CN105791397A - Caching method of ICN (Information-Centric Networking) based on SDN - Google Patents

Abstract

The invention discloses a caching method of an ICN (Information-Centric Networking) based on an SDN, and belongs to the technical field of the Internet. The method is based on a caching mechanism of the ICN (Information-Centric Networking) of the SDN (software defined network). In the ICN, each network node has a storage function; network resources and network states are sensed through an SDN control plane; and cache placement and cache replacement are differentially realized in a whole network range through application of complex network community thoughts. The caching mechanism provided by the invention comprises two parts: a content cache placement mechanism, wherein in the mechanism, ICN data packets determine cache communities of nodes according to the community importance degree of communities to which the ICN data packets belong, and the popularity of to-be-cached contents, and in the same community, cache nodes are determined according to the node community importance degree and the popularity degree of the contents; and a cache content replacement mechanism, wherein in the mechanism, replacement contents are determined according to the node community importance degree of the nodes and the popularity degree of content objects. According to the method, the problems of low caching efficiency and inflexible control of the ICN can be solved well.

Description

A kind of ICN caching method based on SDN

Technical field

The invention belongs to Internet technical field.

Technical background

According to Cisco (Cisco) corporate statistics, global network flow increased 4 times in past 5 years.During 2012～2017, global network flow will grow at top speed with the annual average compound growth rate of 23%.And along with continuing to bring out of new application (such as Internet of Things pattern, cloud computing mode), flow produces and fundamental change also will occur transmission means.Wherein, major part flow all will be derived from the content obtaining class application that user drives.

The transformation of the huge volumes of content obtaining mode that the end to end communication pattern driven by sender to adapt to internet, applications drives to recipient, strengthen the network support to aspects such as safety, quality of service, mobility, extensibilities simultaneously, researcheres propose class new network architectural framework centered by information in recent years, it is referred to as information centre's network (Information-CentricNetworking is called for short ICN).For alleviate current network flow quickly increase the severe pressure that the network bandwidth is caused, ICN adds the built-in buffer memory of the whole network.At present, ICN caching mechanism can be divided into: along road (On-path) caching mechanism and bypass (Off-path) caching mechanism.

According to certain policy selection part of nodes, transferring content being carried out buffer memory on the transmit path along road (On-path) caching mechanism and content, the advantage of this kind of strategy is easy realization, and its shortcoming is only to consider buffer efficiency from individual node；The node that bypass (Off-path) caching mechanism and content are not necessarily along road according to certain policy selection on the transmit path carries out content caching, the advantage of this kind of strategy is from localized network, further improve the buffer efficiency of ICN, its shortcoming for not only not easily to realize in current network framework, and does not account for Internet resources and network state.

For this, patent of the present invention merges under framework at SDN and ICN, sensing network resource and network state, application complex network community thought, in network-wide basis, the buffer memory that realizes of differentiation is placed and buffer memory replacement, preferably resolves ICN buffer efficiency not high, controls the problems such as dumb.

Summary of the invention

It is an object of the invention to merge under framework at SDN and ICN, it is proposed that a kind of brand-new ICN buffer memory thought, solve the problems such as the buffer efficiency of ICN is not high and control is dumb preferably.

From aspects such as corporations, node corporations importance degree, corporations' importance degree and content popularit, the caching method that the present invention proposes is described separately below.

One, corporations: the empirical research of Internet network structure is shown, Internet network structure has obvious corporations characteristic, namely inside same corporations, connects relative close, and the connection between the node between corporations is relatively sparse between node.And most of user content of interest (interest) in same corporations is also similar.The node of different corporations, under same SDN controller, is carried out cache management by the present invention.

Two, node corporations importance degree: based on node with not in a network its represent network adjacent matrix eigenvalue change represent the node importance degree to network community characteristic, as:Wherein, c is corporations' number, v in network_iRepresenting with the router in network for node, the physical link between router is the ith feature vector of the adjacency matrix that limit builds, v_ikRepresent characteristic vector v_iIn kth element, for normalization node corporations importance degree, define IN_k=P_k/ c, meetsThe present invention is under same SDN controller, in same corporations, it is achieved node corporations importance degree mates with content popularit, improves the utilization rate of spatial cache.

