CN109951317B - User-driven popularity perception model-based cache replacement method - Google Patents
User-driven popularity perception model-based cache replacement method Download PDFInfo
- Publication number
- CN109951317B CN109951317B CN201910127784.7A CN201910127784A CN109951317B CN 109951317 B CN109951317 B CN 109951317B CN 201910127784 A CN201910127784 A CN 201910127784A CN 109951317 B CN109951317 B CN 109951317B
- Authority
- CN
- China
- Prior art keywords
- content
- popularity
- cache
- node
- time
- 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.)
- Active
Links
Images
Abstract
The invention discloses a cache replacement method based on a user-driven popularity perception model, which comprises the following steps: s1: establishing a popularity perception model; s2: the controller calculates the popularity of the content according to the life cycle of the request content; s3: the controller judges the propagation stage according to the popularity of the content and calculates the probability of caching the content; s4: and the cache replacement is completed under the control of the controller. A popularity perception model of early propagation of content is established by introducing user behaviors, user social networks and content levels, and the popularity of the content is predicted according to the user behaviors, so that a cache replacement strategy is executed. Compared with the cache replacement strategies such as LCE and LRU, the UDPAM cache replacement strategy has higher cache direct hit rate and lower cache replacement rate in the satellite network.
Description
Technical Field
The invention belongs to the field of communication, and particularly relates to a cache replacement method based on a user-driven popularity perception model.
Background
With the change and development of network environment, the network transmission object is converted from simple text data into streaming media data. In a traditional host-centric network model, the waist structure of an IP protocol stack becomes a bottleneck of data response speed. The continuous development of satellite network technology, and the characteristics of globality and real-time property thereof enable the satellite network to have distinct advantages in data transmission in a cross-space environment. However, the existing satellite network has great limitations in network management, service deployment and the like. Researchers have therefore addressed the above issues by virtue of Software Defined Networking (SDN) and Information Centric Networking (ICN) advantages.
Different from the characteristics of fixed network topology, sufficient cache resources and the like of a ground network, the network topology of the satellite network has high dynamics, and the computing resources and cache resources of the nodes are limited. Due to the mobility of satellite nodes, the direct cache hit rate of a satellite network is reduced by traditional cache replacement strategies such as LCE (lower control element) and LRU (least recently used) and the like; meanwhile, the limitation of cache resources enables the satellite node to have a high cache replacement rate when cache replacement strategies such as LCE and the like are used. And under the condition of limited computing resources, the satellite network has higher queuing delay when processing services, and the performance of the satellite network is reduced. It is therefore important to design a cache replacement strategy that is suitable for satellite networks.
Disclosure of Invention
Aiming at the defects of the prior art, the technical problem to be solved by the invention is to provide a User-Driven Popularity Awareness Model (UDPAM) based cache replacement method, establish a Popularity Awareness Model for content early propagation by introducing User behaviors, User social networks and content levels, and predict the Popularity of the content according to the User behaviors so as to execute a cache replacement strategy. Compared with the cache replacement strategies such as LCE and LRU, the UDPAM cache replacement strategy has higher cache direct hit rate and lower cache replacement rate in the satellite network.
In order to achieve the purpose, the technical scheme of the application is as follows: a cache replacement method based on a user-driven popularity perception model comprises the following steps:
s1: establishing a popularity perception model;
s2: the controller calculates the popularity of the content according to the life cycle of the request content;
s3: the controller judges the propagation stage according to the popularity of the content and calculates the probability of caching the content;
s4: and the cache replacement is completed under the control of the controller.
Further, the existing time of the content, the requesting user, the requesting frequency and the content transmission characteristics are the first characteristics for establishing the popularity perception model.
Further, the presence of contentThe perceptual contribution of time to popularity is formulated as: η ═ λ (t) (. alpha. omega. (t) + (1-a). gamma. (t)s,t))，
Where λ (t) represents time [ t ]s,t]The propagation rate of content through a social network of a certain user, and omega (t) is a random variable of a user request behavior and describes the randomness of content requests in the network; gamma (t)sT) is an increasing part of the popularity of the requested content, the content which becomes popular in the future is sensed, theta reflects the increasing speed of the popularity of the content and is characterized by the request frequency; alpha is expressed by the time of arrival of the request according to the content propagation time and the weight of the random contribution and the growing contribution of the adaptive adjustment and the running degree perception.
