by ISP cache node using the idle bandwidth of ISP cache node.This Invention can accelerate content distribution speed by the method for load transfer, shorten user's request response time, improve network service Quality.
It should be noted that, in this document, relational terms such as first and second and the like are used merely to a reality Body or operation are distinguished with another entity or operation, are deposited without necessarily requiring or implying between these entities or operation In any actual relationship or order or sequence.Moreover, the terms "include", "comprise" or its any other variant are intended to Non-exclusive inclusion, so that the process, method, article or equipment including a series of elements is not only wanted including those Element, but also including other elements that are not explicitly listed, or further include for this process, method, article or equipment Intrinsic element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that There is also other identical elements in process, method, article or equipment including the element.
Each embodiment in this specification is all made of relevant mode and describes, same and similar portion between each embodiment Dividing may refer to each other, and each embodiment focuses on the differences from other embodiments.Especially for device, For electronic equipment and readable storage medium storing program for executing embodiment, since it is substantially similar to the method embodiment, so the comparison of description is simple Single, the relevent part can refer to the partial explaination of embodiments of method.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the scope of the present invention.It is all Any modification, equivalent replacement, improvement and so within the spirit and principles in the present invention, are all contained in protection scope of the present invention It is interior.

Claims (10)

1. a kind of network service method, which is characterized in that the described method includes:
The key feature in acquired each user request is extracted, according to the key feature in each user's request and in advance The corresponding relationship of the Internet Service Provider's ISP network group and key feature first established determines that each user's request is corresponding ISP network group includes at least one ISP cache node in each ISP network group;
For each ISP network group, each user in the ISP network group is requested, according to deeply learning algorithm, is led to It crosses to circuit sequentially and executes the step of choosing ISP cache node, calculating the income of the ISP network group and update policing parameter, really Fixed whether there is makes the maximum revenue of the ISP network group and meets the target ISP cache node of user's request condition;
If it does, being responded by the target ISP cache node to user request and transmitting request data to the use Family;
If it does not, being responded by content distributing network CDN to user request and transmitting request data to the use Family.
2. network service method according to claim 1, which is characterized in that saved in described cached by the target ISP After point responds to user request and transmits request data to the user, the method also includes:
Residual available bandwidth after the bandwidth resources of the target ISP cache node to be updated to transmit the request data, and Available bandwidth in the network state of the target ISP cache node is updated to the residual available bandwidth, network state packet It includes: available bandwidth and whether comprising the request data.
3. network service method according to claim 1 or 2, which is characterized in that it is described according to deeply learning algorithm, The step of choosing ISP cache node, calculating the income of the ISP network group and update policing parameter is executed by circuiting sequentially, It determines whether there is the maximum revenue for making the ISP network group and meets the target ISP cache node of user's request condition, wrap It includes:
The corresponding learning model of ISP network group is arranged in the state parameter for obtaining each ISP cache node in the ISP network group Parameter, the parameter of the learning model includes: value threshold value and cycle-index;
Step A, according to initial policy parameter π (a | s0), choose ISP cache node, and the initial policy parameter π (a | s0) first Beginning state s0Probability when each ISP cache node of lower selection is impartial;
Wherein, original state s0It indicates not yet to start the state to receive an assignment when requesting, a expression pair after the deployment of ISP cache node The user requests selection ISP cache node to carry out matching and data transmission;
Step B obtains the r value of the ISP network group according to the state parameter of selected ISP cache node, and the r value includes Pre-set big value and small value, the big value indicate that selected ISP cache node has the condition of response user's request, institute Stating small value indicates that selected ISP cache node does not have the condition of response user's request;
According to cost function: V (s, a)=Eπ(a|s)[r+ γ W (s ')] determines the value V at state s after selection operation a (s,a);
Wherein, (s a) indicates that the profit expectation of the selection operation a at state s, s indicate all ISP cachings in current environment group to V The network state after operation a is taken in the state of node, s ' expression at current network state s, and policing parameter π (a | s) is indicated The probability of selection operation a at state s, r indicate the reward value of the ISP network group at policing parameter π (a | s), and γ indicates folding Button rate, the value of γ are 0-1, and W (s ') indicates the maximum value reached under next state s ', and r+ γ W (s ') is operating value letter Number;
Step C, according to advantage function: A (s, a)=Q (s, a)-V (s, a), it is determined whether selection operation a;
((s a) indicates to perform operation a at state s Q by s, a)=r+ γ W (s '), Q;
If (s a) is greater than 0, selection operation a to A, executes step D;
If (s a) is not more than 0, not selection operation a, return step A to A;
Step D, according to formula:It calculates from original state s0To state s it is stateful under ISP Whether the average value of the value of group of networks, the average value judged are less than the value threshold value;
If obtained average value is less than the value threshold value, and the number for choosing ISP cache node is less than cycle-index, executes Step E;If obtained average value is less than the value threshold value, and the number for choosing ISP cache node is equal to cycle-index, really There is no the target ISP of the maximum revenue for making the ISP network group and satisfaction response user's request condition in the fixed ISP network group Cache node;
If obtained average value is not less than the value threshold value, the ISP cache node of final choice is determined as target ISP and is delayed Deposit node;
Step E, according to gradient function:
Determination makesChange most fast direction, updates plan Slightly parameter π (a | s), return step B.