Three, corporations' importance degree: represent the importance degree of different corporations under same controller based on interior nodes corporations of corporations importance degree, as:WhereinRepresent corporations C_kInterior interstitial content.The present invention is under same SDN controller, between different corporations, it is achieved corporations' importance degree mates with content popularit, further improves the utilization rate of spatial cache.

Four, in units of content popularit: to be accessed for probability be content popularit to definition content object, node under a controller or corporations, if its properties collection is CP={cp_i| i=1,2 ..., N}, the content cp in CP_iLabel is carried out, as to any i < j, being then accessed for Probability p by the rank of its popularity height_i<p_j.The present invention under same SDN controller, the node place that pitch point importance that the content priority that popularity is high is stored in the corporations that corporations' importance degree is big is big, it is easy to by other node visit.

Accompanying drawing explanation

Fig. 1 is the implementation framework figure of the ICN buffer memory based on SDN；

Fig. 2 is caching method generting machanism.

Concrete implementation

The implementation framework schematic diagram of the present invention is as shown in Figure 1.

The Step1:SDN chain of command perception ICN packet of switch under controller first；

Step2: transfer to controller to complete the generation of caching method, in caching method generation process, as in figure 2 it is shown, mainly complete the work of two aspects, its concrete implementation is as follows:

(1) placement location of content is determined.In same controller, content placement location includes two strategies, a method determining that content caching corporations；Two is the method determining content caching node in corporations.

The method determining content caching corporations: in same controller, is first depending on corporations' importance degree, corporations is ranked up from big to small, if the corporations' set in a controller is IC={IC_i| i=1,2 ..., C}, C represents the number of corporations under same controller, as to any i < j, then IC_i<IC_j, then, according to the content popularit under controller by content be divided into C interval, compare content popularit under the popularity of the content that need to store and whole controller, it is determined that its affiliated interval, so that it is determined that the buffer memory corporations of content simultaneously；

Content caching nodal method is determined: after the corporations completing to need cache contents institute buffer memory, for same corporations, first arranged in descending order by node according to node corporations importance degree, such as IN={IN in corporations_i| i=1,2 ..., N_Ci, then, divide the content into N according to the popularity of the content in corporations_CiIndividual interval, compares popularity and the content popularit under whole corporations of the content that need to store, it is determined that its affiliated interval simultaneously, so that it is determined that the cache node of content.

(2) the replacement content in node is determined.Node corporations importance degree according to node and the popularity of content object, select different content objects to be replaced.Namely determining, by node corporations importance degree, the probability that the content object in nodal cache queue is replaced, node as big in node corporations importance degree, the probability that the content object that popularity is low in buffer queue is replaced is big；And the node that pitch point importance degree is little, the probability that the content object that popularity is high in buffer queue is replaced is big, further improves the storage efficiency of spatial cache.If the spatial cache of selected ICN node is sized to NC object.And the node corporations importance degree of this node is IN_i, and the importance degree of the corporations belonging to this node is IC_j, then this content object is when this node needs buffer memory, will according to Probability p_ijK () replaces the content (k ∈ [1, C]) of the kth position of this nodal cache queue, p_ijK () defines such as formula (1).

$p_{ij}\left(k\right)=\frac{1}{\mathrm{\&alpha;}}{\left(\frac{{\mathrm{IN}}_i}{{\mathrm{IC}}_j}\right)}^{\mathrm{\&beta;}k}---\left(1\right)$

Wherein, α is normalization factor, meets formula (2), and β is probability adjustment factor.

$\underset{k=1}{\overset{C}{\&Sigma;}}{p}_{ij}\left(k\right)=1\&DoubleRightArrow;\mathrm{\&alpha;}=\underset{k=1}{\overset{C}{\&Sigma;}}{\left(\frac{{\mathrm{IN}}_i}{{\mathrm{IC}}_j}\right)}^{\mathrm{\&beta;}k}---\left(2\right)$

Step3: caching method is issued to switch stream table and reaches data surface, complete the execution of caching method.