Further, the controller calculates the popularity of the content according to the life cycle of the content, and for the early propagation of the content, on one hand, the survival of the newly requested content is ensured through ω (t) and on the other hand, γ (t) is ensuredsT) ensuring the increase or decrease of the popularity of the content by adjusting theta; for medium and late dissemination of content, the popularity of the content is calculated according to its request frequency.
Further, the request frequency formula of the content is:
wherein p is [ t ] from the last time the content popularity is calculated to the current times,t]The ratio of the request times N of the content to the time interval in time represents the request frequency of the content in unit time; n is a radical ofsFor the number of other content requests, β represents [ t ]s,t]A rate of change of the content request over time; eta is expressed in [ t ]s,t]And in time, the normalized request frequency of the content is the popularity of the content.
Further, the propagation stage of the content is judged according to the popularity of the content, and the cache probability of the propagation stage is calculated: for in early or middle stagePropagation, the probability of which is calculated as:
ρ＝rand()，
wherein, κiDenotes the probability of a node caching the content under the combined effect of content popularity and node caching space, κiThe larger the content is, the larger the probability that the content is cached by the node is, and the smaller the probability is otherwise; ρ is a random number obeying uniform distribution; if tau is 0, the node does not cache the content, and if tau is 1, the node caches the content; deltaiCaching probabilities are allowed for the nodes.
Further, in the present invention,
wherein C isiRepresenting the total cache space of node i, CimRepresenting the remaining cache space of node i.
Further, a determination is made whether the content is cached: when the number of the cache content nodes is more than half of the total number of the forwarding nodes, the controller finishes decision calculation of the cache nodes, and the controller issues the cache content nodes to the corresponding forwarding nodes through the flow table; if the node still has a cache space, directly caching the content; and if the node has no residual cache space, replacing the content with the lowest popularity according to the popularity information of the content in the cache information table.
Due to the adoption of the technical scheme, the invention can obtain the following technical effects: the method and the device can achieve high cache direct hit rate and low cache replacement rate. The utilization rate of satellite node resources is effectively improved, and a more appropriate cache replacement strategy is provided for the research of a satellite network.
Drawings
FIG. 1 is a flow diagram of a cache replacement policy based on a user-driven popularity aware model in accordance with the present invention;
FIG. 2 is a flow chart of a cache replacement policy;
FIG. 3 is a graph comparing direct hit rates of satellite networks;
FIG. 4 is a diagram of a direct hit rate versus cache for a satellite network;
FIG. 5 is a graph comparing cache replacement rates in a satellite network;
FIG. 6 is a diagram of a satellite network cache replacement rate versus cache space.
Detailed Description
The invention is described in further detail below with reference to the following figures and specific examples: the present application is further described by taking this as an example.
The architecture employs three high-orbit geostationary satellites as SDN controllers.
According to the cache replacement method based on the user-driven popularity perception model, as shown in fig. 1, formulas related to the method are elaborated in detail in combination with a flow chart. The cache replacement method comprises the following steps:
s1: establishing a popularity prediction model according to the existing time of the content, the requesting user, the requesting frequency, the content and other factors;
s2: the SDN controller calculates the popularity of the content according to the life cycle of the request content;
s3: the controller judges the propagation stage according to the popularity of the content and calculates the probability of caching the content;
s4: finishing a cache replacement strategy under the control of the controller;
the existence time of the content, the requesting user of the content, the request frequency and the content propagation characteristics in the S1 are main factors for establishing the popularity perception model. On one hand, the search behavior of the user on the content can be represented by dynamic modeling, namely a power law response function describing the user behavior:
it characterizes the potential impact of the cause of the user's behavior. By definition, the kernel function is memorized
Representing the storage of "reason" to "action" (watching video) for a single individualAre distributed in time. On the other hand, under the influence of the user social network, the content has the characteristic of cascade propagation in the network. The cascade propagation process can be described by the Poisson model of the free-excited Hawkes condition:
λ (t) represents time [ t ]s,t]The propagation rate of content through a user's social network, where μiAnd the normalized influence factor representing the contribution of the user i to the content propagation is described by the connectivity of the user network node. Based on the dynamic analysis, the content dissemination in the network is random in the early stage, and the distribution of the content popularity is also random. Through early distribution, the content may stop spreading after a certain time, or may reach a certain popularity threshold into the middle distribution stage. At this stage, the content popularity is on an exponential growth trend. After the content is propagated through the medium growth, the heat of the content is rapidly reduced. And the content enters a later propagation stage, and the popularity of the content is in a negative index and rapidly declines.
