KR101448586B1 - Method for cloud network resource management, cloud server and cloud network system using the same - Google Patents

Abstract

A cloud network resource management method according to the present invention is a cloud network resource management method in which data is distributed and stored in a plurality of cloud servers. The cloud server receives a request for at least one packet, Based on the inbound response contribution according to the history of each cloud server that has made a packet request to the cloud server in response to the packet request of the cloud server in the past and the drift rate measure regarding the operational ratio of the operational resource of the cloud server, Determining a combination of packet requests that maximize the server utility of the cloud server, determining a combination of the packet requests to the cloud servers that requested the packets in accordance with the determined combination of packet requests, Lt; RTI ID = 0.0 > It may comprise the step of responding.

Description

METHOD FOR CLOUD NETWORK RESOURCE MANAGEMENT, CLOUD SERVER AND CLOUD NETWORK SYSTEM USING THE SAME,

The present invention relates to a cloud network system, and more particularly, to a resource management strategy of a cloud network system.

In addition to the rapidly changing IT technology, smartphones and tablet PCs have spread, and as a result, not only large files such as ordinary files such as web, email and documents, but also large files such as video, The demand of users for multimedia streaming data such as multimedia streaming is increasing exponentially. This has led many network service providers to worry about securing network resources that can provide these large amounts of data to users.

Cloud computing techniques that can efficiently use large amounts of network resources are attracting attention as an alternative to solve these problems.

Cloud computing services, which can integrate various network resources with virtualization technology to form a single cloud network and provide users with various data services through the Internet, are now becoming popular among data service providers.

Cloud computing services can effectively cope with the surge in demand and supply of data, and can supply network data services more effectively even in the era of accelerating big data.

According to data center IP traffic forecasts from Cisco (CISCO), a leading provider of network equipment, the total capacity of the world's data centers has already reached 2.6 ZB (ZB = 1 billion TB) in 2012 and 6.6 ZB . According to this document, about three-quarters of the traffic that originates in the data center is generated by operations such as storage, classification, generation, retrieval, and authentication within the data center, with only one fourth occurring between the data center and the outside Traffic.

In particular, according to CISCO, Internet traffic will more than quadruple in 2014, with the largest portion of video expected to come from video, which is typically delayed-sensitive data, so that multimedia streaming data, including video, Avoid overloading servers in the cloud while ensuring fast and reliable transmission, and using resources as efficiently as possible will determine the feasibility of cloud network services.

Therefore, in order to efficiently operate a cloud network system, an effective resource management strategy is required to cope with such an enormous load.

The present invention is directed to a cloud network resource management method capable of efficiently processing delay sensitive data, not overloading due to load of resource management itself, and capable of flexibly coping with scale variation of a cloud network, And a cloud network system using the same.

According to an aspect of the present invention, there is provided a method of managing a resource of a cloud network in which data is distributedly stored in a plurality of cloud servers,

Wherein the cloud server receiving the at least one packet request receives the at least one packet from the at least one cloud server based on the importance of the data corresponding to each packet request and inbound according to the history in which each of the other cloud servers that have made a packet request to the cloud server responded to the packet request of the cloud server in the past Computing a server utility of the cloud server for each combination of packet requests based on a response rate and a drift rate metric on the operational rate of operational resources of the cloud server;

Determining a combination of packet requests that maximizes a server utility of the cloud server; And

And responding to the packet requests by sending the required packets corresponding to the determined combination of packet requests to the requesting cloud servers, respectively.

According to one embodiment, the importance of the data can be quantified as the relative importance of the data corresponding to the packet request is delay sensitive.

According to one embodiment, the inbound response contribution may be determined based on the history that the other cloud servers have responded to in the past for the packet request made by the cloud server.

According to one embodiment, the inbound response contribution is determined based on the importance of data corresponding to a packet request that the cloud server has made in the past, for other cloud servers that have previously responded to a packet request by the cloud server Can be determined.

According to one embodiment, the inbound response contribution is determined according to the importance of the data requested by the cloud server at the immediately preceding time to the other cloud servers that the cloud server has previously responded to a packet request in the past .

According to one embodiment, the drift rate metric is a measure of the computational resources in which the cloud server is running, and may be quantified such that the cloud server is relatively advantageous as it uses computational resources energy efficient or speed efficiently.

According to one embodiment, the drive rate measure

When the operation ratio is at a predetermined appropriate level, it is positively applied to the server utility,

If it is smaller than the predetermined proper level, it is positively applied to the server utility,

If the predetermined level is exceeded, the value may be negatively applied to the server utility.

According to one embodiment, the server utility for a particular combination of packet requests is a weighted sum of the importance, inbound response contribution, and drive rate metrics of the data corresponding to each combination of packet requests for each packet request Can be calculated by integrating the values.

According to one embodiment, the server utility is calculated by the following equation

Lt; / RTI >

는 공급자인 클라우드 서버

의 서버 효용도이고,

는 패킷을 요청한 수요자인 클라우드 서버들

의 총 개수이며, 데이터 중요도

는 공급자인 클라우드 서버

에 요청된 각각의 데이터

의

번째 패킷

에 관한 데이터 중요도이고, 인바운드 응답 기여도

는 직전 시점에서 수요자였던 클라우드 서버

가 공급자였던 클라우드 서버

에게 했던 요청에 기초한 클라우드 서버

의 인바운드 응답 기여도이며,

는 클라우드 서버

의 구동율 측도이고,

,

,

는 각각 데이터의 중요도

, 인바운드 응답 기여도

및 구동율 측도

의 가중치들일 수 있다.Here, the server utility

Lt; RTI ID = 0.0 >

Of the server utility,

Lt; RTI ID = 0.0 > servers < / RTI &

, And the data importance

Lt; RTI ID = 0.0 >

For each of the requested data

of

Th packet

, And the inbound response contribution

The cloud server

The cloud server

Cloud server based on request to

Inbound response contribution,

The cloud server

Of the driving rate,

,

,

The importance of each data

, Inbound response contribution

And drive ratio measure

Lt; / RTI >

According to one embodiment, the cloud network resource management method comprises:

The method of claim 1, wherein, if some of the data related to the service provided by the cloud server is insufficient, the cloud server may determine at least one of the at least one other cloud server to make a packet request based on the outbound response contribution of the cloud server to each of the other cloud servers. Selecting a cloud server; And

The cloud server may further include a step of requesting a packet necessary for the selected cloud server.

According to one embodiment, the outbound response contribution may be determined based on the history that the cloud server responded to each packet request of other cloud servers in the past.

According to one embodiment, the outbound response contribution is determined according to the importance of data corresponding to each packet request of the other cloud servers to which the cloud server has responded in the past,

The importance of the data can be quantified to be relatively important as the delay sensitivity of the data.

According to one embodiment, the outbound response contribution may be determined according to the importance of data corresponding to a packet request from other cloud servers that the cloud server has responded to immediately before.