4. network service method according to claim 3, which is characterized in that requested by CDN the user described After being responded and transmitting request data to the user, the method also includes:
The request data is cached into the ISP cache node of the final choice, and the ISP of the final choice is cached Whether it is updated to by not including the request data comprising the request comprising the request data in the network state of node Data.
5. network service method according to claim 3, which is characterized in that after stepb, the method also includes:
Policing parameter π (a | s) is sent to parameter server by the ISP network group, the parameter server according to receive from The policing parameter that all ISP network groups of body covering are sent is updated the policing parameter π of parameter server (a | s), and will Updated policing parameter π (a | s) it is sent to all ISP network groups of itself covering.
6. a kind of Web services apparatus, which is characterized in that described device includes:
Grouping module, for extracting the key feature in acquired each user request, according to the pass in each user's request The corresponding relationship of key feature and the Internet Service Provider's ISP network group and key feature that pre-establish determines each use Corresponding ISP network group is requested at family, includes at least one ISP cache node in each ISP network group;
Node determining module requests each user in the ISP network group, for being directed to each ISP network group according to depth Nitrification enhancement is spent, by circuiting sequentially the income and update that execute and choose ISP cache node, calculate the ISP network group The step of policing parameter, it is determined whether there is the maximum revenue for making the ISP network group and meet the target of user's request condition ISP cache node;
Node data transmission module, for determining in the node determining module there are when target ISP cache node, by described Target ISP cache node responds user request and transmits request data to the user;
Network data transmission module is used for when the node determining module determines and target ISP cache node is not present, by interior Content distributing network CDN responds user request and transmits request data to the user.
7. Web services apparatus according to claim 6, which is characterized in that described device further include:
Update module, it is surplus after the request data for being updated to transmit by the bandwidth resources of the target ISP cache node Remaining available bandwidth, and the available bandwidth in the network state of the target ISP cache node is updated to the residue can use band Whether width, network state include: available bandwidth and include the request data.
8. Web services apparatus according to claim 6 or 7, which is characterized in that the node determining module is for obtaining The state parameter of each ISP cache node in the ISP network group, is arranged the parameter of the corresponding learning model of ISP network group, institute The parameter for stating learning model includes: value threshold value and cycle-index;
Step A, according to initial policy parameter π (a | s0), choose ISP cache node, and the initial policy parameter π (a | s0) first Beginning state s0Probability when each ISP cache node of lower selection is impartial;
Wherein, original state s0It indicates not yet to start the state to receive an assignment when requesting, a expression pair after the deployment of ISP cache node The user requests selection ISP cache node to carry out matching and data transmission;
Step B obtains the r value of the ISP network group according to the state parameter of selected ISP cache node, and the r value includes Pre-set big value and small value, the big value indicate that selected ISP cache node has the condition of response user's request, institute Stating small value indicates that selected ISP cache node does not have the condition of response user's request;
According to cost function: V (s, a)=Eπ(a|s)[r+ γ W (s ')] determines the value V at state s after selection operation a (s,a);
Wherein, (s a) indicates that the profit expectation of the selection operation a at state s, s indicate all ISP cachings in current environment group to V The network state after operation a is taken in the state of node, s ' expression at current network state s, and policing parameter π (a | s) is indicated The probability of selection operation a at state s, r indicate the reward value of the ISP network group at policing parameter π (a | s), and γ indicates folding Button rate, the value of γ are 0-1, and W (s ') indicates the maximum value reached under next state s ', and r+ γ W (s ') is operating value letter Number;
Step C, according to advantage function: A (s, a)=Q (s, a)-V (s, a), it is determined whether selection operation a;
((s a) indicates to perform operation a at state s Q by s, a)=r+ γ W (s '), Q;
If (s a) is greater than 0, selection operation a to A, executes step D;
If (s a) is not more than 0, not selection operation a, return step A to A;
Step D, according to formula:It calculates from original state s0To state s it is stateful under ISP Whether the average value of the value of group of networks, the average value judged are less than the value threshold value;
If obtained average value is less than the value threshold value, and the number for choosing ISP cache node is less than cycle-index, executes Step E;If obtained average value is less than the value threshold value, and the number for choosing ISP cache node is equal to cycle-index, really There is no the target ISP of the maximum revenue for making the ISP network group and satisfaction response user's request condition in the fixed ISP network group Cache node;
If obtained average value is not less than the value threshold value, the ISP cache node of final choice is determined as target ISP and is delayed Deposit node;
Step E, according to gradient function:
Determination makesChange most fast direction, updates Policing parameter π (a | s), return step B.
9. a kind of electronic equipment characterized by comprising processor, communication interface, memory and communication bus, wherein described Processor, the communication interface, the memory complete mutual communication by the communication bus;
The memory, for storing computer program;
The processor when for executing the program stored on the memory, realizes any net of claim 1-5 The step of network method of servicing.
10. a kind of computer readable storage medium, which is characterized in that be stored with computer in the computer readable storage medium Program when the computer program is executed by processor, realizes the step of any network service method of claim 1-5 Suddenly.
CN201811504958.9A 2018-12-10 2018-12-10 Network service method, device, electronic equipment and readable storage medium Active CN109474696B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811504958.9A CN109474696B (en) 2018-12-10 2018-12-10 Network service method, device, electronic equipment and readable storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811504958.9A CN109474696B (en) 2018-12-10 2018-12-10 Network service method, device, electronic equipment and readable storage medium