The perceptual contribution formula of the content existence time to the popularity in the S2 is as follows:
η＝λ(t)*(α·ω(t)+(1-α)·γ(ts,t))
where λ (t) represents time [ t ]s,t]The propagation rate of content through a social network of a certain user, and omega (t) is a random variable of a user request behavior and describes the randomness of content requests in the network; gamma (t)sT) is an increasing part of the popularity of the requested content, the content which becomes popular in the future is sensed, theta reflects the increasing speed of the popularity of the content and is characterized by the request frequency; alpha is expressed by the time of arrival of the request according to the content propagation time and the weight of the random contribution and the growing contribution of the adaptive adjustment and the running degree perception.
For the early propagation of the content in S3, on one hand, the survival of the newly requested content is guaranteed by ω (t), and on the other handγ(tsT) to guarantee the increase or decrease of the popularity of the content by the adjustment of theta. For medium and late dissemination of content, the popularity of content is mainly calculated according to its request frequency.
The request frequency formula for content is:
wherein p is [ t ] from the last time the content popularity is calculated to the current times,t]The ratio of the request times N of the content to the time interval in time represents the request frequency of the content in unit time; beta represents [ t ]s,t]A rate of change of the content request over time; eta is expressed in [ t ]s,t]And in time, the normalized request frequency of the content is the popularity of the content. In S3, the propagation stage of the content is determined according to the popularity of the content, and the caching probability of the stage is calculated. For propagation in early or medium stages, the probability is calculated as follows:
ρ＝rand()，
wherein, κiDenotes the probability of a node caching the content under the combined effect of content popularity and node caching space, κiThe larger the content is, the larger the probability that the content is cached by the node is, and the smaller the probability is otherwise; ρ is a random number obeying uniform distribution; if τ is 0, the node does not cache the content, and if τ is 1, the node caches the content.
Determination as to whether or not the content is cached in S4: when the number of the cache nodes with a certain content is more than half of the number of the forwarding nodes, the controller ends the decision calculation of the cache nodes. The controller issues the data to the corresponding forwarding node through the flow table. If the node still has a cache space, directly caching the content; and if the node has no residual cache space, replacing the content with the lowest popularity according to the popularity information of the content in the cache information table. The flow of the cache replacement policy is shown in fig. 2.
The effect of the present invention is demonstrated by way of another example.
According to the SDICSN architecture, a ground network is divided into a plurality of autonomous Areas (AS) according to the characteristics of areas and the like, and a control layer under an SDN framework adopts layered distributed control; the whole satellite network is controlled in real time by taking 3 high-orbit satellites as controllers in the satellite network, an iridium constellation simulation forwarding layer is adopted in the low-orbit satellite network in the simulation process, and satellite nodes of the forwarding layer are in four-neighbor domain communication (2 ISLs and 2 ISLs).
The method comprises the steps that 1000 contents numbered according to popularity are cached in a network, the size range of a content block is 10-100 MB, the initial caching capacity of a forwarding node in the satellite network is 1G, and the number of the initially cached contents of each node is 10-100. The number of ICN clients in the network is N M, each ICN client initiates a request for the content in parallel, and the probability of each ICN client initiating the request is subject to Poisson distribution with lambda being 100.
As shown in fig. 3, the comparison of LCEs and LRUs with the UDPAM cache replacement strategy designed herein in terms of cache hit rate.
As shown in fig. 4, by changing the size of the cache capacity of the satellite node, the influence of the cache capacity of the node on the cache hit rate is verified, that is, the larger the cache capacity of the node is, the more popular contents can be cached by the node under the UDPAM cache replacement policy, and the higher the probability of direct cache hit of the network is.
As shown in fig. 5, the comparison of the cache replacement rate for LCEs and LRUs versus the UDPAM cache replacement strategy designed herein.
As shown in fig. 6, comparing the effect of the node cache capacity on the cache replacement rate under the UDPAM cache replacement policy, it can be seen that the cache replacement rate of the node decreases as the cache capacity increases.
In summary, the following results can be obtained: the method and the device can achieve high cache direct hit rate and low cache replacement rate, effectively improve the utilization rate of satellite node resources, and provide a more appropriate cache replacement strategy for the research of a satellite network.
The above description is only for the purpose of creating a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can substitute or change the technical solution and the inventive concept of the present invention within the technical scope of the present invention.
Claims (2)
1. A cache replacement method based on a user-driven popularity perception model is characterized by comprising the following steps:
s1: establishing a popularity perception model;
s2: the controller calculates the popularity of the content according to the life cycle of the request content;
s3: the controller judges the propagation stage according to the popularity of the content and calculates the probability of caching the content;
s4: finishing cache replacement under the control of the controller;
the existing time, the requesting user, the requesting frequency and the content transmission characteristic of the content are first characteristics for establishing a popularity perception model;
the perceptual contribution of the time of existence of the content to the popularity is formulated as:
wherein
Represents time [ ts,t]The propagation rate of content through a user's social network, where μiThe normalized influence factor representing the contribution of the user i to the content propagation, and omega (t) is a random variable of the user request behavior and describes the randomness of the content request in the network; gamma (t)sT) is an increasing part of the popularity of the requested content, the content which becomes popular in the future is sensed, theta reflects the increasing speed of the popularity of the content and is characterized by the request frequency; alpha adaptively adjusts the random contribution and the growth contribution weight of the row-degree perception according to the content propagation time and is expressed by the arrival time of the request;
the controller calculates the popularity of the content according to the life cycle of the content, and for the early propagation of the content, on one hand, the survival of the newly requested content is ensured through omega (t) and on the other hand, gamma (t) is usedsT) ensuring the increase or decrease of the popularity of the content by adjusting theta; for medium and late dissemination of content, the popularity of the content is calculated according to its request frequency;
the request frequency formula for content is:
wherein p is [ t ] from the last time the content popularity is calculated to the current times,t]The ratio of the request times N of the content to the time interval in time represents the request frequency of the content in unit time; n is a radical ofsFor the number of other content requests, θ represents [ t ]s,t]A rate of change of the content request over time; eta is expressed in [ t ]s,t]Within time, the request frequency of the content normalization is the popularity of the content;
judging the propagation stage of the content according to the popularity of the content, and calculating the cache probability of the stage: for propagation in early or medium stages, the probability is calculated as follows:
wherein, κiDenotes the probability of a node caching the content under the combined effect of content popularity and node caching space, κiThe greater the probability that a node will cache the content, kiThe smaller the probability that the node caches the content is; ρ is a random number obeying uniform distribution; if tau is 0, the node does not cache the content, and if tau is 1, the node caches the content; deltaiAllowing a caching probability for the node;
2. The cache replacement method based on the user-driven popularity perception model according to claim 1, wherein whether the content is cached is determined by: when the number of the cache content nodes is more than half of the total number of the forwarding nodes, the controller finishes decision calculation of the cache nodes, and the controller issues the cache content nodes to the corresponding forwarding nodes through the flow table; if the node still has a cache space, directly caching the content; and if the node has no residual cache space, replacing the content with the lowest popularity according to the popularity information of the content in the cache information table.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910127784.7A CN109951317B (en) | 2019-02-18 | 2019-02-18 | User-driven popularity perception model-based cache replacement method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910127784.7A CN109951317B (en) | 2019-02-18 | 2019-02-18 | User-driven popularity perception model-based cache replacement method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109951317A CN109951317A (en) | 2019-06-28 |
| CN109951317B true CN109951317B (en) | 2022-04-05 |
Family
ID=67006915
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910127784.7A Active CN109951317B (en) | 2019-02-18 | 2019-02-18 | User-driven popularity perception model-based cache replacement method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109951317B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112286488B (en) * | 2020-09-17 | 2023-06-16 | 昆明理工大学 | Buffer memory replacement method based on utility value of data packet |
| CN113472420B (en) * | 2021-06-29 | 2022-05-24 | 大连大学 | Satellite network cache placement method based on regional user interest perception |
| CN113992753B (en) * | 2021-10-17 | 2022-06-10 | 南京理工大学 | Intelligent caching strategy for heaven-earth integrated satellite network node |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101989289A (en) * | 2009-08-06 | 2011-03-23 | 富士通株式会社 | Data clustering method and device |
| CN104281882A (en) * | 2014-09-16 | 2015-01-14 | 中国科学院信息工程研究所 | Method and system for predicting social network information popularity on basis of user characteristics |
| CN105553998A (en) * | 2015-12-23 | 2016-05-04 | 中国电子科技集团公司第三十研究所 | Network attack abnormality detection method |
| CN105897517A (en) * | 2016-06-20 | 2016-08-24 | 广东电网有限责任公司信息中心 | Network traffic abnormality detection method based on SVM (Support Vector Machine) |
| CN108173903A (en) * | 2017-11-30 | 2018-06-15 | 南京邮电大学 | Application of the autonomous system cooperation caching strategy in CCN |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8577993B2 (en) * | 2011-05-20 | 2013-11-05 | International Business Machines Corporation | Caching provenance information |
-
2019
- 2019-02-18 CN CN201910127784.7A patent/CN109951317B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101989289A (en) * | 2009-08-06 | 2011-03-23 | 富士通株式会社 | Data clustering method and device |
| CN104281882A (en) * | 2014-09-16 | 2015-01-14 | 中国科学院信息工程研究所 | Method and system for predicting social network information popularity on basis of user characteristics |
| CN105553998A (en) * | 2015-12-23 | 2016-05-04 | 中国电子科技集团公司第三十研究所 | Network attack abnormality detection method |
| CN105897517A (en) * | 2016-06-20 | 2016-08-24 | 广东电网有限责任公司信息中心 | Network traffic abnormality detection method based on SVM (Support Vector Machine) |
| CN108173903A (en) * | 2017-11-30 | 2018-06-15 | 南京邮电大学 | Application of the autonomous system cooperation caching strategy in CCN |
Non-Patent Citations (1)
| Title |
|---|
| 基于收视行为的互联网电视节目流行度预测模型;朱琛刚;《电子与信息学报》;20171015;第39卷(第10期);第2.2节，图1-2 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109951317A (en) | 2019-06-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhong et al. | A deep reinforcement learning-based framework for content caching | |
| He et al. | QoE-driven content-centric caching with deep reinforcement learning in edge-enabled IoT | |
| CN102624922B (en) | Method for balancing load of network GIS heterogeneous cluster server | |
| CN109951317B (en) | User-driven popularity perception model-based cache replacement method | |
| Fang et al. | Deep-reinforcement-learning-based resource allocation for content distribution in fog radio access networks | |
| WO2023168824A1 (en) | Mobile edge cache optimization method based on federated learning | |
| Björkqvist et al. | Minimizing retrieval latency for content cloud | |
| CN110430440A (en) | Video transmission method, system, computer equipment and storage medium | |
| CN107276788B (en) | Communication model construction method with cache base station based on sleep control | |
| CN104426979A (en) | Distributed buffer scheduling system and method based on social relations | |
| CN108541025B (en) | Wireless heterogeneous network-oriented base station and D2D common caching method | |
| CN113255004A (en) | Safe and efficient federal learning content caching method | |
| Aloqaily et al. | Blockchain and fl-based network resource management for interactive immersive services | |
| CN113411826B (en) | Edge network equipment caching method based on attention mechanism reinforcement learning | |
| CN105227665B (en) | A kind of caching replacement method for cache node | |
| Li et al. | Efficient cooperative cache management for latency-aware data intelligent processing in edge environment | |
| CN117202265A (en) | DQN-based service migration method in edge environment | |
| CN113766540B (en) | Low-delay network content transmission method, device, electronic equipment and medium | |
| CN114095573B (en) | Content copy placement method of CDN-P2P network based on edge cache | |
| CN115361710A (en) | Content placement method in edge cache | |
| Hossen et al. | Mobile task offloading under unreliable edge performance | |
| CN110138863B (en) | Self-adaptive consistency protocol optimization method based on Multi-Paxos grouping | |
| Jia et al. | Cluster-based content caching driven by popularity prediction | |
| CN113672372A (en) | Multi-edge cooperative load balancing task scheduling method based on reinforcement learning | |
| Liu et al. | User-Driven Cache Replacement Strategy for Satellite-Terrestrial Networks Based on SDN |
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 |