According to another aspect of the present invention, there is provided a method for managing a resource of a cloud network in which data is distributedly stored in a plurality of cloud servers,

In the event that some of the data related to the service provided by the first cloud server is insufficient, the first cloud server may send a second cloud server to make a packet request based on the outbound response contribution for each of the other cloud servers Selecting step;

The first cloud server requesting a packet necessary for the selected second cloud server;

The second cloud server receiving the at least one packet request transmits the importance of the data corresponding to each packet request and the importance of the data to the second cloud server in response to the request of the second cloud server in the past Computing a server utility value of the second cloud server for each combination of packet requests based on the inbound response contribution according to the history of the first cloud server and the driving rate measure regarding the operational rate of the operational resource of the second cloud server;

Determining a combination of packet requests that maximizes a server utility of the second cloud server; And

And responding to packet requests by sending each of the required packets for the cloud servers corresponding to the determined combination of packet requests.

According to another aspect of the present invention, there is provided a cloud server in which data is distributedly stored in a cloud network,

An importance of data corresponding to each packet request, an inbound response according to the history in which each of the other cloud servers that have made a packet request to the cloud server responded to the packet request of the cloud server in the past when receiving at least one packet request, A server utility calculating unit for calculating a server utility of the cloud server with respect to each combination of packet requests based on a contribution rate and a drive rate measure related to an operation ratio of operation resources of the cloud server;

A response partner selection unit for determining a combination of packet requests that maximize the server utility; And

And a request response interface that responds to packet requests by sending the required packets corresponding to the determined combination of packet requests to the requesting cloud servers, respectively.

According to one embodiment, the importance of the data can be quantified as the relative importance of the data corresponding to the packet request is delay sensitive.

According to one embodiment, the inbound response contribution is determined based on the importance of data corresponding to a packet request that the cloud server has made in the past, for other cloud servers that have previously responded to a packet request by the cloud server Can be determined.

According to one embodiment, the drift rate metric is a measure of the computational resources in which the cloud server is running, and may be quantified such that the cloud server is relatively advantageous as it uses computational resources energy efficient or speed efficiently.

According to one embodiment, the server utility for a particular combination of packet requests is a weighted sum of the importance, inbound response contribution, and drive rate metrics of the data corresponding to each combination of packet requests for each packet request Can be calculated by integrating the values.

According to one embodiment, the server utility is calculated by the following equation

Lt; / RTI >

는 공급자인 클라우드 서버

의 서버 효용도이고,

는 패킷 요청을 한 수요자인 클라우드 서버들

의 총 개수이며, 데이터 중요도

는 공급자인 클라우드 서버

에 요청된 각각의 데이터

의

번째 패킷

에 관한 데이터 중요도이고, 인바운드 응답 기여도

는 직전 시점에서 수요자였던 클라우드 서버

가 공급자였던 클라우드 서버

에게 했던 요청에 기초한 클라우드 서버

의 인바운드 응답 기여도이며,

는 클라우드 서버

의 구동율 측도이고,

,

,

는 각각 데이터의 중요도

, 인바운드 응답 기여도

및 구동율 측도

의 가중치들일 수 있다.Here, the server utility

Lt; RTI ID = 0.0 >

Of the server utility,

Lt; RTI ID = 0.0 > server < / RTI >

, And the data importance

Lt; RTI ID = 0.0 >

For each of the requested data

of

Th packet

, And the inbound response contribution

The cloud server

The cloud server

Cloud server based on request to

Inbound response contribution,

The cloud server

Of the driving rate,

,

,

The importance of each data

, Inbound response contribution

And drive ratio measure

Lt; / RTI >

According to one embodiment,

A requesting party selection unit for selecting at least one other cloud server to request a packet based on the outbound response contribution for each of the other cloud servers when a portion of the data related to the service provided by the cloud server is insufficient Further included,

And the request response interface may be operable to request packets for the selected cloud server.

According to one embodiment, the cloud server records the importance of the packet requested data and the identification information of the cloud server selected as the requesting party at the time of packet request, and at the time of packet response, the importance of data and response And a request response history management unit for managing the request response history by recording the identification information of the cloud server selected as the other party.

According to another aspect of the present invention, there is provided a cloud network system for distributedly storing data in a plurality of cloud servers,

Wherein each of the plurality of cloud servers comprises:

An importance of data corresponding to each packet request, an inbound response according to the history in which each of the other cloud servers that have made a packet request to the cloud server responded to the packet request of the cloud server in the past when receiving at least one packet request, A server utility calculating unit for calculating a server utility of the cloud server with respect to each combination of packet requests based on a contribution rate and a drive rate measure related to an operation ratio of operation resources of the cloud server;

A response partner selection unit for determining a combination of packet requests that maximize the server utility; And

And a request response interface that responds to packet requests by sending the required packets corresponding to the determined combination of packet requests to the requesting cloud servers, respectively.

According to the cloud network resource management method, the cloud server, and the cloud network system using the same, the delay sensitive data such as the multimedia streaming data can be efficiently processed.

According to the cloud network resource management method, the cloud server, and the cloud network system using the cloud network resource management method of the present invention, in a cloud network, each of the unit servers independently manages optimal resource management without any centralized management server Optimized resource management can also be achieved at the global cloud level.

According to the cloud network resource management method, the cloud server, and the cloud network system using the cloud network resource management method of the present invention, it is possible to pursue flexible and optimized resource management without being affected by the scale change such as expansion or reduction of the cloud network resource.

According to the cloud network resource management method, the cloud server, and the cloud network system using the cloud network resource management method of the present invention, the resource required for the optimized resource management by the individual unit server is very small and the overhead due to the resource management itself rarely occurs .

1 is a conceptual diagram illustrating a cloud network system according to an embodiment of the present invention.
2 is a conceptual diagram illustrating data stored in a cloud network in a cloud network resource management method according to an exemplary embodiment of the present invention.
3 is a conceptual diagram illustrating a data packet that is requested by a certain cloud server to another cloud server in the cloud network resource management method according to an embodiment of the present invention.
FIG. 4 is a conceptual diagram illustrating a plurality of data packet requests that a certain cloud server receives simultaneously from other cloud servers in a cloud network resource management method according to an exemplary embodiment of the present invention.
5 is a conceptual diagram illustrating optional data packet responses performed by an arbitrary cloud server to other cloud servers in a cloud network resource management method according to an exemplary embodiment of the present invention.
6 is a flowchart illustrating a cloud network resource management method according to an exemplary embodiment of the present invention.
FIG. 7 is a graph illustrating simulation results of a cloud network resource management method according to an exemplary embodiment of the present invention. FIG. 7 is a graph illustrating a comparison of the number of repeated data requests until there is a response according to data importance when the congestion is normal.
FIG. 8 is a graph illustrating simulation results of a cloud network resource management method according to an exemplary embodiment of the present invention. FIG. 8 is a graph comparing the number of repeated data requests until there is a response according to data importance when congestion is severe.
FIG. 9 is a graph illustrating simulation results of a cloud network resource management method according to an exemplary embodiment of the present invention and comparing CPU drive ratios of the cloud server when the congestion is normal.
10 is a block diagram illustrating a cloud server according to an embodiment of the present invention.

For the embodiments of the invention disclosed herein, specific structural and functional descriptions are set forth for the purpose of describing an embodiment of the invention only, and it is to be understood that the embodiments of the invention may be practiced in various forms, The present invention should not be construed as limited to the embodiments described in Figs.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The same reference numerals are used for the same constituent elements in the drawings and redundant explanations for the same constituent elements are omitted.

1 is a conceptual diagram illustrating a cloud network system according to an embodiment of the present invention.

The cloud network system 10 of the present invention includes a plurality of cloud servers 11 and a cloud core domain 12 and is capable of providing various cloud services to any user terminals 30 via the network 20 A software as a service (SaaS), a platform as a service (PaaS), or an infrastructure (IaaS: Infrastructure as a Service).

The cloud core domain 12 includes a virtualization device, a data distribution management server, ultra-high speed switches and a firewall. For the outside, the user terminal 30 provides the cloud service as if it is provided from one server. Can be managed.

In this case, the services or data provided by the cloud network system 10 may be appropriately distributed or executed or stored in packets by each of the plurality of cloud servers 11 by the cloud core domain 12.

Data distributed and stored in units of packets may have a predetermined importance and a number of packets. For example, if the data is delay sensitive, it can be given relatively high importance. Also, for convenience of management, packet size can be constant.

Each of the cloud servers 11 stores distributed packets of data relating to at least one service. According to the packet distribution storage policy of the cloud core domain 12, each of the cloud servers 11 stores certain data You may have all of your packets, but you may have only a few, or none at all. In addition, packets of certain data may be redundantly stored in the plurality of cloud servers 11.

2 and 3, FIG. 2 is a conceptual diagram illustrating data stored and stored in a cloud network in a cloud network resource management method according to an embodiment of the present invention. FIG. FIG. 2 is a conceptual diagram illustrating a data packet that a certain cloud server requests to another cloud server in the cloud network resource management method according to the present invention.

In Fig. 2, N _d kinds of data are stored in a certain cloud network system 10. Assume that data is divided and stored in units of packets of a certain size. It is assumed that one kind of data n is divided into K _n packets and distributedly stored in the plurality of cloud servers 11. Each of the cloud servers 11 may have some or all of the respective data, or none at all.

에 대해 어떤 데이터 n의 요청이 있을 경우에, 클라우드 서버

는 데이터 n의 패킷들을 다수 가지고 있지만 일부 패킷들을 가지고 있지 않은데, 다른 클라우드 서버들에 요청하여야 하는 점선으로 표시된 k 번째 패킷을

라고 표현할 수 있다.In Fig. 3, the jth cloud server

In the event that there is a request for any data n for the cloud server

Has a number of packets of data n but does not have some packets, and the kth packet indicated by the dotted line that should be requested to other cloud servers

.

는 부족한 하나 이상의 패킷들을 동시에 여러 클라우드 서버들에 요청하고 그들 중 적어도 일부로부터 필요한 패킷들을 수신한다면, 데이터 획득 시간을 줄일 수 있을 것이다.At this time,

Will be able to reduce data acquisition time if one or more packets that are deficient are requested to multiple cloud servers at the same time and receive necessary packets from at least some of them.

On the other hand, when data is required to provide a service to the user terminal 30, each of the cloud servers 11 can first search whether the data has all the data.

Next, each of the cloud servers 11 can request a packet to another cloud server 11, and the cloud server 11 receiving the request can transmit the packet in response to the request. Here, the algorithm of the packet request and the algorithm of the request response may appear.

4 and 5, FIG. 4 illustrates a cloud network resource management method according to an exemplary embodiment of the present invention, in which a cloud server illustrates a plurality of data packet requests received simultaneously from other cloud servers FIG. 5 is a conceptual diagram illustrating optional data packet responses performed by an arbitrary cloud server to other cloud servers in a cloud network resource management method according to an exemplary embodiment of the present invention. Referring to FIG.

4, the first cloud server 111 and the third cloud server 113 are users who request data packets necessary for providing a cloud service to the user terminal 30, and the second cloud server 112 and the third cloud server 113 A number of cloud servers, including the four cloud server 114, are providers that can hold the necessary data packets and respond to requests.

However, the first to fourth cloud servers 111, 112, 113, and 114 are required to cooperate with each other to play a role of a consumer by accidentally requiring a specific packet at a certain point in time, And not the supplier, and it is also the consumer and supplier according to the situation.

In FIG. 4, for example, with respect to any popular video data n from which a service request is received from a certain user terminal 30, the first cloud server 111 or the third cloud server 113, The second cloud server 112 or the fourth cloud server 114, respectively.

At this time, if the first cloud server 111, which is a consumer, unconditionally makes a request to the neighboring cloud servers 11, each of the first cloud servers 111 may also receive a response to their request in time And the network resources of the cloud network system 10 as a whole will be greatly wasted.

Accordingly, the first cloud server 111, which is a consumer, has a problem of determining which one of the neighboring cloud servers 11 to request a packet to the specific cloud server 11, A packet request algorithm is required to select the second and fourth cloud servers 112 and 114 to send a request among the first and second cloud servers 11 and 11, respectively.

In addition, if the second cloud server 112 or the fourth cloud server 114 has sufficient available resources, it may respond to all requests and send packets, but normally the second cloud server 112 Since the fourth cloud server 114 itself will also correspond to a service request from a certain user terminal 30, the fourth cloud server 114 may request the service from the first cloud server 111 or the third cloud server 113 You will only be able to respond to some of the requests, and therefore a request response algorithm is needed to select the request to respond.

5, the second cloud server 112 is now connected to some of the cloud servers 11 among the plurality of cloud servers 11 including the first cloud server 111 and the third cloud server 113, Lt; RTI ID = 0.0 > 111, < / RTI >

Referring to FIGS. 4 and 5, the first cloud server 111 selectively requests a short packet to other cloud servers 11 having a short packet, and a second cloud Server 112 may optionally respond to packet requests.

In particular, in the present invention, the packet requesting algorithm may be configured such that the first cloud server 111 selects the second and fourth cloud servers 112, 114 to send a request among the cloud servers 11, The cloud server 111 requests a packet among the plurality of cloud servers 11 according to the past outbound response contribution of the first cloud server 111 itself to each of the cloud servers 11 2 and the fourth cloud servers 112 and 114, respectively.

In other words, it is a strategy to preferentially choose the person whom you have previously answered.

To this end, each of the cloud servers 11 receives a request for a packet in the past and responds to the request as a provider role, thereby allowing the remaining cloud servers 11 to transmit the history of the contribution made by the other parties to the data assembly .

Specifically, the outbound response contribution of the first cloud server 111 to the second cloud server 112 is such that each second cloud server 112 has previously requested a packet to the first cloud server 111, 1 how much the first cloud server 111 contributed for the second cloud server 112 based on the history that the first cloud server 111 provided packets to the second cloud server 112 It is a value that is given to be able to quantify.

In particular, in one embodiment, the outbound response contribution granted to the second cloud server 112 may be given relatively high as the importance of the data packet to which the first cloud server 111 has responded to the request is high .

In other words, it is a strategy to request an important data packet to the other party that has previously responded to the important data packet.

According to the embodiment, instead of considering all past hysteresis, the outbound response contribution can be determined taking into account only the previous performance.

가 현재 공급자 클라우드 서버

에 대해 부여하는 아웃바운드 응답 기여도

는 가장 최근인 직전 시점(t-1)에 공급자로서 클라우드 서버

가 수요자로서 클라우드 서버

에 응답을 해 주었던 요청에 상응하는 패킷의 데이터 중요도에 따라 결정될 수 있다. In other words, the current consumer, the cloud server

Current provider cloud server

Outbound response credit given for

Lt; RTI ID = 0.0 > (t-1) < / RTI &

As a consumer,

May be determined according to the data importance of the packet corresponding to the request that responded to the request.

If so, the criterion of the importance of the data may be a problem.

According to the survey, users who use the network service generally begin to feel dissatisfied when the service response time is delayed for more than 3 seconds, and when they are delayed for more than 10 seconds, they easily depart from the service. In addition, if a buffer underrun occurs during playback of media data such as a music video or a movie, the user is not satisfied with such a media service.

Therefore, the importance of the data can be easily recognized when the service is delayed, while the user considers that the service is delayed when the service is delayed. In addition, .

For example, if the video data having a high number of views and a high number of views are high in popularity, users may be dissatisfied with the service if they are delayed, so that they can have a relatively high importance level than the low-volume document data without popularity.

의 중요도

는 수학식 1과 같이 수요자인 어떤 클라우드 서버

가 필요로 하는 데이터 n의 k 번째 패킷인 데이터 패킷

의 함수로써 표현될 수 있다.More generally,

Importance of

As shown in Equation (1), a certain cloud server

Which is the kth packet of the data n required by the data packet

Can be expressed as a function of.

는 최저값

부터 최고값

까지 P 단계로 설정된 값 중 하나인

로 특정될 수 있다. 예를 들어 데이터의 중요도는 최저 1부터 최고 10 사이의 10 단계로 설정된 정수로 특정될 수 있다.Here,

Is the lowest value

Highest value from

One of the values set in step P up to

. ≪ / RTI > For example, the importance of the data can be specified by integers set to 10 levels from 1 to 10.

는 공급자인 클라우드 서버

가 동시적인 데이터 패킷 요청들을 받을 경우에 클라우드 서버

가 우선적으로 응답할 데이터 패킷을 선택하는 기준이 될 수 있다.On the other hand,

Lt; RTI ID = 0.0 >

Lt; / RTI > receives simultaneous data packet requests,

May be a criterion for selecting a data packet to respond preferentially.

Further, if the first cloud server 111 that has successfully responded to the request of the second cloud server 112 with respect to the packet of the data of high importance, the second cloud server 112 transmits the data of high importance at a time And it contributes greatly to providing the service to the user without delay. On the other hand, even if the first cloud server 111 responds successfully within a short period of time to the request of the second cloud server 112 for packets of data of low importance, the second cloud server 112 The first cloud server 111 has a low contribution to the second cloud server 112 because a quick response is not necessary.

For example, when the first cloud server 111 has successfully responded to a request from the second cloud server 112 regarding a packet having insufficient data of importance 10 in the past, for example, the second cloud server 112, The outbound response contribution to the user can be determined to be 10.

The outbound response contribution to the second cloud server 112 of the first cloud server 111 is determined by the data packet transmitted by the first cloud server 111 in response to the request of the second cloud server 112 in the past, As shown in FIG.

For this packet requesting algorithm, each of all the cloud servers 11 may continuously provide outbound response contributions for each of the remaining cloud servers 11. [

Meanwhile, the cloud server 11 acting as a supplier can simultaneously receive a packet request from other cloud servers 11 serving as a plurality of customers.

In this case, the cloud server 11 acting as a provider can preferentially respond to some requests rather than responding to all of the packet requests of the cloud servers 11 serving as the consumer. In the cloud network system 10, For maximum efficiency, it is necessary to effectively select which request to respond first.

If the cloud servers 11 acting as all suppliers perform their own best responses, such a cloud network system 10 may be said to be the most efficient operating state as a whole.

In view of this, the request response algorithm of the present invention is a method in which the cloud server 11, which is the provider, selects the utility of the response in selecting packet requests to be answered during requests of the cloud servers 11, It is an algorithm that determines the combination of packet requests to make larger.

Specifically, the second cloud server 112, which is the provider, firstly determines the importance of the requested data, second, the inbound response contribution of each of the cloud servers 11 that made the packet request, and third, It is possible to determine a combination of packet requests that maximize the server utility of the second cloud server 112, which is a provider calculated based on the driving rate measure of the second cloud server 112 itself.

는 지연 민감성 정도에 따라 상대적인 값으로 데이터마다 적절하게 부여될 수 있다. As described earlier, the importance of the requested data

May be appropriately given for each data with a relative value depending on the degree of delay sensitivity.

는, 과거에 수요자로서 클라우드 서버

의 요청에 대해 공급자로서 클라우드 서버

가 응답한 내력에 기초하여, 현재의 공급자인 클라우드 서버

가 현재의 수요자인 클라우드 서버

에 대하여 과거에 기여를 받은 정도를 수치화한 값으로서, 실시예에 따라 인바운드 응답 기여도는 과거에 클라우드 서버

로부터 응답을 받았던 데이터의 중요도에 따라 결정될 수 있다.Inbound response contribution

In the past, as a consumer,

Lt; RTI ID = 0.0 > server < / RTI &

On the basis of the internal history of the cloud server

Lt; RTI ID = 0.0 >

And the inbound response contribution rate according to the embodiment is the value obtained in the past by the cloud server

And the importance of the received data.

In other words, the second cloud server 112, which is a provider, is configured so that when the first cloud server 111, which is the consumer at the present time, has been a provider in the past, the second cloud server 112 The inbound response contribution to the current consumer, the first cloud server 111, may be determined according to the history of the response to the first cloud server 111.

According to the embodiment, instead of considering all past hysteresis, the inbound response contribution can be determined considering only the previous response.

는 현 시점에 수요자인 클라우드 서버

의 요청에 대하여, 직전 시점 t-1에 공급자였던 클라우드 서버

가 직전 시점 t-1의 클라우드 서버

의 요청에 응답해 준 데이터의 중요도를 가지고, 현 시점의 수요자인 클라우드 서버

의 인바운드 응답 기여도

를 결정할 수 있다.In other words, at this point,

At the present time, the cloud server

To the cloud server < RTI ID = 0.0 >

Lt; RTI ID = 0.0 > t-1 < / RTI &

Which is the consumer of the current time,

Inbound response contribution of

Can be determined.

은 공급자인 제2 클라우드 서버(112)의 연산 자원의 가동 비율에 관한 측도(measure)이다.Finally,

Is a measure of the operating ratio of the computational resources of the second cloud server 112 that is the supplier.

이 50%일 때에 서비스 시간이 2초 소요되고 CPU 구동율

이 70%인 때에는 서비스 시간이 3초 소요된다고 하면, CPU 구동율

이 90%일 때에 서비스 시간은 비례에 따라 4초 소요되는 것이 아니라 그보다 훨씬 긴 10초 이상 소요된다고 한다. 또한 관찰에 따르면 CPU 구동율

이 70%가 넘으면 응답 속도는 느려질 뿐 아니라 불규칙적으로 변하기도 한다.According to the survey, the CPU operating rate

Is 50%, the service time is 2 seconds and the CPU drive rate

Is 70%, assuming that the service time takes 3 seconds, the CPU driving rate

At 90%, the service time does not take 4 seconds in proportion, but takes longer than 10 seconds. Also, according to observations,

If it exceeds 70%, the response speed will not only slow but also change irregularly.

이 높을수록 에너지 효율적일 것이므로, 클라우드 서버(11)의 구동율 측도

는 지연 시간 및 에너지 효율 측면들에서 다소 트레이드오프 관계에 있다고 할 수 있다. 다만 CPU 부하가 작을 때에 소비 전력을 줄일 수 있는 다른 전력 관리 기술들을 적용할 수 있고, 앞의 예시에서 CPU 구동율

이 90%일때 서비스 지연 시간이 길어지면 결과적으로 동일한 에너지로 수행한 작업량은 오히려 줄어들 수 있으므로, 이 경우에 에너지 효율보다는 지연 시간이 더 큰 문제라고 할 수 있다.As described above, if the service delay time is 10 seconds or more, the user is dissatisfied with the service, and on the other hand, the CPU drive rate

The energy efficiency of the cloud server 11 can be improved,

Can be said to be somewhat trade-off in terms of delay time and energy efficiency. However, other power management techniques that can reduce power consumption when the CPU load is small can be applied, and in the above example,

Is 90%, the longer the service delay time, the lower the work done with the same energy, so the delay time is bigger than the energy efficiency in this case.

은 어떤 적정 값, 예를 들어 70%를 유지하거나 넘지 않는 것이 바람직하다.Therefore, the CPU driving rate

Is preferably maintained at or above a certain appropriate value, for example 70%.

가 부담하는 부하는 거의 동일하다고 볼 수 있으므로, 어떤 공급자 클라우드 서버

의 CPU 구동율

은 공급자 클라우드 서버

가 처리해야 하는 요청 및 응답의 수가 늘어나면 비례적으로 증가한다고 할 수 있다. 이는 다음 수학식 2와 같이 표현될 수 있다.On the other hand, for requesting and responding to packets of the same size,

The load on the provider cloud server is almost the same,

CPU drive rate

Provider cloud server

Can be said to increase proportionally as the number of requests and responses to be processed increases. This can be expressed by the following equation (2).

는 현 시점 t에 공급자 클라우드 서버

가 처리할 응답들에 상응하는 패킷들의 집합이고,

는 현 시점 t에 공급자 클라우드 서버

가 처리하는 응답의 개수이며,

는 소정의 비례 상수이다.here

Lt; RTI ID = 0.0 > t < / RTI &

Is a set of packets corresponding to the responses to be processed,

Lt; RTI ID = 0.0 > t < / RTI &

Is the number of responses processed by the <

Is a predetermined proportional constant.

을 총 M 개의 구간 별로 나누고 수치화한 구동율 측도

를 도입할 수 있다.At this time, for convenience of calculation, the CPU driving rate

Is divided into a total of M sections and a driving rate measure

Can be introduced.

의 CPU 구동율

을 대표하는 구동율 측도

는 M+1 개의 경계값들

중 이웃하는 두 개의 경계값들

로 구분되는 M 개의 구간들을 각각 대표하는 M 개의 측도값들

,

, ...,

, ...,

중 하나의 측도값

을 가질 수 있다.Here, the provider cloud server

CPU drive rate

Lt; RTI ID = 0.0 >

Lt; RTI ID = 0.0 > M + 1 &

Two neighboring boundary values

M < / RTI > measures < RTI ID = 0.0 >

,

, ...,

, ...,

≪ / RTI >

Lt; / RTI >

는 제2 클라우드 서버(112)가 연산 자원을 에너지 효율적 또는 속도 효율적으로 사용할수록 상대적으로 유리하도록 수치화될 수 있다.According to an embodiment,

Can be quantified to be relatively advantageous as the second cloud server 112 uses computational resources energy-efficient or speedily.

은 공급자 클라우드 서버

에서 최적의 에너지 내지 지연시간 효율을 보이는 구동율 구간에서 양의 값을 부여받고, 반면에 과부하를 일으키거나 서비스 지연 내지 서버의 다운을 초래할 수 있는 구동율 구간들에서는 음의 값을 부여받을 수 있다.For example,

Provider cloud server

A positive value can be given in a driving rate interval showing optimal energy or delay time efficiency, while a negative value can be given in driving rate intervals in which an overload occurs or a service delay or a server is brought down .

이 음수인 경우에는 해당 클라우드 서버의 효용도가 감소될 것이고, 그 결과 상대적으로 낮은 효용도를 산출시키는 패킷 요청 및 응답의 조합은 탈락하게 된다.For example,

If the number is negative, the utility of the corresponding cloud server will be reduced, and as a result, the combination of the packet request and response that yields a relatively low utility will be dropped.

은 CPU 구동율

구간마다 데이터 중요도

의 최대값

및 최소값

의 함수로 부여될 수 있다. 수학식 3은 그러한 구동율 측도

의 측도값들

을 예시한 것이다.Further, according to an embodiment, each of the measure values

Is the CPU drive rate

Data Importance per Section

Maximum value of

And minimum value

As shown in FIG. Equation (3) shows that such a drive ratio measure

≪ / RTI >

.

를 구성하는 변수들, 즉 데이터 중요도, 인바운드 응답 기여도 및 구동율 측도는 서로 동등한 수리적 의미를 가질 수 있다. 수학식 3에서 구동율 측도

는 CPU 구동율이 50%에서 70% 사이일 때 가장 높고, CPU 구동율이 50% 미만이면 양호한 수준이지만, CPU 구동율이 70%를 초과하면 음수로 부여된다.In this case, the server utility of Equation (4)

The data importance, the inbound response contribution, and the drift rate measure may have mathematical meanings equivalent to each other. In equation (3), the driving-

Is the highest when the CPU driving rate is between 50% and 70%, and is good when the CPU driving rate is less than 50%, but is negative when the CPU driving rate exceeds 70%.

, 인바운드 응답 기여도

및 구동율 측도

가 결정되면, 각각의 가능한 경우에 대해 공급자 클라우드 서버

의 서버 효용도

는 총

개의 수요자 클라우드 서버들

이 공급자 클라우드 서버

에게 요청한 각각의 데이터

의

번째 패킷

의 중요도

, 직전 시점에서 수요자였던 클라우드 서버

가 공급자였던 클라우드 서버

에게 했던 요청에 기초한 클라우드 서버

의 인바운드 응답 기여도

및 구동율 측도

의 함수로서, 예를 들어, 다음 수학식 4와 같이 연산될 수 있다.Thus, for each possible packet request and response, the importance of the data

, Inbound response contribution

And drive ratio measure

For each possible case, the provider cloud server < RTI ID = 0.0 >

Server utility

Gun

Consumer cloud servers

This provider cloud server

Each data requested to

of

Th packet

Importance of

, The cloud server

The cloud server

Cloud server based on request to

Inbound response contribution of

And drive ratio measure

For example, the following equation (4).

가 가장 크게 되도록 클라우드 서버들

의 패킷 요청 및 응답 조합이 결정될 수 있다. 이때,

,

,

는 각각 데이터의 중요도

, 인바운드 응답 기여도

및 구동율 측도

의 가중치들로서, 상황에 따라 적절하게 가변할 수 있다.According to Equation (4), the server utility

The cloud servers < RTI ID = 0.0 >

Lt; / RTI > may be determined. At this time,

,

,

The importance of each data

, Inbound response contribution

And drive ratio measure

And can be appropriately varied depending on the situation.

In this embodiment, the server utility is determined by how important the currently requested data is, how important the data that the other party has responded to in the past, and how much computing resources are available.

Since the exemplary server utility calculation formula according to Equation (4) is only a few thousands of times of arithmetic operations in hundreds of times, the calculation resources to be burdened by the cloud server for calculating the server utility and selecting the response partner are small, Management overhead is negligible.

In this way, the second cloud server 112, as a provider, can select packet requests and respond to selected packet requests to maximize server utility.

6 is a flowchart illustrating a cloud network resource management method according to an exemplary embodiment of the present invention.

6, in the cloud network resource management method of the present invention, when some packets among the data related to the service provided by the first cloud server 111 to the user terminal 30 are insufficient in step S61, Based on the outbound response contribution of the first cloud server 111 to each of the other cloud servers 11 of the second cloud server 11, the second cloud server 112 to make the packet request.

Here, the outbound response contribution can be determined based on the history in which the first cloud server 111 responded to each request of the cloud servers 11 in the past.

Depending on the embodiment, the outbound response contribution may be determined according to the importance of the data corresponding to the request of the cloud servers 11 that the previous first cloud server 111 has responded to.

Here, the importance of the data can be quantified to be relatively important as the delay sensitivity of the data.

In this embodiment, the outbound response contribution means whether the first cloud server 111 has responded to the request of the critical cloud server 11 in the past.

Also, according to an embodiment, the outbound response contribution may be determined according to the importance of the data corresponding to the request of the cloud servers 11 that the first cloud server 111 has responded to immediately before.

In step S62, the first cloud server 111 requests a packet necessary for the selected second cloud server 112. [

In step S63, the second cloud server 112 receiving the at least one packet request transmits the importance of the data corresponding to each packet request to the cloud servers 11 (including the first cloud server 111) Based on the inbound response contribution according to the history that each has responded to the request of the second cloud server 112 in the past and the drift rate measure with respect to the operational ratio of the operational resources of the second cloud server 112, The server utility of the second cloud server 112 is calculated.

According to an embodiment, the importance of the data can be quantified as the relative importance of the data is delayed sensitivity.

Depending on the embodiment, the inbound response contribution may be determined based on the history that the cloud servers 11 have responded to in the past for the request of the second cloud server 112. [

According to an embodiment, the inbound response contribution may be determined based on the importance of the data previously requested by the second cloud server 112 to the cloud servers 11 that have previously responded to the request of the second cloud server 112 . ≪ / RTI >

In this embodiment, the inbound response contribution means that the second cloud server 112 has received a response to a request for important data from the partner cloud servers 11 in the past.

Also, in accordance with the embodiment, the inbound response contribution may be provided to the cloud servers 11 that have previously responded to a packet request in the past by the second cloud server 112, ) May be determined according to the importance of the data requested.

In these embodiments, it is meaningful that the cloud server 11, which is the provider, calculates the server utility by considering the importance of the data to be currently responded and the importance of the data received in the past.

In accordance with an embodiment, the drift rate measure is a measure of the computational resources in which the second cloud server 112 is running.

In accordance with an embodiment, the drift rate metric can be quantified such that the second cloud server 112 is relatively advantageous to use computational resources energy efficient or speed efficiently.

According to an embodiment, the drift rate measure is largest when the actuation ratio is a predetermined moderate level, relatively low when it is less than the predetermined moderate level, and may be negatively given if it exceeds a predetermined moderate level.

Depending on the embodiment, the server utility for a particular combination of packet requests may be computed, for example, as shown in equation (4), by summing the weighted sum of the importance of the data, the inbound response contribution and the drift rate metric.

In step S64, a combination of packet requests that maximize the server utility of the second cloud server 112 is determined. In step S65, the cloud servers 11 corresponding to the determined combination packet requests Lt; RTI ID = 0.0 > requests. ≪ / RTI >

7 and 8 are simulation results according to a cloud network resource management method according to an embodiment of the present invention. Fig. 7 shows the results of the simulation when the congestion degree is normal, Fig. 8 shows the response Are graphs that compare the number of repeated data requests until there is.

First, the simulation environment is a cloud network system 10 in which 100 data, in which data importance is first given as a natural number from 1 to 10, is distributed and stored in 100 cloud servers 11, respectively. Each of the cloud servers 11 has only 40% of packets of some data and the remaining 60% packets are distributed to other cloud servers 11. [ Therefore, a certain cloud server 11 must request packets to a plurality of other cloud servers 11 to integrate the complete data.

It is assumed that each cloud server 11 has a CPU drive rate of 0% to 100% and a response rate of 10% of a CPU drive rate in response to one packet.

In addition, the drive ratio measure is 11 when the drive ratio is 50% to 70%, 5.5 when it is 0 to 50%, -5 when it is 70% to 80%, -10, 90% %, It is specified in advance to have a value of -15.

Referring to FIG. 7, in order to simulate a congestion condition, each cloud server repeats a request until a response is received, when it is set to transmit 20 data packet requests at a time to other cloud servers These graphs are based on the importance of data.

The results of the cloud network resource management algorithm based on the server utility of the present invention represented by a rhomboid pointer obtained the data of high importance with fewer than two requests while the data of low importance was obtained by repeating the request more than four times .

In contrast, existing cloud network resource management algorithms based on tit-for-tat, represented by hollow triangle pointers, which do not take into account server utility, or cloud network resource management algorithms based on random request-based cloud network resources In the case of the management algorithm, all the data were obtained with the number of requests repeated slightly more than twice regardless of the importance.

Meanwhile, referring to FIG. 8, in order to simulate a congestion situation, when each cloud server is set to transmit 30 data packet requests at a time to other cloud servers, the request is repeated until a response is received These graphs are based on the importance of data.

Results of cloud network resource management algorithms based on the server utility of the present invention, represented by rhomboid pointers, obtained high importance data in less than two requests despite the increased network congestion, Was obtained by repeatedly requesting more than 6 times, which was somewhat higher.

On the other hand, in the cloud network resource management algorithm based on the existing 팃-format based on the empty triangle pointer, which does not consider the server utility, or the cloud network resource management algorithm based on the random request indicated by the X-pointer, All the data, regardless of the number of requests, was obtained in three repetitive times, which is slower than in Fig.

FIG. 9 is a graph illustrating simulation results of a cloud network resource management method according to an exemplary embodiment of the present invention and comparing CPU drive ratios of the cloud server when the congestion is normal.

Referring to FIG. 9, in order to simulate a situation where the congestion is normal, when each cloud server is set to transmit 20 data packet requests at a time to other cloud servers, a cloud based on server effectiveness indicated by rhombic pointers The results of the network resource management algorithm show that the balance between speed and energy is maintained while the CPU drive rate remains strictly below 70%.

On the other hand, existing cloud-based cloud network resource management algorithms, represented by hollow triangle pointers, which do not take into account the server utility, showed more than 75% CPU share, and random- In the case of the network resource management algorithm, the CPU share was maintained at about 70%.

In this simulation, each cloud server performs only requests and responses, and does not consider the load due to the cloud service to be allocated. Therefore, in the case of conventional cloud network resource management algorithms, if each cloud server performs only a little additional work There is a possibility of entering the overload state, and the service quality as a whole may be greatly deteriorated.

10 is a block diagram illustrating a cloud server implementing a cloud network resource management method according to an embodiment of the present invention.

10, the cloud server 100 includes an outbound response contribution calculating unit 101, a requesting party selecting unit 102, a server utility calculating unit 103, a response partner selecting unit 104, a request response interface 105, a request response history management unit 106, and a resource management unit 107.

The outbound response contribution calculation unit 101 calculates the outbound response contribution of the other cloud servers based on the history of the past responses to the past packet requests from each of the other cloud servers.

Depending on the embodiment, the outbound response contribution may be determined according to the importance of data corresponding to packet requests of other cloud servers that the cloud server 100 has responded to in the past.

According to the embodiment, the importance of the data can be quantified to be relatively important as the delay sensitivity of the data.

Also, according to the embodiment, the outbound response contribution may be determined according to the importance of data corresponding to a request from other cloud servers that the cloud server 100 has responded to immediately before the history of the past history.

The requesting party selection unit 102 can select a relative cloud server to which to make a packet request among the cloud servers based on the outbound response contribution of each of the other cloud servers.

When the at least one packet request is received, the server utility calculating unit 103 calculates the importance of data corresponding to each packet request, the history of each of the cloud servers that have made a packet request in response to the request of the cloud server 100 in the past The server utility of the cloud server 100 for each combination of packet requests is computed based on the inbound response contribution according to the inbound response contribution and the drift rate measure with respect to the operational ratio of the computational resources of the cloud server 100.

According to an embodiment, the importance of the data can be quantified as the relative importance of the data is delayed sensitivity.

Depending on the embodiment, the inbound response contribution may be determined based on the history that other cloud servers have responded to in the past to the requests of the cloud server 100. [

Depending on the embodiment, the inbound response contribution may be determined based on the importance of the data previously requested by the cloud server 100 to other cloud servers that have previously responded to the request of the cloud server 100 .

In addition, according to the embodiment, the inbound response contribution may be calculated based on the data of the cloud server 100 requested by the cloud server 100 at the immediately preceding time, among the past history, with respect to the other cloud servers that have received a response from the cloud server 100 immediately before Can be determined according to importance.

According to an embodiment, the drift rate measure is a measure of the computational resources in which the cloud server 100 is running.

In accordance with an embodiment, the drift rate metric can be quantified to be relatively advantageous as the cloud server 100 uses the computational resources energy efficient or speed efficiently.

According to an embodiment, the drift rate measure is largest when the actuation ratio is a predetermined moderate level, relatively low when it is less than the predetermined moderate level, and may be negatively given if it exceeds a predetermined moderate level.

Depending on the embodiment, the server utility for a particular combination of packet requests may be computed, for example, as shown in equation (4), by summing the weighted sum of the importance of the data, the inbound response contribution and the drift rate metric.

Then, the response partner selecting unit 104 determines a combination of packet requests that maximize the server utility of the cloud server 100, and selects counterpart cloud servers corresponding to the determined packet requests.

The request response interface 105 transmits a request signal of a required data packet to the selected requesting partner cloud servers at the time of packet request outbound, Lt; / RTI >

The request response interface 105 may also receive packet requests inbound from other cloud servers, or may receive response packets inbound.

The request response history management unit 106 records the importance of the packet requested data and the identification information of the requesting party cloud server at the time of requesting the packet, and at the time of packet response, the importance of the data corresponding to the received packet and the identification information of the response partner cloud server So that the request response history can be managed.

The resource management unit 107 can detect and manage the usage amount of operation resources in the cloud server 100. [

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, Modification is possible. Accordingly, the spirit of the present invention should be understood only in accordance with the following claims, and all of the equivalent or equivalent variations will fall within the scope of the present invention.

Further, the apparatus according to the present invention can be implemented as a computer-readable code on a computer-readable recording medium. A computer-readable recording medium includes all kinds of recording apparatuses in which data that can be read by a computer system is stored. Examples of the recording medium include a ROM, a RAM, an optical disk, a magnetic tape, a floppy disk, a hard disk, a nonvolatile memory, and the like, and a carrier wave (for example, transmission via the Internet). The computer-readable recording medium may also be distributed over a networked computer system so that computer readable code can be stored and executed in a distributed manner.

10 Cloud Network Systems
11, 111, 112, 100 Cloud servers
101 outbound response contribution calculation unit 102 request destination selection unit
103 Server utility calculation unit 104 Response partner selection unit
105 Request response interface 106 Request response history management section
107 Resource Management Department
12 cloud core domain
20 Network
30 user terminal

Claims (25)

CLAIMS 1. A resource management method of a cloud network in which data is distributedly stored in a plurality of cloud servers,
Wherein the cloud server receiving the at least one packet request receives the at least one packet from the at least one cloud server based on the importance of the data corresponding to each packet request and inbound according to the history in which each of the other cloud servers that have made a packet request to the cloud server responded to the packet request of the cloud server in the past Computing a server utility of the cloud server for each combination of packet requests based on a response rate and a drift rate metric on the operational rate of operational resources of the cloud server;
Determining a combination of packet requests that maximizes a server utility of the cloud server; And
And responding to the packet requests by sending the required packets corresponding to the determined combination of packet requests to the requesting cloud servers, respectively. The method of claim 1, wherein the importance of the data is expressed as relative importance of data corresponding to a packet request as delay sensitivity. 3. The method of claim 2, wherein the inbound response contribution is determined based on the history of responses to other packet requests by the cloud server in the past. 4. The method of claim 3, wherein the inbound response contribution is determined based on the importance of data corresponding to a packet request in the past to other cloud servers that have previously responded to a packet request by the cloud server Wherein the cloud network resource management method comprises the steps of: [Claim 4] The method of claim 4, wherein the inbound response contribution is determined based on the importance of data requested by the cloud server at a point in time immediately before the cloud server has responded to the packet request in the past Wherein the cloud network resource management method comprises the steps of: The method of claim 1, wherein the drift rate metric is a measure of the computational resources in which the cloud server is running, wherein the cloud server is quantified to be relatively advantageous as the computational resources are used energy- Resource management method. 7. The method of claim 6, wherein the drive rate measure
When the operation ratio is at a predetermined appropriate level, it is positively applied to the server utility,
If it is smaller than the predetermined proper level, it is positively applied to the server utility,
And if the predetermined level is exceeded, a negative value is applied to the server utility. The method of claim 1, wherein the server utility for a particular combination of packet requests comprises a weighted sum of the importance, inbound response contribution, and drift rate metrics of each packet request for each packet request, And calculating the cloud network resources by integrating the cloud network resources. The method as claimed in claim 8, wherein the server utility is expressed by the following equation

Lt; / RTI >
Here, the server utility

Lt; RTI ID = 0.0 >

Of the server utility,

Lt; RTI ID = 0.0 > server < / RTI >

, And the data importance

Lt; RTI ID = 0.0 >

For each of the requested data

of

Th packet

, And the inbound response contribution

The cloud server

The cloud server

Cloud server based on request to

Inbound response contribution,

The cloud server

Of the driving rate,

,

,

The importance of each data

, Inbound response contribution

And drive ratio measure

Wherein the weights are weights of the cloud network resources. The method according to claim 1,
When the cloud server lacks some of the data related to the service provided by the cloud server, the cloud server selects at least one other cloud server to make a packet request based on the outbound response contribution for each of the other cloud servers ; And
Further comprising the step of the cloud server requesting a packet necessary for the selected cloud server. 11. The method of claim 10, wherein the outbound response contribution is determined based on the history of the cloud server responding to each packet request of the other cloud servers in the past. 12. The method of claim 11, wherein the outbound response contribution is determined by the importance of data corresponding to each packet request of the other cloud servers to which the cloud server has responded in the past,
Wherein the importance of the data is quantified to be relatively important as the delay sensitivity of the data increases. 13. The method of claim 12, wherein the outbound response contribution is determined according to the importance of data corresponding to a packet request of other cloud servers that the cloud server has previously responded to. CLAIMS 1. A resource management method of a cloud network in which data is distributedly stored in a plurality of cloud servers,
In the event that some of the data related to the service provided by the first cloud server is insufficient, the first cloud server may send a second cloud server to make a packet request based on the outbound response contribution for each of the other cloud servers Selecting step;
The first cloud server requesting a packet necessary for the selected second cloud server;
The second cloud server receiving the at least one packet request transmits the importance of the data corresponding to each packet request and the importance of the data to the second cloud server in response to the request of the second cloud server in the past Computing a server utility value of the second cloud server for each combination of packet requests based on the inbound response contribution according to the history of the first cloud server and the driving rate measure regarding the operational rate of the operational resource of the second cloud server;
Determining a combination of packet requests that maximizes a server utility of the second cloud server; And
And responding to the packet requests by sending each of the required packets for the cloud servers corresponding to the determined combination of packet requests. A computer-readable recording medium storing a program for implementing a cloud network resource management method according to any one of claims 1 to 14 in a computer device. As a cloud server in which data is distributed in a cloud network,
An importance of data corresponding to each packet request, an inbound response according to the history in which each of the other cloud servers that have made a packet request to the cloud server responded to the packet request of the cloud server in the past when receiving at least one packet request, A server utility calculating unit for calculating a server utility of the cloud server with respect to each combination of packet requests based on a contribution rate and a drive rate measure related to an operation ratio of operation resources of the cloud server;
A response partner selection unit for determining a combination of packet requests that maximize the server utility; And
And a request response interface that responds to packet requests by sending the required packets corresponding to the determined combination of packet requests to the requesting cloud servers, respectively. 17. The cloud server of claim 16, wherein the importance of the data is expressed as relative importance of data corresponding to the packet request as delay sensitivity. 18. The method of claim 17,
Wherein the inbound response contribution is determined based on the importance of data corresponding to a packet request made in the past to other cloud servers that have previously responded to a packet request by the cloud server in the past Cloud server. 17. The method of claim 16, wherein the drift rate metric is a measure of computational resources running on the cloud server, wherein the cloud server is numerically quantified to be relatively advantageous as computing resources are used energy- . 17. The method of claim 16, wherein the server utility for a particular combination of packet requests is a weighted sum of the importance, inbound response contribution, and drive rate metrics of the data corresponding to each combination of packet requests, And a cloud server for computing the cloud server. The server according to claim 20, wherein the server utility is expressed by the following equation

Lt; / RTI >
Here, the server utility

Lt; RTI ID = 0.0 >

Of the server utility,

Lt; RTI ID = 0.0 > server < / RTI >

, And the data importance

Lt; RTI ID = 0.0 >

For each of the requested data

of

Th packet

, And the inbound response contribution

The cloud server

The cloud server

Cloud server based on request to

Inbound response contribution,

The cloud server

Of the driving rate,

,

,

The importance of each data

, Inbound response contribution

And drive ratio measure

Wherein the weighting factors are weighting factors of the server. 18. The method of claim 16,
A requesting party selection unit for selecting at least one other cloud server to request a packet based on the outbound response contribution for each of the other cloud servers when a portion of the data related to the service provided by the cloud server is insufficient Further included,
Wherein the request response interface is operable to request a packet required for the selected cloud server. The method according to claim 16, further comprising the steps of: recording the importance of the packet requested data and the identification information of the cloud server selected as the requesting party at the time of requesting the packet; Further comprising a request response history management unit for managing the request response history by recording the identification information of the server. A computer-readable recording medium having recorded thereon a computer device capable of operating as a cloud server belonging to a cloud network according to any one of claims 16 to 23. A cloud network system for distributedly storing data in a plurality of cloud servers,
Wherein each of the plurality of cloud servers comprises:
An importance of data corresponding to each packet request, an inbound response according to the history in which each of the other cloud servers that have made a packet request to the cloud server responded to the packet request of the cloud server in the past when receiving at least one packet request, A server utility calculating unit for calculating a server utility of the cloud server with respect to each combination of packet requests based on a contribution rate and a drive rate measure related to an operation ratio of operation resources of the cloud server;
A response partner selection unit for determining a combination of packet requests that maximize the server utility; And
And a request response interface that responds to packet requests by sending the required packets corresponding to the determined combination of packet requests to the requesting cloud servers, respectively.

Images