Publications (2)

Publication Number Publication Date
CN109474696A true CN109474696A (en) 2019-03-15
CN109474696B CN109474696B (en) 2020-05-08

Family

ID=65676106

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811504958.9A Active CN109474696B (en) 2018-12-10 2018-12-10 Network service method, device, electronic equipment and readable storage medium

Country Status (1)

Country Link
CN (1) CN109474696B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112118278A (en) * 2019-06-04 2020-12-22 杭州海康威视系统技术有限公司 Computing node access method and device, electronic equipment and machine-readable storage medium
CN112446501A (en) * 2020-10-30 2021-03-05 北京邮电大学 Method, device and system for obtaining cache allocation model in real network environment

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101764747A (en) * 2009-12-24 2010-06-30 北京世纪互联宽带数据中心有限公司 Intelligent dispatching method, device and system of content delivery network flow
CN102891804A (en) * 2012-10-16 2013-01-23 中兴通讯股份有限公司南京分公司 Control strategy adjustment method and system
CN103475719A (en) * 2013-09-12 2013-12-25 北京科技大学 Content distribution method for minimizing cross-domain flows in CDN-P2P fusion network
CN103825916A (en) * 2012-11-16 2014-05-28 腾讯科技(深圳)有限公司 Resource downloading method and resource downloading system
CN103825975A (en) * 2014-02-26 2014-05-28 北京太阳庄科技有限公司 Cdn node distribution server and system
CN104618506A (en) * 2015-02-24 2015-05-13 庄奇东 Crowd-sourced content delivery network system, method and device
US9172713B2 (en) * 2008-09-24 2015-10-27 Neustar, Inc. Secure domain name system
CN107483630A (en) * 2017-09-19 2017-12-15 北京工业大学 A kind of construction method for combining content distribution mechanism with CP based on the ISP of edge cache

