KR20190025285A - System and method for allocating resource statistically and providing QoS guarantee for Inter-Data Center WAN - Google Patents

Abstract

Disclosed are a system for statistical resource allocation and quality of service (QoS) guarantee, and a method thereof. The present invention provides a solution that can guarantee bandwidth in a statistical way in a wide area network environment between data centers, while achieving high network resource utilization and requested QoS. According to the present invention, high network resource utilization can be achieved while ensuring a requested QoS level by statistically ensuring a service request with the QoS level of less than 100%.

Description

[0001] The present invention relates to a statistical resource allocation and service quality assurance system in a wide-area inter-data center network,

The present invention relates to a statistical resource allocation and service quality assurance system and a method thereof in a wide-area inter-data center network, and more particularly, to a system and method for securing a bandwidth in a data center wide area network environment by statistical methods, And a system and method for providing a solution that can achieve the requested quality of service.

Most service providers (SPs), such as Gamevil and Netflix, are building and managing their own data centers (DCs) for their businesses. However, in order to expand the infrastructure, additional costs such as purchasing additional hardware and software are consumed. Accordingly, the majority of service providers (SPs) are migrating their service infrastructure to data centers (DCs) managed by a cloud service provider (CSP) to reduce costs. Cloud service providers (CSPs) are deployed across multiple global data centers to provide global coverage and inter-datacenter wide-area networks (Inter-DC WANs) with data centers (DCs) DC). Inter-DC WANs between data centers are being shared by multiple services that generate large amounts of traffic.

Conventional aggregate traffic management uses a fixed bandwidth allocation method shown in FIG. 1A or a multiplexing method shown in FIG. 1B.

The fixed bandwidth allocation method allocates a fixed bandwidth to the user to guarantee the quality of service (QoS) requested by the user. However, such a method of allocating network resources is not a good solution for a service provider (SP) or a cloud service provider (CSP). Service providers (SPs) are difficult to determine how much bandwidth they need because the service requests are complex and change over time. Even if they already know the statistics of the traffic patterns, it is still difficult to specify their needs at the determined values. Furthermore, as shown in the graph of FIG. 1 (a), allocating the fixed bandwidth has a problem that the network resources can not be efficiently used.

The multiplexing method is not a fixed bandwidth allocation for a user's service request but a method for improving network utilization by mixing all the traffic. However, the multiplexing method has a problem that it is difficult to control each traffic so as not to exceed the requested demand.

Korean Patent Laid-Open No. 2004-0076822 (Huawei Technology Company) Sep. 3, 2004 Patent Document 1 discloses a method of providing a service with guaranteed service quality in an IP access network, and Patent Document 1 discloses a network control layer layer is requesting network resources to the access network in response to a service request that requires guaranteed QoS, the edge router determines whether the access network can provide sufficient resources for the service And when the resources are sufficient, the edge routers transmit QoS parameters to access network end devices that control the QoS to obtain guaranteed QoS.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a data center wide area communication network that provides a solution that can guarantee a bandwidth in a statistical manner in a wide area network environment between data centers while achieving high network resource utilization and requested quality of service And to provide a method for guaranteeing quality of service and a method for the same.

According to an aspect of the present invention, there is provided a central control apparatus for a statistical resource allocation and service quality assurance system in a wide-area inter-data center wide area network system, comprising: service request information including data flow information and service level information, ; A path obtaining unit that obtains a data flow path corresponding to the data flow information based on the service level information upon receiving the service request information through the receiver; A routing update unit updating the routing information based on the data flow path obtained through the path obtaining unit and providing the updated routing information to a plurality of data centers; And a resource allocation manager for providing the service request information to the source data server so that a source data server based on the data flow information generates a data packet flow based on the service level information, Tags the priority of the packet according to the bandwidth based on the service level information.

The service level information is information indicating a service quality level of the data flow, and may include an average and a standard deviation of a peak rate of a flow along a normal distribution, and a quality of service (QoS) level.

The source data server may tag the priority of one of high priority and low priority in a packet according to a service quality level based on the service level information.

The source data server may tag the packet with a higher priority if it is currently using a bandwidth lower than the bandwidth based on the quality of service level and if the bandwidth is currently higher than the bandwidth based on the quality of service level, It can be tagged.

The relay device included in the data center has a high priority queue, a low priority queue, and a first-in first out (FIFO) queue. If the priority of a packet received from the outside is high Inserting the received packet into the high priority queue, inserting the received packet into the low priority queue if the received packet has a low priority, and transmitting the packet in the high priority queue to the first priority queue And if there is no packet in the high priority queue, the packet in the low priority queue may be delivered to the first-in pre-processing queue.

The relay device may drop the packet in the low priority queue if the first-in-first-out queue is full of packets.

The source data server may allocate a token according to the bandwidth based on the service level information and generate a data packet flow based on the service level information by consuming the corresponding token when generating a data packet flow.

According to another aspect of the present invention, there is provided a statistical resource allocation and service quality assurance method in a wide area network between a data center, Receiving service request information from the user terminal; Obtaining a data flow path corresponding to the data flow information based on the service level information; Updating the routing information based on the obtained data flow path, and providing the updated routing information to a plurality of data centers; And providing the service request information to the source data server such that the source data server based on the data flow information generates a data packet flow based on the service level information, Tag priority on packets according to bandwidth based on service level information.

The service level information is information indicating a service quality level of the data flow, and may include an average and a standard deviation of a peak rate of a flow along a normal distribution, and a quality of service (QoS) level.

The source data server may tag the priority of one of high priority and low priority in a packet according to a service quality level based on the service level information.

The source data server may tag the packet with a higher priority if it is currently using a bandwidth lower than the bandwidth based on the quality of service level and if the bandwidth is currently higher than the bandwidth based on the quality of service level, It can be tagged.

The relay device included in the data center has a high priority queue, a low priority queue, and a first-in first out (FIFO) queue. If the priority of a packet received from the outside is high Inserting the received packet into the high priority queue, inserting the received packet into the low priority queue if the received packet has a low priority, and transmitting the packet in the high priority queue to the first priority queue And if there is no packet in the high priority queue, the packet in the low priority queue may be delivered to the first-in pre-processing queue.

The relay device may drop the packet in the low priority queue if the first-in-first-out queue is full of packets.

The source data server may allocate a token according to the bandwidth based on the service level information and generate a data packet flow based on the service level information by consuming the token when generating a data packet flow.

According to an aspect of the present invention, there is provided a computer program for use in a computer readable recording medium, the computer program causing the computer to execute any one of the methods.

According to the statistical resource allocation and service quality assurance system and method therefor in the inter-data center wide area network, statistically ensuring service requests with a service quality (QoS) level of 100% or less, Level network resource utilization can be achieved.

1 is a diagram for explaining an example of a traffic management method according to a conventional service request.
2 is a block diagram for explaining a statistical resource allocation and service quality assurance system in a inter-data-center wide area network according to a preferred embodiment of the present invention.
3 is a block diagram illustrating the configuration of the statistical resource allocation and service quality assurance system shown in FIG.
FIG. 4 is a view for explaining an example of the configuration of the statistical resource allocation and service quality assurance system shown in FIG.
5 is a block diagram for explaining the configuration of the data center shown in FIG.
Fig. 6 is a diagram for explaining an example of the configuration of the data center shown in Fig. 5. Fig.
7 is a block diagram for explaining the configuration of the central control unit shown in Fig.
8 is a view for explaining service request information according to a preferred embodiment of the present invention.
9 is a diagram for explaining an admission control according to a preferred embodiment of the present invention.
FIG. 10 is a view for explaining a resource allocation management operation according to a preferred embodiment of the present invention.
11 is a view for explaining an example of statistical resource allocation and service quality assurance operations according to a preferred embodiment of the present invention.
FIG. 12 is a view for explaining an example of an operation of tagging and processing priority to a packet according to a preferred embodiment of the present invention.
13 is a view for explaining an example of an operation of processing a tagged packet having a priority according to a preferred embodiment of the present invention.
FIG. 14 is a diagram for explaining a loss ratio according to priority tagging according to a preferred embodiment of the present invention.
FIG. 15 is a flowchart illustrating a statistical resource allocation and service quality assurance method in a wide-area inter-data communication network according to a preferred embodiment of the present invention.
16 is a graph for comparing statistical resource allocation and service quality assurance methods according to a preferred embodiment of the present invention.
17 is a view for explaining an example of an emulation environment for testing performance of statistical resource allocation and service quality assurance operations according to a preferred embodiment of the present invention.
18 is a view for explaining an example of a testbed environment for testing performance of statistical resource allocation and service quality assurance operations according to a preferred embodiment of the present invention.
FIG. 19 is a diagram for explaining the number of acceptance of a service request as a result of an experiment in the emulation environment shown in FIG.
FIG. 20 is a diagram for explaining link utilization, which is an experiment result in the emulation environment shown in FIG.
FIG. 21 is a diagram for explaining a loss guarantee flow as a result of an experiment in the emulation environment shown in FIG.
22 is a diagram for explaining the number of service requests accepted as a result of an experiment in the test bed environment shown in FIG.
FIG. 23 is a diagram for explaining link utilization, which is an experiment result in the test bed environment shown in FIG.
FIG. 24 is a diagram for explaining a loss guarantee flow as a result of an experiment in the test bed environment shown in FIG. 18; FIG.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, preferred embodiments of a statistical resource allocation and service quality assurance system and a method thereof in a wide-area inter-data center wide area network according to the present invention will be described in detail with reference to the accompanying drawings.

First, a statistical resource allocation and service quality assurance system in a inter-data center wide area network according to a preferred embodiment of the present invention will be described with reference to FIG.

2 is a block diagram for explaining a statistical resource allocation and service quality assurance system in a inter-data-center wide area network according to a preferred embodiment of the present invention.

Referring to FIG. 2, a statistical resource allocation and service quality assurance system (hereinafter referred to as 'statistical resource allocation and service quality assurance system') 100 in a inter-data center wide area network And is connected to the terminal 200.

The statistical resource allocation and service quality assurance system 100 is a system operated by a cloud service provider (CSP), which provides a solution capable of guaranteeing bandwidth and achieving high network resource utilization in a data center wide area network environment .

That is, the statistical resource allocation and service quality assurance system 100 receives service request information from the user terminal 200 through the communication network 300. Here, the service request information includes data flow information and service level information. The data flow information is information about the source and destination of the data flow requesting the service. The service level information is information indicating the level of quality of service (QoS) of the data flow.

The statistical resource allocation and service quality assurance system 100 obtains a data flow path corresponding to the service request information using statistical network abstraction (SNA). Here, the data flow path refers to a path through which data is transmitted between a source and a destination of a data flow according to service request information.

In addition, the statistical resource allocation and service quality assurance system 100 provides service request information to the source data server so that the source data server generates a data packet flow according to the service request information using statistical network virtualization (SNA) do. At this time, the source data server tags the priority of the packet according to the bandwidth based on the service level information.

The user terminal 200 is a device operated by a service provider SP. The user terminal 200 transmits service request information including information on the level of a desired service quality (QoS) to the user terminal 200 through a communication network 300, To the system 100.

The user terminal 200 may be a desktop computer, a notebook computer, a workstation, a palmtop computer, an Ultra Mobile Personal Computer (UMPC), a tablet PC, a personal digital assistant (PDA) A smart phone, a mobile phone, or the like, and a terminal equipped with a microprocessor and having computing capability.

The communication network 300 includes a data network as well as a telephone network including a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN) It does not matter whether it is wired or wireless, and it does not matter what communication method you use.

3 to 6, the statistical resource allocation and service quality assurance system in the inter-data center wide area network according to the preferred embodiment of the present invention will be described in detail.

FIG. 3 is a block diagram for explaining the configuration of the statistical resource allocation and service quality assurance system shown in FIG. 2. FIG. 4 is a block diagram for explaining an example of the configuration of the statistical resource allocation and service quality assurance system shown in FIG. FIG.

3, the statistical resource allocation and service quality assurance system 100 may include a plurality of data centers 130 coupled to a central control unit 110 and a central control unit 110.

The central control unit 110 manages a plurality of data centers 130 and maintains global network information. That is, when receiving the service request information from the user terminal 200, the central control unit 110 determines whether to accept the service request information based on the current network. Then, the central control unit 110 obtains a data flow path corresponding to the service request information using statistical network abstraction (SNA). The central control unit 110 also provides service request information to the source data center 130 so that the source data server uses statistical network virtualization (SNA) to generate a data packet flow according to the service request information.

Referring to an example of the configuration of the statistical resource allocation and service quality assurance system shown in Fig. 4, the central control unit 110 is connected to five data centers 130-1 to 130-5, 130-1 to 130-5. Each of the five data centers 130-1 to 130-5 can be connected to each other via a link.

FIG. 5 is a block diagram for explaining the configuration of the data center shown in FIG. 3, and FIG. 6 is a diagram for explaining an example of the configuration of the data center shown in FIG.

5, the data center 130 includes a local control device 131, a plurality of relay devices 133 connected to the local control device 131, and a plurality of data servers 135 connected to the relay device 133 .

The local control device 131 is connected to a plurality of data servers 135 via a plurality of relay devices 133 to manage a plurality of data servers 135. [ That is, the local control unit 131 updates the routing information of the flow table according to the command of the central control unit 110. The local control unit 131 provides the service request information provided from the central control unit 110 to the data server 135 in the data center 130. In addition, the local controller 131 monitors link information such as available bandwidth, link failure, and the like, and provides the monitored link information to the central control unit 110.

Referring to an example of the configuration of the data center shown in Fig. 6, the first data center 130-1 includes a local control device 131, three relay devices 133-1 to 133-3, (135-1 to 135-6).

The local control device 131 is connected to the first relay device 133-1 and the second relay device 133-2. The local control device 131 is connected to the first data server 135-1, the second data server 135-2, and the third data server 135-3.

7 to 10, a central control device according to a preferred embodiment of the present invention will be described in more detail.

7 is a block diagram for explaining the configuration of the central control unit shown in Fig.

7, the central control unit 110 may include a receiving unit 110, a path obtaining unit 113, a routing updating unit 115, and a resource allocation managing unit 117.

The receiving unit 110 receives service request information from the user terminal 200. Here, the service request information includes data flow information and service level information. Service level information refers to information indicating the level of quality of service (QoS) of the data flow. For example, the service level information may include an average and standard deviation of the peak rates of flows along a normal distribution, a quality of service (QoS) level, and so on .

8 is a view for explaining service request information according to a preferred embodiment of the present invention.

More specifically, the service request R is composed of n flows, as shown in FIG. 8, and is defined as R = {f ₁ , ..., f _n }. Each flow f _m is defined as f _m = {src _m , dst _m , μ _m , σ _m , ε _m }. Where src _m represents the source dataset (DC) of the flow and dst _m represents the destination datacenter (DC) of the flow. _m denotes the average of the peak rates of the flow along the normal distribution and σ _m denotes the peak rate of the flow along the normal distribution, rates of standard deviation. epsilon _m is a violation probability of quality of service (QoS), which represents an epsilon value. That is, the epsilon is a quality of service (QoS) factor indicating the proportion of the amount of loss packets in the entire packet transmitted. For example, when epsilon _m is 0.03, the service provider (SP) may be provided with a guaranteed service in which the percentage of loss packets in the entire packet is 3 percent. Thus, 1-epsilon _m indicates the possibility of guaranteeing the required bandwidth. If the user wishes to construct a rectangular topology consisting of DC ₁ , DC ₂ , DC ₄ and DC ₅ as shown in FIG. 8A, A service request consisting of different flows can be made. Then, the central control unit 110 can obtain the data flow path corresponding to the service request information as shown in FIG. 8 (b).

7, when receiving the service request information through the receiving unit 110, the path obtaining unit 113 obtains a data flow path corresponding to the data flow information based on the service level information.

That is, the path obtaining unit 113 may obtain the data flow path corresponding to the data flow information while satisfying the service level information based on the bandwidth of the link between the data centers 130 and the service information pre-allocated to the link. Here, the link refers to a link that is currently held. Of course, the link may include all the links that are currently held and newly purchased.

At this time, the path obtaining unit 1130 can obtain the data flow path corresponding to the data flow information using the k-shortest path algorithm. Here, k is the number of links included in the path.

More specifically, admission control according to the present invention acquires a data flow path corresponding to service request information based on a currently held link.

That is, a service request R = {f ₁ , f ₂ , ..., f _n } is given. Where f _i is a normal distributed random variable N _i = N (μ _i , σ _i ² ). The capacity of link l is defined as c _l , and the weight of link _l is defined as w _l . Here, the weight may be set to a value associated with a cost such as a price or a delay that the cloud service provider (CSP) is interested in. F _l ^old refers to a set of already allocated flows with a path containing link l. and ε _l ^min is the minimum service quality (QoS) violation probability in the flow of link 1. Then, the admission control according to the present invention the notation _{"I Pm, l = 1:} 1 if l ∈ Pm, and 0 otherwise" using, the k-shortest path for each flow i P _i = (p _1, p ₂ , ..., p _k ). Finally, the admission control according to the present invention obtains a data flow path optimized for the service request information through the following [Table 1].

As can be seen in [Table 1] above, the objective function (5.1) is to minimize the average weight of all flows in the network. In other words, the optimization problem is to output a minimum weighted path for each flow of accepted service requests. The constraint of (5.3) indicates that the sum of the amounts of loss requested by each user is equal to the amount of loss of flows. The loss requested by each user can be computed in (5.2). Because of the statistical multiplexing gain, the sum of the bandwidth required to provide the requested service is less than the bandwidth of the flow service. In this way the loss rate requested by the user is guaranteed.

이고, 이는 적분 형태이다. 그래서 (5.3)의 항은 비선형이된다. 따라서, 이 비선형 제약을 선형화할 필요가 있다. 비선형이라면, 문제를 해결하는데 많은 시간이 소요될 수 있다. 이는 시스템의 확장성과 자원 효율성을 저하시킬 수 있다. 그래서, 비선형 제약을 선형화하는 것이 중요하다.However, the constraint of (5.3) has a non-linear term. The flow rate on a particular link is

, Which is an integral form. Thus, the term in (5.3) becomes nonlinear. Therefore, it is necessary to linearize this nonlinear constraint. If it is nonlinear, it can take a lot of time to solve the problem. This can degrade system scalability and resource efficiency. Thus, it is important to linearize nonlinear constraints.

First, by using the normalization of the normal distribution, Eq. (5.2) can be expressed by the Q function and the phi function of the standard normal distribution as shown below (5.5). And these functions can be easily linearized by the Taylor series as in (5.6) below.

가 있다.

는 k-shortest path 중에서 단지 하나의 가능한 경로(path)를 의미한다. 따라서,

은 미리 모든 가능한 k-shortest path를 반영하는 것을 의미한다. 따라서,

에 의해, 원래의 제약은 충분한 제약이 된다. 그래서, 원래의 문제를 위배하지 않는다. 이러한 프로세스는 아래의 [표 3]과 같다.However, in the normalization process of the normal distribution, the denominator of A in (5.5)

.

Means only one possible path in the k-shortest path. therefore,

Means to reflect all possible k-shortest paths in advance. therefore,

, The original constraint is a sufficient constraint. So, it does not violate the original problem. This process is shown in [Table 3] below.

In this way, the constraint (5.3) can be linearized.

9 is a diagram for explaining an admission control according to a preferred embodiment of the present invention.

In the Inter-DC WAN between data centers, most multiplexing schemes are trying to optimize the traffic received based on the current network resources. As shown in FIG. 9 (b), the cloud service provider (CSP) normally allocates four links when receiving a service request from DC ₁ to DC ₇ (see FIG. 9 (a)). However, this path assignment may be harmful to accepting the next service request. For this reason, provisioning bandwidth between DC ₁ and DC _{4 and} between DC ₄ and DC ₇ , as shown in Figure 9 (c), may be efficient in accepting service requests.

In summary, the path obtaining unit 113 can obtain the data flow path corresponding to the service request information through the following Table 4. [Table 4]

Step 1: Sort service requests in descending order according to total demands

Step 2: If there are no remaining service requests, stop

Step 3: Select the service request with the greatest total network resource demands

Step 4: Obtain optimized data flow path according to Admission Control or Network Provisioning according to the present invention

Step 5: When you are finished, go to Step 2

Referring back to FIG. 7, the routing update unit 115 updates the routing information based on the data flow path obtained through the path obtaining unit 113. Here, the routing information can be delivered in compliance with the OpenFlow Protocol. For example, a service request consists of a plurality of flows, each of which consists of a source and a destination. More specifically, it includes a source server (SRC_IP), a source port (SRC_PORT), a destination server (DSP_IP), and a destination port (DSP_PORT). Each of the local control devices 131 has a table for confirming the departure server, the source port, the destination server, the destination port, and the route to the destination server and port. For example, it has an ordered pair table like (SRC_IP, SRC_PORT, DST_IP, DST_PORT, output_port). This information is called routing rules, or routing information. According to the OpenFlow Protocol, the first part (SRC_IP, SRC_PORT, DST_IP, DST_PORT) is called the matching part, and the rear part (output_port) is called the action part.

Then, the routing update unit 115 provides the updated routing information to the plurality of data centers 130. At this time, the routing update unit 115 may provide the updated routing information to the plurality of local control units 131 that manage the plurality of data centers 130, respectively.

The resource allocation management unit 117 provides the service request information to the source data server 135 so that the source data server 135 based on the data flow information generates the data packet flow based on the service level information.

At this time, the resource allocation management unit 117 may provide the service request information to the source data server 135 through the local control unit 131 managing the data center 130 including the source data server 135.

The origin data server 135 may allocate tokens according to the bandwidth based on the service level information and generate the data packet flow based on the service level information when consuming the corresponding token when generating the data packet flow.

The resource allocation management operation according to the present invention will now be described in more detail with reference to FIG.

FIG. 10 is a view for explaining a resource allocation management operation according to a preferred embodiment of the present invention.

The resource allocation management operation according to the present invention uses a statistical rate limiter such as a kind of token bucket.

The resource allocation management algorithm can be divided into two parts. The first part is to create tokens based on traffic information. The local control device 131 or the data server 135 constructs a histogram following the normal distribution N (mu, [sigma]) upon receipt of the flow rate information such as ([mu], [sigma] Create tokens. To this end, the maximum rate possible can be determined, for example, by μ + σ · F ^-1 (0.9999).

이다. 히스토그램의 i번째 인터벌 내의 토큰의 개수는

이다.The token width (token width) and the entire token can then be determined. The local control device 131 or the data server 135 divides the entire area of the value following the normal distribution into a series of determined values using the token width value. That is, the type of the token is

to be. The number of tokens in the i-th interval of the histogram is

to be.

In this way, the local control device 131 or the data server 135 has tokens conforming to the normal distribution N ([mu], [sigma]). This part is described in the upper part of Algorithm 1 according to [Table 5] below.

Referring to FIG. 10, when the data server 135 removes the corresponding token from all the tokens, if there is no corresponding token, the data server 135 removes the min token on the corresponding token. Also, if there is no token on the token, the data server 135 removes the max token under the token.

An example of statistical resource allocation and service quality assurance operations according to a preferred embodiment of the present invention will now be described with reference to FIG.

11 is a view for explaining an example of statistical resource allocation and service quality assurance operations according to a preferred embodiment of the present invention.

As shown in Fig. 11, the network topology consists of a central control unit (CC) and four data centers (DC1 to DC4). The first link L1 connects the first data center DC1 and the second data center DC2. The second link L2 connects the first data center DC1 and the third data center DC3. The third link L3 connects the second data center DC2 and the third data center DC3. The fourth link L4 connects the second data center DC2 and the fourth data center DC4. The fifth link L5 connects the third data center DC3 and the fourth data center DC4. It is assumed that the bandwidth of all the links L1 to L5 is 100 Mbps. It is assumed that the network topology shown in FIG. 11 is small in size, so that there is no local control device for managing each data center, and a central control unit (CC) manages each data center.

The statistical resource allocation and service quality assurance operation according to the preferred embodiment of the present invention will be described below under the network topology.

That is, when a new service request is received, the present invention checks whether there is an extra resource that can accept the service request. If the service request is accepted, the central control unit (CC) acquires the data flow path corresponding to the service request and updates the routing information based on the data flow path so that the service can be serviced . Then, the central control unit (CC) limits the rate of flow to use the link only for the corresponding request.

For example, if the service request is (1, 2, 5, 4, 0.01), then it is desired to send a data flow from the first data center DC1 to the second data center DC2, Means that you want to follow a normal distribution with a standard deviation of 4. And, the meaning of 0.01 means that 99% follow this normal distribution, and 1% does not follow, which indicates that loss can occur. In other words, 99% is guaranteed, and 1% is a possibility of QoS violation in the sense that it is ok to lose.

The sum of service requests already existing in the third link LC3 follows a normal distribution with an average of 90 and a variance of 60 and the minimum value of the quality of service (QoS) violation possibility factor of these service requests is assumed to be 0.15 do. At this time, the probability that the bandwidth of the third link LC3 exceeds 100 Mbps is 0.1, which is 0.15 or less, which is the possibility of QoS violation, and thus it can be seen that the service is normally performed.

At this time, it is assumed that a new request (3, 2, 10, 25, 0.05) comes in. Then, the k-shortest path should be calculated. Assuming that k is 1, there is only a path using the third link LC3. Upon receipt of the new service request, the sum of the flows in the third link (LC3) will follow a normal distribution with an average of 100 and a variance of 85, and the minimum value of the probability of a quality of service (QoS) violation will be changed from 0.15 to 0.05. However, in this case, the probability that the bandwidth of the third link LC3 exceeds 100 Mbps exceeds 0.05, and the service request is rejected.

On the other hand, suppose that a new request (3, 2, 2, 16, 0.14) comes in. Then, the k-shortest path should be calculated. Assuming that k is 1, there is only a path using the third link LC3. When the new service request is received, the sum of the flows flowing through the third link (LC3) follows a normal distribution with an average of 92 and a variance of 76, and the minimum value of the probability of a quality of service (QoS) violation is changed from 0.15 to 0.14. In this case, the probability of the bandwidth of the third link LC3 exceeding 100 Mbps is 13% even if a flow according to a new request is added, and the service request is accepted because it is less than the minimum value of the possibility of QoS violation.

If the new request is accepted, then the traffic generated in the third data center DC3 will have to be forwarded to the second data center DC2 via the third link LC3, thus updating the routing information and updating the updated routing information To the data center.

Then, the third data center DC3 generates a data packet according to a normal distribution having an average of 2 and a variance of 16 according to a new request. To do this, the token bucket method is used to transmit the corresponding data packet while consuming the token.

In addition, the origin data server 135 tags the packet priority according to the bandwidth based on the service level information.

That is, the origin data server 135 may tag the priority of one of the high priority and the low priority in the packet according to the service quality level based on the service level information.

In more detail, the origin data server 135 tags the packet with a higher priority if it is currently using a bandwidth lower than the bandwidth based on the quality of service level, and if the bandwidth is higher than the bandwidth based on the quality of service level, A low priority can be tagged.

As such, the origin data server 135 may tag the priority to a packet and generate a data packet flow based on the service level information.

In addition, the repeater 133 included in the data center 130 has a high priority queue, a low priority queue, and a first in first out (FIFO) queue.

If the priority of the packet received from the outside is high, the relay apparatus 133 inserts the received packet into the high priority queue. If the priority of the received packet is low, Can be inserted into the rank queue.

In addition, the relay apparatus 133 forwards the packet in the high priority queue to the first-inning preprocessing queue before the packet in the low priority queue, and the packet in the low priority queue if there is no packet in the high priority queue It can be passed to the first-in-first-out queue.

At this time, the relay apparatus 133 can drop the packet in the low priority queue before the packet in the high priority queue if the packet is full in the first-in pre-processing queue.

12 to 14, an operation of tagging and processing the priority of a packet according to a preferred embodiment of the present invention will be described.

FIG. 12 is a view for explaining an example of an operation of tagging and processing priority to a packet according to a preferred embodiment of the present invention.

More specifically, admission control and rate enforcement according to the present invention are performed, but this is not sufficient to provide quality of service (QoS) to each service provider (SP). This is because loss may occur in the network regardless of the quality of service level, i.e., the quality of service (QoS) violation probability (epsilon value). A loss rather than an epsilon value is related to the instantaneous rate of the same time for each user. Thus, there is a need to make losses in the epsilon ratio.

If such a mechanism is implemented only in the network switch, that is, the relay device 133, the network switch becomes more complicated, which is not practical. Thus, it is a solution that the end host, i. E. The origin data server 135, and the network switch, or relay device 133, cooperate for the epsilon differentiation service.

For an epsilon-loss differentiating service with low network switch complexity, the end host, i. E., Source data server 135, tags the packets with low priority according to the epsilon ratios. This means that packets corresponding to the epsilon ratios among the entire transmitted packets are tagged with low priority and the remaining packets are tagged with high priority.

In the network switch, that is, the relay device 133, low priority tagged packets are dropped first because low priority tagged packets are processed since high priority tagged packets are processed. To this end, the network switch, or relay device 133, requires two priority queues for loss services corresponding to the epsilon value.

13 is a view for explaining an example of an operation of processing a tagged packet having a priority according to a preferred embodiment of the present invention.

Referring to FIG. 13, the first host (Host 1) tags low priority to three packets out of the ten packets transmitted outward according to the epsilon value of 0.3, and tags high priority to seven packets.

Then, the second host (Host 2) tags low priority to five packets out of the ten packets transmitted outward according to the epsilon value of 0.5 and high-priority tags to five packets.

Thereafter, the network switch that receives the packets from the first host (Host 1) and the second host (Host 2) inserts the packets with the high priority tagged into the high priority queue and the packets with the low priority tagged with the low And inserts it into the priority queue. The network switch then forwards the packets in the high priority queue to the first-in-first-out queue before the packets in the low-priority queue and for the packets in the low-priority queue if there are no packets in the high- To the queue. At this time, the network switch drops the packet in the low priority queue before the packet in the high priority queue if the packet is full in the first-in first-out processing queue.

FIG. 14 is a diagram for explaining a loss ratio according to priority tagging according to a preferred embodiment of the present invention. In the graph of FIG. 14, the x-axis is the flow id and the y-axis is the amount of loss.

Referring to FIG. 14, it can be seen that the actual loss rate is not proportional to the epsilon value when the priority of the packet is not tagged, and thus the quality of service (QoS) is not guaranteed (left graph in FIG. 14).

On the other hand, if the priority of the packet is tagged, it can be confirmed that the actual loss ratio is proportional to the epsilon value (the right graph in FIG. 14).

A method of allocating statistical resources and ensuring service quality in a wide-area inter-data center network according to a preferred embodiment of the present invention will be described with reference to FIG.

FIG. 15 is a flowchart illustrating a statistical resource allocation and service quality assurance method in a wide-area inter-data communication network according to a preferred embodiment of the present invention.

Referring to FIG. 15, the central control unit 110 receives service request information including data flow information and service level information (S110). Here, the data flow information refers to information about a source and a destination of a data flow requesting a service. The service level information is information indicating the level of quality of service (QoS) of the data flow. For example, the service level information may include an average and standard deviation of the peak rates of flows along a normal distribution, a quality of service (QoS) level, and so on .

Then, the central control unit 110 obtains a data flow path corresponding to the data flow information based on the service level information (S120). Here, the data flow path refers to a path through which data is transmitted between a source and a destination of a data flow according to service request information.

That is, the central control unit 110 can obtain the data flow path corresponding to the data flow information while satisfying the service level information based on the bandwidth of the link between the data centers 130 and the service information pre-allocated to the link. Here, the link refers to a link that is currently held. Of course, the link may include all the links that are currently held and newly purchased.

At this time, the central control unit 110 can acquire the data flow path corresponding to the data flow information using the k-shortest path algorithm. Here, k is the number of links included in the path.

The central control unit 110 then updates the routing information based on the obtained data flow path and provides the updated routing information to the plurality of data centers 130 (S130). At this time, the central control unit 110 may provide the updated routing information to the plurality of local control units 131 that manage the plurality of data centers 130, respectively.

Then, the central control device 110 provides the service request information to the source data server 135 so that the source data server 135 based on the data flow information generates the data packet flow based on the service level information (S140 ). At this time, the central control unit 110 may provide the service request information to the source data server 135 through the local control unit 131 managing the data center 130 including the source data server 135.

Then, the source data server 135 tags the priority of the packet according to the bandwidth based on the service level information (S150). That is, the origin data server 135 may tag one of the high priority and the low priority in the packet according to the service quality level based on the service level information. In more detail, the origin data server 135 tags the packet with a higher priority if it is currently using a bandwidth lower than the bandwidth based on the quality of service level, and if the bandwidth is higher than the bandwidth based on the quality of service level, A low priority can be tagged. In addition, the relay apparatus 133 included in the data center 130 has a high priority queue, a low priority queue, and a first-come pre-processing queue. If the priority of the packet received from the outside is high, the relay apparatus 133 inserts the received packet into the high priority queue. If the priority of the received packet is low, Can be inserted into the rank queue. In addition, the relay apparatus 133 forwards the packet in the high priority queue to the first-inning preprocessing queue before the packet in the low priority queue, and the packet in the low priority queue if there is no packet in the high priority queue It can be passed to the first-in-first-out queue. At this time, the relay apparatus 133 can drop the packet in the low priority queue before the packet in the high priority queue if the packet is full in the first-in first-out processing queue.

Then, the source data server 135 generates a data packet flow according to the service request information (S160). That is, the source data server 135 may allocate tokens according to the bandwidth based on the service level information, and may generate the data packet flow based on the service level information by consuming the corresponding token when generating the data packet flow.

Performance of statistical resource allocation and service quality assurance operations according to a preferred embodiment of the present invention will now be described with reference to FIGS.

16 is a graph for comparing statistical resource allocation and service quality assurance methods according to a preferred embodiment of the present invention.

In order to evaluate the performance of the statistical resource allocation and service quality assurance operation according to the present invention, the service request D _s = <μ, σ, ε> according to the present invention and the fixed bandwidth allocation method (hereinafter referred to as "fixed bandwidth allocation method" referred to as the fixing method "), a service request D _d = μ + σ · F determined according to ^- and ^{1 (1-ε)} is assumed to be the same, where, F is the standard normal distribution N ~ <0, 1 ^2> cumulative distribution of Function (CDF). This is because the present invention and the fixed bandwidth allocation method have the same quality of service (QoS) level as shown in Fig.

Emulation Settings

17 is a view for explaining an example of an emulation environment for testing performance of statistical resource allocation and service quality assurance operations according to a preferred embodiment of the present invention.

Emulates the Google B4 inter-DC WAN as shown in Figure 17 using Mininet. The Google B4 inter-DC WAN has 12 data centers (DCs) and 19 Inter-DC links per city. OVS switch and ONOS (Open Network Operation System) are used as Central Controller. Each link capacity is 200 Mbps, and 3-shortest paths are considered to accept the service request. 400 service requests are generated, and the origin and destination of each service request are uniformly randomly determined. Each service request has only one flow, and the values of the mean and standard deviation of each flow are uniformly distributed in a given range. The results were obtained through 15 experiments. The emulation was done through a high-performance computer with quad-core Intel i5-6500 3.2 Ghz, 32 GB RAM and a 256 GB SSD hard disk to get meaningful results.

Test bed  Set

18 is a view for explaining an example of a testbed environment for testing performance of statistical resource allocation and service quality assurance operations according to a preferred embodiment of the present invention.

The test bed includes four data centers (DC1 to DC4), four switches (SW), five inter-DC links , And a central control unit (CC). The central control unit (CC) consists of dual core Intel E8400 3.00 GHz, 8 GB memory and 300 GB hard disk. Each data center (DC1 to DC4) consists of five data servers, an Iptime H50008 8-port Gigabit Switching Hub, and an open virtual switch (OVS). Each OVS consists of 12 Intel Xeon X5690 3.47GHz CPUs, 20G RAM, 1TB hard disk, and five 1 Gigabit Ethernet NICS. All of the links in the inter-DC WAN are assigned the same weight and the same 300 Mbps capacity. The Iperf application is used to generate traffic.

Emulation experiment result

FIG. 19 is a view for explaining the number of acceptance of a service request as a result of an experiment in the emulation environment shown in FIG. 17, FIG. 20 is a view for explaining link utilization, which is an experiment result in the emulation environment shown in FIG. FIG. 21 is a diagram for explaining a loss guarantee flow which is an experiment result in the emulation environment shown in FIG.

In Figure 19 (a), the flow is generated to have a standard deviation of 5 to 6 and an epsilon of 0.01 to 0.04. Referring to FIG. 19 (a), the number of accepted service requests is about 20% larger than the fixed method according to the present invention. As the average increases, the number of accepted service requests decreases due to network bandwidth limitations.

In Figure 19 (b), the flow is generated to have an average of 10 to 25 and an epsilon of 0.01 to 0.04. Referring to FIG. 19 (b), the number of accepted service requests is about 25% larger than the fixed method according to the present invention.

In Figure 19 (c), the flow is generated to have an average of 105 to 30 and a standard deviation of 10 to 13. Referring to FIG. 19 (c), the number of accepted service requests is about 25% larger than the fixed method of the present invention.

Thus, the present invention can process more service requests than the fixed method.

In FIG. 20, the flow is generated to have an average of 10 to 30, a variance of 1 to 4, and an epsilon of 0.0001 to 0.02. Referring to FIG. 20, the present invention can use the link about 10% more efficiently than the fixed method.

Referring to FIG. 21, it can be seen that the present invention can guarantee the quality of service (QoS) of almost all flows.

Test bed  Experiment result

FIG. 22 is a view for explaining the number of acceptance of a service request as a result of an experiment in the testbed environment shown in FIG. 18, FIG. 23 is a view for explaining the link utilization, which is an experiment result in the testbed environment shown in FIG. FIG. 24 is a diagram for explaining a loss guarantee flow which is a result of an experiment in the test bed environment shown in FIG. 18; FIG.

Referring to FIGS. 22, 23 and 24, it can be seen that the test bed test results show similar trends to the emulation test results. That is, the present invention can handle more service requests, improve link efficiency, and guarantee quality of service (QoS) of almost all flows.

The present invention can also be embodied as computer-readable codes 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 is stored. Examples of the computer-readable recording medium include a ROM, a RAM, a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device. In addition, the computer-readable recording medium may be distributed to computer devices connected to a wired / wireless communication network, and a computer-readable code may be stored and executed in a distributed manner.

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, It will be understood by those skilled in the art that various changes may be made and equivalents may be substituted without departing from the scope of the appended claims.

100: statistical resource allocation and service quality assurance system,
110: central control device,
111: receiving unit, 113: path obtaining unit,
115: routing update unit, 117: resource allocation management unit,
130: data center, 131: local control device,
133: Relay device, 135: Data server

Claims (15)

As a central control device of statistical resource allocation and service quality assurance system in a wide-area inter-data center network,
A receiving unit for receiving service request information including data flow information and service level information from a user terminal;
A path obtaining unit that obtains a data flow path corresponding to the data flow information based on the service level information upon receiving the service request information through the receiver;
A routing update unit updating the routing information based on the data flow path obtained through the path obtaining unit and providing the updated routing information to a plurality of data centers; And
A resource allocation manager for providing the service request information to the source data server so that a source data server based on the data flow information generates a data packet flow based on the service level information;
/ RTI &gt;
Wherein the source data server tags a priority in a packet according to a bandwidth based on the service level information. The method of claim 1,
Wherein the service level information is information indicating a level of service quality of a data flow, the average and standard deviation of peak rates of flow along a normal distribution, and a Quality of Service (QoS) level. 3. The method of claim 2,
Wherein the source data server tags priority of one of a high priority and a low priority in a packet according to a service quality level based on the service level information. 4. The method of claim 3,
Wherein the source data server is configured to tag a packet with a higher priority if it is currently using a bandwidth lower than the bandwidth based on the quality of service level and to assign a packet with a lower priority if the bandwidth currently used is higher than the bandwidth based on the quality of service level The central control device. 5. The method of claim 4,
The relay device included in the data center includes a high priority queue, a low priority queue, and a first-in first out (FIFO) queue, If the priority of the received packet is low, inserts the received packet into the low priority queue, and inserts the packet in the high priority queue into the high priority queue, Queue, and if there is no packet in the high priority queue, forwards the packet in the low priority queue to the first-in first-out queue. The method of claim 5,
Wherein the relay device drops a packet in the low priority queue when the first pre-processing queue is full of packets. The method of claim 1,
Wherein the source data server allocates a token according to the bandwidth based on the service level information and generates a data packet flow based on the service level information when consuming the token when generating a data packet flow. A method for statistical resource allocation and quality of service of a central control device,
Receiving service request information including data flow information and service level information from a user terminal;
Obtaining a data flow path corresponding to the data flow information based on the service level information;
Updating the routing information based on the obtained data flow path, and providing the updated routing information to a plurality of data centers; And
Providing the service request information to the source data server such that a source data server based on the data flow information generates a data packet flow based on the service level information;
/ RTI &gt;
Wherein the source data server tags priorities in packets according to bandwidth based on the service level information. 9. The method of claim 8,
Wherein the service level information is information indicating a level of service quality of a data flow, the average of the peak rate of the flow along a normal distribution, the standard deviation, and the quality of service (QoS) Statistical resource allocation and quality assurance method in communication network. The method of claim 9,
Wherein the source data server is operable to statistically (at least partially) statistically in a data center inter-wide area network, tagging a priority of one of a high priority and a low priority in a packet according to a service quality level based on the service level information. Resource allocation and service quality assurance. 11. The method of claim 10,
Wherein the source data server is configured to tag a packet with a higher priority if it is currently using a bandwidth lower than the bandwidth based on the quality of service level and to assign a packet with a lower priority if the bandwidth currently used is higher than the bandwidth based on the quality of service level And statistical resource allocation and quality of service guarantee in inter-data center wide area network. 12. The method of claim 11,
The relay device included in the data center includes a high priority queue, a low priority queue, and a first-in first out (FIFO) queue, If the priority of the received packet is low, inserts the received packet into the low priority queue, and inserts the packet in the high priority queue into the high priority queue, Queue in a high priority queue, and forwards the packet in the low priority queue to the first-in first-out queue if there is no packet in the high priority queue. The method of claim 12,
Wherein the relay device drops a packet in the low priority queue when the first-in-first-out queue is full of packets, in the inter-data center wide area network. 9. The method of claim 8,
Wherein the source data server allocates a token according to the bandwidth based on the service level information and generates a data packet flow based on the service level information by consuming the token when generating a data packet flow, Statistical resource allocation and quality assurance method in communication network. A computer program stored in a computer readable recording medium for causing a computer to execute a method of statistical resource allocation and service quality assurance in a inter-data center wide area network as claimed in any one of claims 8 to 14.

Images