Claims (2)

1., based on an ICN caching method of SDN, from aspects such as corporations, node corporations importance degree, corporations' importance degree and content popularit, the caching method that the present invention proposes is described respectively, it is characterized in that:

One corporations: the empirical research of Internet network structure is shown, Internet network structure has obvious corporations characteristic, namely inside same corporations, connects relative close, and the connection between the node between corporations is relatively sparse between node；And the most of user content interest of interest in same corporations is also similar.The node of different corporations, under same SDN controller, is carried out cache management by the present invention；

Two node corporations importance degrees: based on node with not in a network its represent network adjacent matrix eigenvalue change represent the node importance degree to network community characteristic, as:Wherein, c is corporations' number, v in network_iRepresenting with the router in network for node, the physical link between router is the ith feature vector of the adjacency matrix that limit builds, v_ikRepresent characteristic vector v_iIn kth element, for normalization node corporations importance degree, define IN_k=P_k/ c, meetsThe present invention is under same SDN controller, in same corporations, it is achieved node corporations importance degree mates with content popularit, improves the utilization rate of spatial cache；

Three corporations' importance degrees: represent the importance degree of different corporations under same controller based on interior nodes corporations of corporations importance degree, as:WhereinRepresent corporations C_kInterior interstitial content；The present invention is under same SDN controller, between different corporations, it is achieved corporations' importance degree mates with content popularit, further improves the utilization rate of spatial cache；

In units of four content popularit: to be accessed for probability be content popularit to definition content object, node under a controller or corporations, if its properties collection is CP={cp_i| i=1,2 ..., N}, the content cp in CP_iLabel is carried out, as to any i < j, being then accessed for Probability p by the rank of its popularity height_i<p_j；The present invention under same SDN controller, the node place that pitch point importance that the content priority that popularity is high is stored in the corporations that corporations' importance degree is big is big, it is easy to by other node visit.

2. the ICN caching method based on SDN according to claim 1, is characterized in that:

The Step1:SDN chain of command perception ICN packet of switch under controller first；

Step2: transfer to controller to complete the generation of caching method, in caching method generation process, as in figure 2 it is shown, mainly complete the work of two aspects, its concrete implementation is as follows:

(1) placement location of content is determined.In same controller, content placement location includes two strategies, a method determining that content caching corporations；Two is the method determining content caching node in corporations；

The method determining content caching corporations: in same controller, is first depending on corporations' importance degree, corporations is ranked up from big to small, if the corporations' set in a controller is IC={IC_i| i=1,2 ..., C}, C represents the number of corporations under same controller, as to any i < j, then IC_i<IC_j, then, according to the content popularit under controller by content be divided into C interval, compare content popularit under the popularity of the content that need to store and whole controller, it is determined that its affiliated interval, so that it is determined that the buffer memory corporations of content simultaneously；

Content caching nodal method is determined: after the corporations completing to need cache contents institute buffer memory, for same corporations, first arranged in descending order by node according to node corporations importance degree, such as IN={IN in corporations_i| i=1,2 ..., N_Ci, then, divide the content into N according to the popularity of the content in corporations_CiIndividual interval, compares popularity and the content popularit under whole corporations of the content that need to store, it is determined that its affiliated interval simultaneously, so that it is determined that the cache node of content；

(2) the replacement content in node is determined.Node corporations importance degree according to node and the popularity of content object, select different content objects to be replaced；Namely determining, by node corporations importance degree, the probability that the content object in nodal cache queue is replaced, node as big in node corporations importance degree, the probability that the content object that popularity is low in buffer queue is replaced is big；And the node that pitch point importance degree is little, the probability that the content object that popularity is high in buffer queue is replaced is big, further improves the storage efficiency of spatial cache；If the spatial cache of selected ICN node is sized to NC object.And the node corporations importance degree of this node is IN_i, and the importance degree of the corporations belonging to this node is IC_j, then this content object is when this node needs buffer memory, will according to Probability p_ijK () replaces the content (k ∈ [1, C]) of the kth position of this nodal cache queue, p_ijK () defines such as formula (1)；

$p_{ij}\left(k\right)=\frac{1}{\mathrm{\&alpha;}}{\left(\frac{{\mathrm{IN}}_i}{{\mathrm{IC}}_j}\right)}^{\mathrm{\&beta;}k}---\left(1\right)$

Wherein, α is normalization factor, meets formula (2), and β is probability adjustment factor.

$\underset{k=1}{\overset{C}{\&Sigma;}}{p}_{ij}\left(k\right)=1\&DoubleRightArrow;\mathrm{\&alpha;}=\underset{k=1}{\overset{C}{\&Sigma;}}{\left(\frac{{\mathrm{IN}}_i}{{\mathrm{IC}}_j}\right)}^{\mathrm{\&beta;}k}---\left(2\right)$

Step3: caching method is issued to switch stream table and reaches data surface, complete the execution of caching method.