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9172713B2 (en) * 2008-09-24 2015-10-27 Neustar, Inc. Secure domain name system
CN101764747A (en) * 2009-12-24 2010-06-30 北京世纪互联宽带数据中心有限公司 Intelligent dispatching method, device and system of content delivery network flow
CN102891804A (en) * 2012-10-16 2013-01-23 中兴通讯股份有限公司南京分公司 Control strategy adjustment method and system
CN103825916A (en) * 2012-11-16 2014-05-28 腾讯科技(深圳)有限公司 Resource downloading method and resource downloading system
CN103475719A (en) * 2013-09-12 2013-12-25 北京科技大学 Content distribution method for minimizing cross-domain flows in CDN-P2P fusion network
CN103825975A (en) * 2014-02-26 2014-05-28 北京太阳庄科技有限公司 Cdn node distribution server and system
CN104618506A (en) * 2015-02-24 2015-05-13 庄奇东 Crowd-sourced content delivery network system, method and device
CN107483630A (en) * 2017-09-19 2017-12-15 北京工业大学 A kind of construction method for combining content distribution mechanism with CP based on the ISP of edge cache

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
HUA FAN,XIANGYANG GONG,WENDONG WANG等: "A NOVEL P2P TRAFFIC OPTIMIZATION METHOD", 《2010 3RD IEEE INTERNATIONAL CONFERENCE ON BROADBAND NETWORK AND MULTIMEDIA TECHNOLOGY(IC-BNMT)》 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112118278A (en) * 2019-06-04 2020-12-22 杭州海康威视系统技术有限公司 Computing node access method and device, electronic equipment and machine-readable storage medium
CN112118278B (en) * 2019-06-04 2023-07-04 杭州海康威视系统技术有限公司 Computing node access method, device, electronic equipment and machine-readable storage medium
CN112446501A (en) * 2020-10-30 2021-03-05 北京邮电大学 Method, device and system for obtaining cache allocation model in real network environment
CN112446501B (en) * 2020-10-30 2023-04-21 北京邮电大学 Method, device and system for acquiring cache allocation model in real network environment

Also Published As

Publication number Publication date
CN109474696B (en) 2020-05-08

Similar Documents

Publication Publication Date Title
US20170142177A1 (en) Method and system for network dispatching
CN101296152B (en) Data scheduling method and system of equity linked network
CN110830564B (en) CDN scheduling method, device, system and computer readable storage medium
US8938495B2 (en) Remote management system with adaptive session management mechanism
CN104834722A (en) CDN (Content Delivery Network)-based content management system
CN107404541B (en) Method and system for selecting neighbor node in peer-to-peer network transmission
CN104769550B (en) The method and system of bandwidth allocation in system for cloud computing
CN114760482B (en) Live broadcast source returning method and device
CN110336885A (en) Fringe node distribution method, device, dispatch server and storage medium
CN104901989A (en) Field service providing system and method
CN111083050B (en) Data stream transmission method and device based on software defined network
WO2024066626A1 (en) Route planning method and apparatus for real time audio/video network
CN110519090A (en) A kind of accelerator card distribution method, system and the associated component of FPGA cloud platform
CN101710904A (en) P2p flow optimization method and system thereof
CN105407162A (en) Cloud computing Web application resource load balancing algorithm based on SLA service grade
CN109474696A (en) A kind of network service method, device, electronic equipment and readable storage medium storing program for executing
CN108156257A (en) A kind of information-pushing method and device
CN115037696B (en) Data transmission method, device, electronic equipment and storage medium
EP3902279A1 (en) Audio and video stream distribution method in audio and video stream distribution system, and dynamic parent node
WO2016180284A1 (en) Service node allocation method, device, cdn management server and system
Shi et al. CoLEAP: Cooperative learning-based edge scheme with caching and prefetching for DASH video delivery
CN108512765B (en) Network content diffusion method based on network node distributed Pagerank
CN115412737A (en) Method and device for determining live broadcast source return relay node
CN115665262A (en) Request processing method and device, electronic equipment and storage medium
CN104092735A (en) Cloud computing data access method and system based on binary tree

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant