JP2013168139A - Load balancing device, load balancing method and hierarchized data center system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To use load traffic transmitted to a current data center corresponding to a load balancing device in a plurality of data centers, which is used in a hierarchical data center system configured by connecting a plurality of data centers by a layering method. Provided are a load balancing device, a load balancing method, and a hierarchical data center system that perform allocation to a server.
A load balancer allocates load traffic transmitted to a current data center based on resource usage states of the current data center and a lower data center. According to the present invention, the diversity and flexibility of the configuration of the hierarchical data center system is improved, and resources are idle as close to the user as possible according to the resource usage state of each data center constituting the hierarchical data center system. The user's request is processed by the data center in the state, the load in the hierarchical data center system is further balanced, the response speed of the user request is improved, and thereby the service efficiency of the hierarchical data center system can be improved. .
[Selection] Figure 1

Description

  The present invention relates to a data center load balancing technique, and more particularly, a load balancing device used for load balancing between sub-networks in a hierarchical data center system, a load balancing method, and a hierarchical data center system using the load balancing device and the load balancing method. About.

  Hierarchical data center systems are now widely used in network applications, especially in the networking of goods that are growing day by day. In order to improve the service efficiency of such a hierarchical system, load balancing between the sub-networks is necessary.

  In Patent Document 1, aggregator is used to provide a set of branch facilities and load balance in a manner that reduces the hierarchy format and branch (ie, site or data center in the later description) facility layout, and between aggregators A technique has been proposed in which facility information of different branch offices is identified by interactive processing so that resource requests can balance the entire load across all branch offices and branch facilities.

  Patent Document 2 proposes a technique for selecting a service providing site based on a user's IP address.

US2008 / 0034111A1 US7,523,170B1

  However, the above prior art has the following problems. First, when collecting facility information of each branch by connecting to each branch via an aggregator, the diversity and flexibility of the network hierarchy configuration is limited. In addition, as a site selection method, a newly arrived connection request is processed based only on the existing connection and session state without considering the distinction between service types. Even if there is a possibility that the service types are different in such a processing method, the load balance processing obtained by them is the same. For example, a service that generates a large amount of network traffic is assigned to a site away from the user, and thereby network resources are occupied, the network delay increases, and the service providing efficiency of the overall hierarchical configuration decreases.

  In order to solve the above technical problem, the present invention improves the diversity and flexibility of the configuration of the hierarchical data center system, and improves the service efficiency of the hierarchical data center system. An object is to provide a balancing method and a hierarchical data center system.

  The present invention also provides a load balancing device, a load balancing method, and hierarchical data that can distinguish the service type of the service request and improve the service efficiency of the hierarchical data center system in order to solve the above technical problem. The purpose is to provide a center system.

  In order to achieve the above object, a load balancing apparatus according to the present invention is used in a hierarchical data center system configured by connecting a plurality of data centers by a hierarchical method, and the load in the plurality of data centers is used. A load balancing device for allocating load traffic transmitted to a current data center corresponding to a balancing device, the current data center, and a lower layer connected to the current data center and positioned below the current data center The load traffic transmitted to the current data center is allocated based on the resource usage state with the data center.

  The load balancing method according to the present invention for achieving the above object is used in a hierarchical data center system configured by connecting a plurality of data centers by a hierarchical method, and present data in the plurality of data centers. A load balancing method for allocating load traffic transmitted to a center, the resource usage state between the current data center and a lower data center connected to the current data center and located below the current data center And a allocating step of allocating load traffic transmitted to the current data center based on a determination result of the determining step.

  In order to achieve the above object, a hierarchical data center system according to the present invention is configured by connecting a plurality of data centers by a hierarchical method, and includes hierarchical data including a plurality of load balancing devices respectively corresponding to the data centers. Each of the load balancers is based on the resource usage status of a current data center corresponding to the load balancer and a lower layer data center connected to the current data center and located below the current data center. Then, the load traffic transmitted to the current data center is allocated.

  According to the load balancing device, the load balancing method, and the hierarchical data center system of the present invention, the diversity and flexibility of the configuration of the hierarchical data center system is improved, and the resources of each data center constituting the hierarchical data center system Depending on usage, user requests are handled by data centers that are as close to the user as possible and resources are idle, which balances the load in the tiered data center system and improves the response speed of user requests, thereby The service efficiency of the data center system can be improved.

  In the load balancing apparatus described above, based on a link state between the current data center and the upper data center connected to the lower data center and the current data center and located in the upper layer of the current data center, Allocation may be performed for load traffic currently transmitted to the data center.

  In the load balancing method described above, in the determination step, a link between the current data center and the upper data center connected to the lower data center and the current data center and located on the upper layer of the current data center is further provided. The state may be determined.

  As a result, data can be transferred via a link in an idle state, the pressure on the network link of the transferred service can be relieved, and the service efficiency of the hierarchical data center system can be further improved.

  In the above-described load balancing device, the load traffic transmitted to the current data center may be assigned based on the protocol type of the packet constituting the load traffic.

  In the load balancing method, in the determination step, a protocol type of a packet constituting the load traffic may be further determined.

  This ensures service quality more reliably, reduces the pressure on the network link of the forwarded service, prevents greater response delays on the forwarded service due to network congestion, thereby providing services for hierarchical data center systems Efficiency can be further improved.

  In the load balancing device described above, the protocol type of the packet constituting the load traffic may be determined using the header information of the network layer.

  In the load balancing apparatus, in the determination step, the protocol type of a packet constituting the load traffic may be determined using network layer header information.

  Thereby, the processing speed of the load balancer can be accelerated using the header information of the network layer.

  In the load balancing apparatus described above, when the resource usage state of the current data center is equal to or less than a predetermined first threshold, the load balancing apparatus may allocate the load traffic transmitted to the current data center to the current data center. Well, if the resource usage status of the current data center is greater than the first threshold and less than or equal to a predetermined second threshold greater than the first threshold, the load balancer The current data center based on a link state between the data center and the upper data center connected to the current data center and located in the upper layer of the current data center and / or a protocol type of a packet constituting load traffic Allocation for load traffic transmitted to If the resource usage state of the current data center is greater than a predetermined second threshold, the load balancer can determine whether the load transmitted to the current data center is based on the resource usage state of the lower layer data center. Traffic may be assigned to the lower layer data center or the upper layer data center.

  In the load balancing device described above, when it is determined in the determination step that the resource usage state of the current data center is equal to or less than a predetermined first threshold, the allocation step determines whether the load traffic transmitted to the current data center is currently If it is determined in the determination step that the resource usage state of the current data center is less than a predetermined second threshold greater than the first threshold and greater than the first threshold, The allocating step constitutes a link state and / or load traffic between the current data center and the lower data center and the upper data center connected to the current data center and located on the upper layer of the current data center. Based on packet protocol type The load traffic transmitted to the current data center may be allocated. If it is determined in the determination step that the resource usage state of the current data center is greater than a predetermined second threshold, the allocation The step may allocate the load traffic transmitted to the current data center to the lower data center or the upper data center based on the resource usage state of the lower data center.

  This allows you to optimize load traffic allocation to better match data center performance by using multiple thresholds to quantize data center performance, which makes load traffic more flexible In addition, the service efficiency of the hierarchical data center system can be further improved by appropriately allocating.

  In the load balancing apparatus described above, a communication interface that transmits and receives a packet including at least load traffic and a status message, a transfer unit that changes header information including at least a source address and a destination address of the packet, and resources of the current data center A storage device storing a use state and a resource use state of the lower layer data center acquired by the state message, a resource use state stored in the storage device, and header information of a packet received by the communication interface And a processing unit for controlling allocation of packets constituting the load traffic of the transfer unit.

  Thereby, the load balancing device of the present invention can be realized with a simple configuration.

  According to the load balancing apparatus, load balancing method, and hierarchical data center system according to the present invention, a newly arrived service request is distinguished, and different assignments are made to services of different protocol types based on the current site state and network state. Adopting a mechanism, a site close to the user responds to a service request that requires more network resources, and a site far from the user responds to a service request that requires less network resources. Reduced demand for network resources for communication between different sites. Further, in the present invention, network layer header information is mainly used to distinguish newly arrived service information, so the processing speed of the load balancer may reach the router linear speed. Thus, areas such as a large number of sensor report messages due to disaster, or a large number of report messages due to errors in the sensor node's sleep function, or simultaneous large clicks on the user's hotspot content, and thus malicious attacks from the network, etc. In the event of sudden and sudden heavy loads, the technical solution of the present invention is used to quickly distribute and process some loads to data centers in other areas, thereby overloading the system in those areas. The situation can be improved. Finally, the layered configuration of the system moves the end site closer to the user side and processes most of the services in the normal state leaving the local end site, thereby improving the response speed of the local service and improving the upper layer data. Performance requirements for the center load balancer are reduced.

It is a block diagram of the hierarchical data center system concerning 1st embodiment. It is a module figure of one specific Example of the load balancing apparatus concerning 1st embodiment. It is an example of the message format of a status message. It is an example of a resource use state. It is an example of load allocation mapping information. 3 is a flowchart of a load balancing method according to the first embodiment. It is a flowchart of one specific Example of the load balancing method concerning 1st embodiment. It is an example of a link state. There is an example of threshold information. It is a flowchart of one specific Example of the load balancing method concerning 4th embodiment. It is a time series figure of an example of data allocation processing. It is a time series figure of an example of data allocation processing. It is a time series figure of an example of data allocation processing. It is a time series figure of an example of data allocation processing. It is a time series figure of an example of data allocation processing. It is a time series figure of an example of data allocation processing.

[1. First Embodiment]

Hereinafter, the first embodiment of the present invention will be described in detail with reference to the drawings.
<1-1. Overall Configuration of Hierarchical Data Center System>

  The overall configuration of the hierarchical data center system according to this embodiment will be specifically described below with reference to FIG. FIG. 1 is a configuration diagram of a hierarchical data center system according to the first embodiment.

  In the present embodiment, the hierarchical data center system is composed of a plurality of data centers (that is, sites) connected by a hierarchical method. Of these, the connection here only needs to be able to communicate with the data center, and includes various communication connection methods such as a wired connection method and a wireless connection method. For example, in FIG. 1, the hierarchical data center system includes seven data centers 11 101, data center 21 102, data center 22 103, data center 31 104, data center 32 105, data center 33 106, and data center 34 107. It consists of a data center. These data centers are connected by a tree-type topology, of which the site located in the root direction (upward in FIG. 1) is called the parent site (upper data center), and the leaf direction (in FIG. 1) Sites located below) are called child sites (lower data centers), and each data center communicates only with its parent site and child sites. For example, data center 11 101 has child site data center 21 102 and data center 22 103, data center 21 102 has child sites data center 31 104 and data center 32 105, and data center 22. Reference numeral 103 denotes a data center 33 106 and a data center 34 107 which are child sites.

In the hierarchical data center system according to the present embodiment, the data centers are distributed at different physical locations and are connected to local users belonging to the data centers. For example, in FIG. 1, the data center 31 104 is connected to a sensor group 1 109 composed of sensor nodes 108 and a user group 1 111 composed of user terminals 110. At the same time, the data center 31 104 can further connect with other local users, and the other six sites in the figure can also connect with various users or user groups (not shown).
<1-2. Data Center>

  Hereinafter, the data center according to this embodiment will be described in detail with reference to FIG.

  In this embodiment, the data center can connect with a local user and transmit load traffic and the like with the local user. Further, the data center can be further connected to other data centers, and load traffic and status messages can be transmitted to and from other data centers. The data center includes a load balancer (balancer), a data center intranet, and a server group (local server group). Among them, the load balancer assigns load traffic transmitted to the data center. Specifically, for example, load traffic allocation is handled by the data center, or load traffic is transferred to an upper data center or a lower data center connected to the data center, and the data center intranet is allocated by a load balancer. The load traffic is transmitted to the service group of the data center, and the serve group processes the load traffic transmitted via the data center intranet.

For example, in FIG. 1, data center 31 104 provides sensor group 1 109 with a storage service that stores data collected by sensors, for example, and provides services such as Internet access, video on demand, and email to user group 1 111. To provide. These services are gathered to become load traffic 112, which is transmitted upstream from sensor group 1 109 and user group 1 111 to data center 31 104, or conversely, from data center 31 104 to sensor group 1 109 and user group 1 111. Or transmitted. The data center 31 104 allocates to the load traffic 112 through the balancing device 31 113 and distributes some or all of the load traffic to the server group 31 115 belonging to the data center (site) via the data center intranet 31 114. Then, load traffic that needs to be processed by other data centers (sites) is transferred to the connected upper layer data center (parent site, data center 21 102 in FIG. 1) or lower layer data center (child site). The remaining data centers have a configuration similar to data center 31 104, and for the sake of clarity, the associated load balancer of some data centers, namely balancer 11 116, balancer 21 117, balance, is shown in the figure. Only device 22 118 and balance device 32 119 are shown. In addition, each data center belonging to the same hierarchical data center system can communicate its own site status to the other party by exchanging the status message 120 with the parent site and the child site. The status message 120 will be described in detail later.
<1-3. Load balancing device>

Hereinafter, the load balancing apparatus according to the present embodiment will be described in detail.
<1-3-1. Configuration of load balancer>

  In this embodiment, the load balancer is used in a hierarchical data center system configured by connecting a plurality of data centers by a layering method, and the current data center corresponding to the load balancer in the plurality of data centers. Allocation is performed for load traffic transmitted to.

  Among them, the load balancer allocates the load traffic transmitted to the current data center based on the resource usage status of the current data center and the lower data center connected to the current data center and located in the lower layer of the current data center. Do.

Specifically, for example, if the current data center resource usage is idle, the load balancer assigns load traffic to the current data center, and if the current data center resource usage is not idle, and the lower layer If the data center resource usage is idle, the load balancer allocates load traffic to the lower data center, and the current data center resource usage is not idle, and the lower data center resource usage is idle. Otherwise, the load balancer allocates load traffic to the upper data center that is connected to the current data center and is located above the current data center.
<1-3-2. One Specific Example of Load Balancing Device>

  Hereinafter, one specific example of the load balancing apparatus according to the present embodiment will be described with reference to FIGS.

FIG. 2 is a module diagram of one specific example of the load balancing apparatus according to the first embodiment. As shown in FIG. 2, the load balancing device 21 117 according to this specific embodiment includes a communication interface 201, a transfer unit 202, a processing unit 203, and a storage device 204. A similar structure can be adopted for other load balancing devices.
<1-3-2-1. Communication interface>

  The communication interface 201 transmits and receives a packet including at least load traffic and a status message 120. The communication interface 201 may be, for example, a wired or wireless data communication interface.

  FIG. 3 shows an example of the message format of the status message. Specifically, FIG. 3 shows an example of the message format of the status message 120 sent by the communication interface and reporting the resource usage status to the upper data center. . As shown in FIG. 3, the status message 120 includes a site ID identifying the current data center (eg, a serial number indicating the current data center), and a host's IP address indicating the only public network IP address used outside the current data center. A virtual IP, a resource state indicating a current data center resource use state, and a service list indicating a service type that can be provided by the current data center are included.

Of these, the resource state in the status message 120 further includes a CPU idle percentage indicating the average CPU idle rate of the local server of the current data center, a storage idle percentage indicating the average storage idle rate of the local server of the current data center, and the current A request processing rate indicating an average request processing rate of a local server in the data center may be included.
<1-3-2-2. Transfer unit>

The transfer unit 202 changes the header information including at least the source address and the destination address of the packet. For example, the transfer unit 202 transfers the packet by changing the header information of the packet based on the load allocation mapping information (which will be described in detail later) stored in the storage device 204. The transfer unit 202 is, for example, a dedicated or general-purpose processor or an integrated circuit.
<1-3-2-2-3. Storage device>

  The storage device 204 stores the resource usage status of the current data center and the resource usage status of the lower data center acquired by the status message 120. The storage device 204 is a readable / writable storage device such as a RAM or an HDD.

  FIG. 4 is an example of a resource usage state. Specifically, FIG. 4 is an example of a site capacity table showing the resource usage state of the current data center and the lower data center stored in the storage device 204. As shown in FIG. 4, the site capacity table includes an index, a site ID for identifying the current data center or lower layer data center (for example, a serial number indicating the current data center or lower layer data center), the current data center or lower layer data center. Virtual IP of host indicating public network IP address to be used outside, CPU idle percentage indicating average CPU idle rate of local server of current data center or lower layer data center, average storage device of local server of current data center or lower layer data center Storage idle percentage indicating the idle rate, request processing speed indicating the average request processing speed of the local server in the current data center or lower data center, current data set Total load indicating the site total load terpolymers or lower data center, and the service list indicating services that can be provided by the type of the current data center or lower data centers are included.

  Among them, the total load of the data center may be, for example, an arithmetic average value represented by the following formula 1 of the CPU usage rate and the storage device usage rate of the local server of the data center.

Total load = AVG (CPU usage rate, storage device usage rate)
= 1-AVG (CPU idle percentage, storage idle percentage)
(In the equation, AVG (x, y) indicates that an arithmetic average value of x and y is taken, CPU usage rate = 1-CPU idle percentage, and storage device usage rate = 1-storage device idle percentage.)

FIG. 5 is an example of a load allocation mapping table. Specifically, FIG. 5 is a load allocation mapping table of services currently stored in the storage device 204 and accepted by the data center. The load allocation mapping table includes an index, a client terminal IP indicating the IP address of a user terminal located in a user group or a sensor node located in a center group or an agent thereof, and an intranet of a server located in a local server group in a current data center. Server terminal IP indicating the IP address or virtual IP address of the host of the transfer destination data center, next hop IP indicating the transfer destination IP address, next hop physical address indicating the data link layer address of the next hop of the path arriving at the transfer destination, Ingress port indicating the transmission layer port number when the packet currently arrives at the data center, egress port indicating the transmission layer port number when the packet is transferred, and the transmission layer protocol type of the packet Contains the protocol type indicated. Among them, the transmission layer port number may be a TCP port number or a UDP port number, and the protocol type may be TCP or UDP.
<1-3-2-2-4. Processing unit>

  The processing unit 203 controls the allocation of the packet constituting the load traffic of the transfer unit 202 based on the resource usage state stored in the storage device 204 and the header information of the packet received by the communication interface. The processing unit 203 is, for example, a general-purpose processor such as a CPU or MPU or a dedicated integrated circuit.

Specifically, for example, the processing unit 203 can read the total load (see FIG. 4) stored in the storage device 204, and the total load of the current data center is equal to or less than a predetermined threshold A. The transfer unit 202 controls to transfer the packet constituting the load traffic to the local server of the current data center, the total load of the current data center is greater than the predetermined threshold A, and the total load of the lower data center is predetermined When the threshold value B is equal to or less than the threshold value B, the transfer unit 202 controls to transfer the packet constituting the load traffic to the lower layer data center, and the total load of the current data center is larger than the predetermined threshold value A and If the total load is greater than the predetermined threshold B, the transfer unit 202 The packets constituting the Ikku control to transfer to the upper layer data center. Among them, the threshold A and the threshold B may be the same value or different values.
<1-4. Load balancing method>

Hereinafter, the load balancing method according to the present embodiment will be described in detail with reference to FIG.
<1-4-1. Process of load balancing method>

FIG. 6 is a flowchart of the load balancing method according to the first embodiment. As shown in FIG. 6, the load balancing method according to the present embodiment is used in a hierarchical data center system configured by connecting a plurality of data centers by a hierarchical method, and present data in the plurality of data centers. A load balancing method for allocating load traffic transmitted to a center, the resource usage state of the current data center and a lower level data center connected to the current data center and located below the current data center A determination step S1 for determining the resource usage state, and an allocation step S2 for allocating load traffic currently transmitted to the data center based on the determination result of the determination step S1.
<1-4-2. One specific implementation of load balancing method>

  Hereinafter, one specific example of the load balancing method according to the present embodiment will be described with reference to FIG. FIG. 7 is a flowchart of one specific example of the load balancing method according to the first embodiment. In the figure, the load balancing method according to this specific embodiment can be realized by the above-described specific embodiment of the load balancing device, and refers to the above-mentioned specific embodiment of the load balancing device.

  First, the communication interface 201 of the load balancer receives a packet constituting the load traffic and transmits this packet to the processing unit 203 (step S901).

  Next, the processing unit 203 checks the load allocation mapping table stored in the storage device 204 (step S902).

  Subsequently, the processing unit 203 determines whether there is a record in which the client terminal IP address or the server terminal IP address and the packet source address are the same (step S903).

  If the determination result in step S903 is YES, header processing and packet transfer are performed according to the matching record via the transfer unit 202 and the communication interface 201 (step S904).

  Conversely, if the determination result in step S903 is NO, the packet direction is further checked to determine whether it is an uplink packet (step S905).

  If the determination result in step S905 is NO, it means that the packet is a downstream packet from the server and not a request from the user side, and can be abandoned directly (step S906).

  Conversely, if the determination result in step S905 is YES, this means that the request is from a new user, and it is necessary to distribute it to the local server of the current data center or transfer it to another data center. Accordingly, the processing unit 203 checks information such as the current resource usage status of the current data center, for example, CPU idle percentage, storage device idle percentage, requested processing speed, etc. A comparison is made with a plurality of predefined thresholds to determine if the data center is currently idle. As a specific determination method, for example, the processing unit 203 checks the site capacity table stored in the storage device 204 and compares the total load of the local server of the current data center with a predetermined threshold as described above. It can be adopted (step S907).

  If the determination result in step S907 is YES, it means that the data center is currently idle and the request may be processed, and the processing unit 203 sends the packet to the local server via the transfer unit 202. Send. When there are a plurality of local servers, for example, the local server can be selected by selecting a server having the largest number of idle resources (step S908).

  Thereafter, the processing unit 203 updates the load allocation mapping table by writing the IP addresses of the client terminal and the server terminal, the next hop address and port information, etc. included in the request (step S909).

  If the determination result in step S907 is NO, it means that the current data center is busy, and the processing unit 203 selects one data center from the upper data center and the lower data center of the current data center. Need to be transferred. In order to determine the destination data center to be transferred, the processing unit 203 checks the site capacity table stored in the storage device 204, for example information such as CPU idle percentage, storage idle percentage, requested processing speed, etc. One or more terms in the information need to be compared with one or more predefined thresholds to determine if there is an idle lower level data center. As a specific determination method, for example, a method of comparing the total load of the local server in the lower layer data center with a predetermined threshold as described above can be employed (step S910).

  If the determination result in step S910 is NO, this means that there is no lower data center that is currently idle in the data center, and the current data center is busy, so the processing unit 203 sends a request to the upper data center. Transfer is selected (step S911).

  When the determination result of step S910 is YES, the processing unit 203 transmits the packet to the lower layer data center via the transfer unit 202. When there are a plurality of lower layer data centers, for example, one lower layer data center can be selected by selecting a lower layer data center having the largest number of idle resources (step S912).

  Similarly, after steps S911 and S912, the processing unit 203 needs to update the load allocation mapping table (step S909).

Among them, in the specific example of the load balancing method shown in FIG. 7, in step S901, load traffic is received, and in steps S902, S903, and S905, it is determined whether the received load traffic is a new request, step S907, S910 corresponds to determination step S1 in the load balancing method shown in FIG. 6, and steps S908, S911, and S912 correspond to allocation step S2 in the load balancing method shown in FIG.
<1-5. Effect of First Embodiment>

According to the load balancing device, the load balancing method, and the hierarchical data center system according to the present embodiment, the diversity and flexibility of the configuration of the hierarchical data center system is improved, and each data center constituting the hierarchical data center system is improved. Depending on the resource usage status, the user requests are handled by a data center that is as close to the user as possible and the resources are idle, further balancing the load in the hierarchical data center system and improving the response speed of user requests. As a result, the service efficiency of the hierarchical data center system can be improved.
[2. Second Embodiment]

Hereinafter, a second embodiment of the present invention will be specifically described with reference to the drawings.
<2-1. Features of Second Embodiment>

  This embodiment is based on the first embodiment, and allocates load traffic based on not only the resource usage state of the data center but also the link state between the data centers. Hereinafter, differences between the present embodiment and the first embodiment will be specifically described, and the description of the first embodiment will be referred to for the same points between the present embodiment and the first embodiment.

  The configuration of the hierarchical data center system and the data center according to this embodiment is the same as that of the first embodiment, and a description thereof is omitted here.

  The load balancer according to the present embodiment is further compared to the load balancer according to the first embodiment, based on the link state between the current data center, the lower layer data center, and the upper layer data center. Allocation is performed for load traffic currently transmitted to the data center.

Compared with the load balancing method according to the first embodiment, the load balancing method according to the present embodiment further determines the link state between the current data center, the lower data center, and the upper data center in the determination step, In the allocation step, the load traffic is allocated based on the determination result for the link state in the determination step.
<2-2. One specific example of the second embodiment>

  Hereinafter, one specific example of the load balancing apparatus and the load balancing method according to the present embodiment will be described.

  As one specific example of the load balancer according to this embodiment, the storage device 204 is further connected to the current data center based on the load balancer of the specific example of the first embodiment. The information on the link state with other data centers is stored. FIG. 8 shows an example of the link state. Specifically, FIG. 8 is a network link state table that records the link state from the current data center 21102 to the upper layer data center and the lower layer data center. The network link state table includes a link ID that identifies a connection between a current data center and an upper data center and a lower data center (for example, a network link serial number), an upper data center or a lower data center connected to the link (for example, A remote site ID identifying the data center serial number) and a link load indicating the bandwidth utilization of the link.

As one specific example of the load balancing method according to the present embodiment, based on the load balancing method of the specific example of the first embodiment, it is determined in step S907 that the current data center is not idle. In this case, the processing unit 203 further determines the current data center, upper layer data center, and lower layer based on the network link state table stored in the storage device 204 in the load balancer according to the specific example of the present embodiment. It is determined whether the link state is idle by comparing the link state with the data center with a predetermined threshold. When it is determined that the link state is the idle state, the process proceeds to step S910, and the resource usage state of the lower layer data center is determined. If it is determined that none of the link states are idle, the process proceeds to step S908, and the load traffic is transmitted to the local server of the current data center via the transfer unit 202.
<2-3. Effect of Second Embodiment>

According to the present embodiment, in the hierarchical data center system, the load balancer allocates not only the resource usage state but also the load traffic currently transmitted to the data center based on the link state. Therefore, according to the load balancing device, the load balancing method, and the hierarchical data center system according to the present embodiment, while having the effect of the first embodiment, at the same time, by transferring data via the link in the idle state, The pressure on the network link of the transferred service can be relieved and the service efficiency of the hierarchical data center system can be further improved.
[3. Third Embodiment]

The third embodiment of the present invention will be specifically described below.
<3-1. Features of Third Embodiment>

  This embodiment is based on the first embodiment, and allocates load traffic not only based on the resource usage state of the data center but also based on the protocol type of the packet constituting the load traffic. Hereinafter, differences between the present embodiment and the first embodiment will be specifically described, and the description of the first embodiment will be referred to for the same points between the present embodiment and the first embodiment.

  The configuration of the hierarchical data center system and the data center according to this embodiment is the same as that of the first embodiment, and a description thereof is omitted here.

  Compared with the load balancer according to the first embodiment, the load balancer according to the present embodiment further reduces the load traffic transmitted to the current data center based on the protocol type of the packet constituting the load traffic. Make an assignment.

Compared with the load balancing method according to the first embodiment, the load balancing method according to the present embodiment further determines the protocol type of the packet constituting the load traffic in the determination step, and in the allocation step, further in the determination step. The load traffic is allocated based on the determination result for the packet protocol type.
<3-2. One specific example of the third embodiment>

  Hereinafter, one specific example of the load balancing apparatus and the load balancing method according to the present embodiment will be described.

  Examples of the protocol types of the packets constituting the load traffic include TCP and UDP. Among them, the TCP protocol is generally insensitive to delay and used for services with a small amount of data, for example, web page browsing, file transmission, mail transmission, etc. The UDP protocol is generally sensitive to delay and the amount of data. It is used for transmission of large services such as real-time chat, video on demand, and Internet telephone. Therefore, processing of UDP protocol type services currently left in the data center further guarantees the quality of the service, reduces the pressure on the network link of the transferred service, and has a greater response delay on the transfer service due to network congestion Can be prevented.

  Based on the above rules, if it is determined in step S907 that the current data center is not idle based on the load balancing method of the specific example of the first embodiment, the processing unit 203 further receives in step S901. The protocol type indicated by the protocol field in the IP header of the received packet is checked. If the packet protocol type is TCP, the process advances to step S910 to determine the resource usage state of the lower layer data center. If the packet protocol type is UDP, the process proceeds to step S908, and the load traffic is transmitted to the local server in the current data center via the transfer unit 202.

As described above, since the header information of the network layer is used when determining the protocol type of the packet, there is a possibility that the processing speed of the load balancer is accelerated and reaches the linear speed of the router. Specifically, when the determination is performed using only the header information of the network layer, the processing speed of the load balancer is significantly improved as compared with the case where the determination is performed using the application layer information.
<3-3. Effect of Third Embodiment>

According to the present embodiment, in the hierarchical data center system, the load balancer allocates not only the resource usage state but also load traffic currently transmitted to the data center based on the protocol type of the packet constituting the load traffic. To do. Further, when determining the protocol type, network layer header information is used. Therefore, according to the load balancing device, the load balancing method, and the hierarchical data center system according to the present embodiment, the effect of the first embodiment is obtained, and at the same time, the quality of service is further ensured and the service to be transferred is transmitted. Reduce the pressure on the network link, prevent greater response delay on transfer service due to network congestion, accelerate the load balancer processing speed, and thereby further improve the service efficiency of hierarchical data center system .
[4. Fourth Embodiment]

The fourth embodiment of the present invention will be specifically described below with reference to the drawings.
<4-1. Features of Fourth Embodiment>

  This embodiment combines and improves the first, second, and third embodiments, and allocates load traffic based on the data center resource usage status, link status, and protocol type of the packets that make up the load traffic. . Hereinafter, differences between the present embodiment and the above-described embodiment will be specifically described, and the description of the above-described embodiment will be referred to for the same points between the present embodiment and the above-described embodiment.

  The configuration of the hierarchical data center system and the data center according to this embodiment is the same as that of the first embodiment, and a description thereof is omitted here.

  Compared with the load balancer according to the first embodiment, the load balancer according to the present embodiment further includes the link status between the current data center, the lower layer data center and the upper layer data center and the packet constituting the load traffic. Based on the protocol type, allocation is performed for load traffic transmitted to the current data center.

Compared with the load balancing method according to the first embodiment, the load balancing method according to the present embodiment further includes the link state and load traffic between the current data center, the lower data center, and the upper data center in the determination step. The protocol type of the packet to be configured is determined, and in the assignment step, the load traffic is assigned based on the determination result for the link state and the packet protocol type in the determination step.
<4-2. One specific example of the fourth embodiment>

  Hereinafter, one specific example of the load balancing apparatus and the load balancing method according to the present embodiment will be described.

  As one specific example of the load balancing device according to the present embodiment, threshold information is further stored in the storage device 204 based on the load balancing device of the specific example of the first embodiment. . FIG. 9 is an example of threshold information. Specifically, FIG. 9 is a predefined load threshold table in which the current data center stored in the storage device 204 accepts services. The predefined load threshold table includes a CPU usage rate threshold 11 and a threshold 12, a storage device usage rate threshold 21 and a threshold 22, and a total load threshold 1 and a threshold 2. Among them, the total load threshold value 1 and the threshold value 2 are arithmetic average values shown in the following equations of the corresponding CPU usage rate and storage device usage rate.

  Threshold m = AVG (threshold 1 m, threshold 2 m), m = {1, 2}

  By rationally selecting the value of the threshold nm (n, m = {1, 2}), for example, the CPU when the server average response time is increased twice as much as the threshold average 11 and the threshold 12, respectively, compared to the case of light load. Installed in the usage rate and the storage device usage rate, and the threshold value 21 and the threshold value 22 are respectively set in the CPU usage rate and the storage device usage rate when the server average response time is increased by 100 times compared to the light load. Threshold value 1 and threshold value 2 can be used to represent the start of deterioration of server average performance and the situation of serious deterioration.

  One specific example of the load balancing method according to this embodiment will be described with reference to FIG. FIG. 10 is a flowchart of one specific example of the load balancing method according to the fourth embodiment. In the figure, the load balancing method according to this specific embodiment can be realized by the above-described specific embodiment of the load balancing device, and refer to the related description of the above-mentioned specific embodiment of the load balancing device. .

  First, the communication interface 201 of the load balancer receives a packet constituting the load traffic, and transmits the received packet to the processing unit 203 (step S1001).

  Next, the processing unit 203 checks the load allocation mapping table stored in the storage device 204 (step S1002).

  Subsequently, the processing unit 203 determines whether there is a record in which the client terminal IP address or the server terminal IP address is the same as the packet source address (step S1003).

  If the determination result in step S1003 is YES, header processing and packet transfer are performed according to the matching record via the transfer unit 202 and the communication interface 201 (step S1004).

  Conversely, if the determination result in step S1003 is NO, the packet direction is further checked to determine whether the packet direction is an uplink packet (step S1005).

  Conversely, if the determination result in step S1005 is NO, it means that the packet is a downstream packet from the server and not a request from the user side, and can be abandoned directly (step S1006).

  On the other hand, if the determination result in step S1005 is YES, it means that the packet is a request from a new user, and it is necessary to distribute it to a local server in the current data center or another data center. Accordingly, the processing unit 203 checks information such as the current data center resource usage status, eg, CPU idle percentage, storage idle percentage, requested processing speed, etc., and sets one or more items in the above information as a set. Compare with a predefined threshold to determine if the data center is currently idle. As a specific determination method, for example, the processing unit 203 checks the site capacity table stored in the storage device 204 and a pre-defined load threshold table, and totals the total load of the local server of the current data center as described above. A method of comparing with the load thresholds 1 and 2 is adopted.

  At this time, if the determination result of step S1007 is that the total load is equal to or less than the threshold value 1, it means that the average performance of the local server of the data center is currently in a good state and the request may be processed. The processing unit 203 transmits the packet to the local server via the transfer unit 202. When there are a plurality of local servers, for example, one local server can be selected by selecting a server having the largest number of idle resources (step S1008).

  Thereafter, the processing unit 203 updates the load allocation mapping table by writing the IP addresses of the client terminal and server terminal, the next hop address and port information, the protocol type, etc. included in the request (step S1009).

  If the determination result of step S1007 is that the total load is larger than the total load threshold 2, it means that the average performance of the local server of the current data center is in a bad state, and the upper data center and lower data center of the current data center are in a bad state. It is necessary to select one data center from among them for transfer. To determine the destination data center to transfer, the processing unit 203 checks the site capacity table stored in the storage device 204, for example, information such as CPU idle percentage, storage idle percentage, requested processing speed, etc. One or more terms in the information need to be compared with one or more predefined thresholds to determine if there is an idle lower level data center. As a specific determination method, for example, as described above, a method of comparing the total load of the local server in the lower layer data center with a predetermined threshold may be employed (step S1010).

  If the determination result in step S1010 is NO, it means that there is no lower data center in the current data center and the current data center is busy, so the processing unit 203 forwards the request to the upper data center. (Step S1011).

  If the determination result of step S1010 is YES, the processing unit 203 transmits the packet to the lower layer data center via the transfer unit 203. When there are a plurality of lower layer data centers, for example, one lower layer data center can be selected by selecting a lower layer data center having the largest number of idle resources (step S1012).

  If the determination result of step S1007 is that the total load is greater than threshold 1 and less than or equal to threshold 2, it means that the average performance of the local server in the data center is currently in a general state between good and bad, It is necessary to appropriately reduce the load on the data center and transfer some newly increased load to the upper data center or lower data center of the current data center. At this time, whether or not the processing unit 203 checks the protocol type displayed by the protocol field in the network link state table stored in the storage device 204 and the IP header of the packet, and transfers the request represented by the packet It is necessary to judge. For the specific determination method, refer to the descriptions of the second and third embodiments (step S1013).

  If the determination result in step S1013 is that there is an idle link and the packet protocol type is TCP, the process proceeds to step S1010 to check whether there is an idle lower layer data center. Conversely, if the determination result in step S1013 is that there is no idle link and the packet protocol type is UDP, the process moves to step S1008, and the processing unit 203 transfers the packet to the local server via the transfer unit 202.

  Similarly, after steps S1008, S1011 and S1012, the processing unit 203 needs to update the load allocation mapping table 2 (step S1009).

Among these, in the specific example of the load balancing method shown in FIG. 10, in step S1001, load traffic is received, and in steps S1002, S1003, and S1005, it is determined whether the received load traffic is a new request, and step S1007, S1013 and S1010 correspond to the determination step S1 in the load balancing method shown in FIG. 6, and steps S1008, S1011 and S1012 correspond to the assigning step S2 in the load balancing method shown in FIG.
<4-3. Effect of Fourth Embodiment>

According to the present embodiment, in the hierarchical data center system, the load balancer not only uses the resource usage state, but also loads that are currently transmitted to the data center based on the link state and the protocol type of the packet that constitutes the load traffic. Assign to traffic. Therefore, this embodiment combines the effects of the first, second and third embodiments. Note that by quantizing the data center performance using multiple thresholds, load traffic allocation can be optimized to better match the data center performance, thereby making the load traffic more flexible and Appropriately allocated, the service efficiency of the hierarchical data center system can be further improved.
<5. Load assignment processing of the present invention>

  Hereinafter, based on the time-series diagrams of FIGS. 11 to 16, allocation processing in different situations for load traffic of the load balancer in the hierarchical data center system according to the present invention will be described. 11 and 12 show that the load balancer 31 113 distributes the load traffic 112 of the TCP and UDP types to the local server of the current data center 31 104, that is, the server group 31 115, so that the load traffic 112 is transmitted by the current data center 31 104. Indicates the situation to be processed. FIGS. 13 and 14 show a situation in which TCP and UDP type load traffic 112 is transferred to the parent site (upper layer data center), that is, the data center 21 102, and the load traffic 112 is processed by the parent site. 15 and 16 transfer TCP and UDP type load traffic 112 to the parent site or data center 21 102, which further forwards it to the other child site (lower data center) or data center 32 105, and finally A situation where load traffic 112 is processed by the other child sites will be described.

  The above-described load allocation can be realized by the load allocation apparatus and the load allocation method according to the first to fourth embodiments described above, and in particular, can be realized by the load allocation apparatus and the load allocation method according to the fourth embodiment.

  As shown in FIG. 11, first, a connection request is transmitted from the user group 1 111 (which may be one or a plurality of users or agents thereof) (301), and a SYN message 302 is sent to the current data center 31 104. Send. After the communication interface 201 of the balancing device 31 113 of the data center 31 104 receives the SYN message 302, the processing unit 203 is based on the information stored in the storage device 204 and the header information of the received SYN message 302. , Thereby processing the connection request by the local site (current data center) (303) and deciding to select one destination server from the local server group 31 115, for example, having the most idle resources Select one server. Thereafter, the processing unit 203 updates the load allocation mapping table by writing the IP addresses of the client terminal and the server terminal, the next hop address, port information, and the like included in the request. Subsequently, the forwarding unit 202 processes the SYN message 302 based on the updated load allocation mapping table and sets the destination IP address as the next hop IP address, ie, the IP address of the destination server selected by the processing unit 203 (usually the data center Private IP address), the destination physical address is placed at the next hop physical address, that is, the physical address of the switch immediately adjacent to the destination server, and the source IP address and the source physical address are balanced. 113 IP address and physical address. The processed SYN message 302 is transferred to the data center intranet 31 114 connected to the destination server via the communication interface 201 and further transferred to the destination server via the data center intranet 31 114. In the response message transmitted from the destination server, the destination IP address is set to the next hop IP address, that is, the client terminal IP address through processing in the opposite direction, and the destination physical address is set to the next hop physical address, that is, the path connected to the user group 1 111 The source IP address and the source physical address are set at the IP address and physical address of the balancing device 31 113. The processed response message SYN-ACK 304 is transmitted to the user group 1 111 via the communication interface 201. After receiving the response message, the user group 1 111 establishes a connection with the client terminal (305) and transmits the response message ACK 306 again. If the response message ACK 306 arrives, the balancing device 31 113 forwards to the destination server based on the existing record in the load allocation mapping table, thereby establishing a connection to the server terminal (307) and providing service to the user. (308) to allow the user to obtain the service (309). Among them, the upstream and downstream load traffic 310 is transferred based on the existing record in the load allocation mapping table in the same manner when passing through the balancing device 31 113.

  FIG. 12 is similar to FIG. 11 except that the protocol type of load traffic is UDP. As shown in FIG. 12, first, transmission of a packet is started by the user group 1 111 (which may be one or a plurality of users or their agents) (401), and the upstream traffic 402 is transferred to the data center 31 104. Send. Here, the center group 1 109 (which may be one or a plurality of sensor nodes or agents thereof) may transmit the packet. The upstream traffic 402 may be a video on demand request transmitted from a user terminal, sensor data reported from a sensor node, or data service traffic using another UDP protocol. After receiving the upstream traffic 402 by the communication interface 201 of the balancing device 31 113 of the data center 31 104, the processing unit 203 is based on the information stored in the storage device 204 and the received header information of the upstream traffic 402. , Thereby identifying that this is a new data flow, processing the connection request by the local site (403), and determining to select one destination server from the local server group 31 115; For example, one server having the most idle resources is selected. Thereafter, the processing unit 203 updates the load allocation mapping table by writing the IP addresses of the client terminal and the server terminal, the next hop address, port information, and the like included in the request. The subsequent transfer process is similar to that shown in FIG. 11, and the upstream traffic 402 processed by the transfer unit 202 is transmitted to the destination server. The destination server starts providing the service to the user (404), and the downlink traffic 407 transmitted to the user enables the user to acquire the service through processing in the opposite direction (406). Among them, the upstream and downstream load traffic 405 and 407 are transferred based on the existing records in the load allocation mapping table when passing through the balancing device 31 113.

  FIG. 13 is similar to FIG. 11 except that the parent site data center 21 102 of the data center 31 104 receives the request. As shown in FIG. 13, first, a connection request is transmitted by user group 1 111 (one or a plurality of users or agents thereof) (501), and a SYN message 502 is transmitted to data center 31 104. To do. After receiving the SYN message 502 by the communication interface 201 of the balancing device 31 113 of the data center 31 104, the processing unit 203 is based on the information stored in the storage device 204 and the header information of the received SYN message 502. , And notices that the request exceeds the processing capacity, and decides to process the connection request by the parent site or data center 21102 (503). Thereafter, the load allocation mapping table (see the table of FIG. 13) is updated by the processing unit 203, the SYN message 502 is processed based on the updated load allocation mapping table by the transfer unit 202, and processed by the communication interface 201. The SYN message 502 is transferred to the data center 21 102. The processing method performed by the balancing device 21 117 in response to the SYN message 502 in the data center 21 102 is similar to that in FIG. 11, and a server at the local site is selected to process the connection request (504). Thereafter, the data center 21 102 transmits a response message SYN-ACK 505 to the user, and is further transmitted to the user group 1 111 via the balance device 31 113, of which the balance device 31 113 is transmitted via the balance device 31 113. Modifies header address information based on existing records in the load allocation mapping table. After receiving the response message SYN-ACK 505, the user group 1 111 establishes a connection with the client terminal (506), and transmits the response message ACK 507 again. When the response message ACK 507 passes through the balancer 31 113, the balancer 31 113 modifies the header address information based on the existing record in the load allocation mapping table. Finally, the server terminal establishes a connection (508), starts providing services to the user (509), and allows the user to acquire the service (510). Among them, the upstream and downstream load traffic 514 is transferred based on the existing record in the load allocation mapping table in the same manner when passing through the balancing device 31 113.

  FIG. 14 is similar to FIG. 12 except that it is the parent site data center 21 102 of the data center 31 104 that receives requests. As shown in FIG. 14, first, transmission of a packet is started by the user group 1 111 (which may be one or a plurality of users or agents thereof) (601), and the upstream traffic 602 is transferred to the data center 31 104. Send. After receiving the upstream traffic 602 by the communication interface 201 of the balancing device 31 113 of the data center 31 104, the processing unit 203 performs the upstream traffic based on the information stored in the storage device 204 and the received header information of the S upstream traffic 602. Analyze 602 and notice that the request exceeds the processing capacity and decide to process the connection request by the parent site or data center 21102 (603). Thereafter, the load allocation mapping table is updated by the processing unit 203, the upstream traffic 602 is processed based on the updated load allocation mapping table by the transfer unit 202, and the processed upstream traffic 602 is processed by the communication interface 201 in the data center 21. 102. In the data center 21 102, the processing performed by the balancing device 21 117 for the upstream traffic 602 is similar to FIG. 12, which identifies this as a new data flow, selects a server at the local site and The connection request is processed (604), and the load allocation mapping table is updated. Thereafter, the data center 21 102 starts providing the service to the user (605) and enables the user to acquire the service (510). Among them, the upstream and downstream load traffic 608 and 607 are transferred based on the existing record in the load allocation mapping table in the same manner when passing through the balancing device 31 113.

  FIG. 15 is similar to FIGS. 11 and 13 except that the sibling site data center 32 105 of the data center 31 104 receives the request. As shown in FIG. 15, first, a connection request is issued by the user group 1 111 (which may be one or a plurality of users or their agents) (701), and a SYN message 702 is transmitted to the data center 31 104. To do. After receiving the SYN message 702 by the communication interface 201 of the balancing device 31 113 of the data center 31 104, the processing unit 203 is based on the information stored in the storage device 204 and the header information of the received SYN message 702. , And notices that the request exceeds the processing capacity, and decides to process the connection request by the parent site or data center 21102 (703). Thereafter, the processing unit 203 updates the load allocation mapping table, the transfer unit 202 processes the SYN message 702 based on the updated load allocation mapping table, and the communication interface 201 converts the processed SYN message 702 to the data center 21. 102. Similarly, after receiving the SYN message 702 by the balancing device 21 117 of the data center 21 102, notice that the request exceeds the processing capacity and process the connection request by other child sites or data centers 32 105. Is determined (704). Thereafter, the load allocation mapping table is updated, the SYN message 702 is processed based on the updated load allocation mapping table, and the processed SYN message 702 is transferred to the data center 32 105. In the data center 32 105, the processing method performed by the balancing device 32 119 for the SYN message 702 is similar to that in FIG. 11, and a server at the local site is selected to process the connection request (705). Thereafter, the data center 32 105 transmits a response message SYN-ACK 706 to the user and further transmitted to the user group 1 111 via the balancing device 21 117 and the balancing device 31 113, of which the balancing device 21 117 and the balancing device 31. When passing through 113, the header address information is modified based on the existing records in the respective load allocation mapping table. The user group 1 111 establishes a connection with the client terminal after receiving the response message SYN-ACK 706 (707), and transmits the response message ACK 708 again. When the response message ACK 708 passes through the balancing device 31 113 and the balancing device 21 117, the header address information is modified based on the existing record in the respective load allocation mapping table. Finally, the server terminal establishes a connection (709), starts providing services to the user (710), and allows the user to acquire the service (711). Among them, the upstream and downstream load traffic 712 is transferred based on the existing record in the load allocation mapping table when passing through the balancing device 31 113 and the balancing device 21 117.

  FIG. 16 is similar to FIG. 12 and FIG. 14 except that the sibling site data center 32 105 of the data center 31 104 receives the request. As shown in FIG. 16, first, transmission of a packet is started by the user group 1 111 (which may be one or a plurality of users or their agents) (801), and the upstream traffic 802 is transferred to the data center 31 104. Send. After receiving the upstream traffic 802 by the communication interface 201 of the balancing device 31 113 of the data center 31 104, the processing unit 203 is based on the information stored in the storage device 204 and the received header information of the upstream traffic 802. And recognizes that the request exceeds the processing capacity, and decides to process the connection request by the parent site, that is, the data center 21102 (803). Thereafter, the load allocation mapping table is updated by the processing unit 203, the upstream traffic 802 is processed based on the updated load allocation mapping table by the transfer unit 202, and the processed upstream traffic 802 is processed by the communication interface 201 in the data center 21. 102. Similarly, after receiving the upstream traffic 802 by the balancing device 21 117 of the data center 21 102, notice that the request exceeds the processing capacity and process the connection request by the other child sites or data centers 32 105. Is determined (804). Thereafter, the load allocation mapping table is updated, the upstream traffic 802 is processed based on the updated load allocation mapping table, and the processed upstream traffic 802 is transferred to the data center 32 105. In the data center 32 105, the processing performed by the balancing device 21 117 for the upstream traffic 802 is similar to FIG. 12, which identifies this as a new data flow, selects the local site server and The connection request is processed (805), and the load allocation mapping table is updated. Thereafter, the data center 32 105 starts providing the service to the user (806) and allows the user to acquire the service (807). Among them, the upstream and downstream load traffic 808 and 809 are transferred based on the existing records in the load allocation mapping table when passing through the balancing device 31 113 and the balancing device 21 117.

  As described above, according to the load balancing apparatus, the load balancing method, and the hierarchical data center system according to the present invention, the above-described load allocation process can be realized. This improves the service efficiency of the hierarchical data center system.

  In each of the above-described embodiments, the load balancer of the current data center can acquire the resource usage status and link status of other data centers by the following method. For example, each data center load balancer collects local data center resource data and local data center egress router port bandwidth data in real time, and the collected information is stored in a storage device (eg, storage device). Site capacity table and network link status table), send status information to your upper data center at any time or periodically (eg every 10 seconds) to report the above information, Update the information in your storage device.

  In each of the embodiments described above, the number of data centers (that is, sites) included in the hierarchical data center system is seven. However, the present invention is not limited to this, and the hierarchy is determined by the hierarchical method. Any other number may be used as long as the data center system can be configured.

  In each of the above-described embodiments, the data centers included in the hierarchical data center system are connected by a tree type topology, but the present invention is not limited to this, and the data centers may be connected by a hierarchical method. Well, it may be connected by other connection methods such as a mixed connection method.

  In each of the above-described embodiments, the load balancer is installed in the corresponding data center (first data center). However, in the present invention, the load balancer corresponds to the data center (first data center). As long as it can be allocated to the load traffic transmitted to the network, it may be installed outside the data center.

  In each of the above-described embodiments, the message format and the table format shown in the figure are adopted. However, the present invention is not limited to this, and it is only necessary to display corresponding information. A table format may be adopted.

  In each of the above-described embodiments, a method of calculating the resource usage state of the data center based on the arithmetic average value of the CPU usage rate and the storage device usage rate of the local server of the data center is employed. For example, the calculation is based on one or more of the CPU usage rate, the storage device usage rate, and the requested processing speed of the local server, or by adding other parameters. Also good.

Claims (12)

For load traffic transmitted to a current data center corresponding to a load balancing device in the plurality of data centers, which is used in a hierarchical data center system configured by connecting a plurality of data centers by a layering method. A load balancer for performing the allocation,
The load balancer is configured to load traffic transmitted to the current data center based on a resource usage state between the current data center and a lower data center connected to the current data center and positioned below the current data center. A load balancer characterized in that assignment is made to. The load balancer further includes the current data center based on a link state between the current data center and an upper data center connected to the lower data center and the current data center and located in an upper layer of the current data center. 2. The load balancing apparatus according to claim 1, wherein allocation is performed for load traffic transmitted to the center. 3. The load according to claim 1, wherein the load balancer further allocates the load traffic transmitted to the current data center based on a protocol type of a packet constituting the load traffic. 4. Equilibrium device. The load balancing apparatus according to claim 3, wherein the load balancing apparatus determines a protocol type of a packet constituting the load traffic using header information of a network layer. If the resource usage state of the current data center is less than or equal to a predetermined first threshold, the load balancer assigns load traffic transmitted to the current data center to the current data center;
If the resource usage state of the current data center is less than a predetermined second threshold value that is greater than the first threshold value and greater than the first threshold value, the load balancer is configured to include the current data center and the lower data center. And transmission to the current data center based on the link status between the upper data center connected to the current data center and located in the upper layer of the current data center and / or the protocol type of the packet constituting the load traffic The assigned load traffic,
When the resource usage state of the current data center is larger than a predetermined second threshold, the load balancer converts the load traffic transmitted to the current data center based on the resource usage state of the lower layer data center to the lower layer data. 2. The load balancing apparatus according to claim 1, wherein the load balancing apparatus is assigned to a center or the upper data center. A communication interface for sending and receiving packets including at least load traffic and status messages;
A transfer unit for changing header information including at least a source address and a destination address of the packet;
A storage device that stores the resource usage status of the current data center and the resource usage status of the lower data center acquired by the status message;
A processing unit that controls allocation of packets constituting the load traffic of the transfer unit based on a resource usage state stored in the storage device and header information of the packet received by the communication interface. The load balancing apparatus according to claim 1. A load balancing method used in a hierarchical data center system configured by connecting a plurality of data centers by a hierarchical method, and allocating load traffic transmitted to a current data center among the plurality of data centers Because
A determination step of determining a resource usage state of the current data center and a lower data center connected to the current data center and located in a lower layer of the current data center;
An allocation step of allocating load traffic transmitted to the current data center based on a determination result of the determination step. In the determination step, a determination is further made on a link state between the current data center and the lower data center and the upper data center connected to the current data center and positioned in the upper layer of the current data center. The load balancing method according to claim 7. 9. The load balancing method according to claim 7 or 8, wherein in the determination step, determination is further made for a protocol type of a packet constituting load traffic. 10. The load balancing method according to claim 9, wherein in the determination step, a protocol type of a packet constituting the load traffic is determined using header information of the network layer. In the determination step, when it is determined that the resource usage state of the current data center is equal to or less than a predetermined first threshold, the allocation step allocates load traffic transmitted to the current data center to the current data center,
In the determination step, when it is determined that the resource usage state of the current data center is less than a predetermined second threshold value that is greater than the first threshold value and greater than the first threshold value, the allocation step determines that the current data center Based on the link status between the center and the lower layer data center and the upper layer data center connected to the current data center and located in the upper layer of the current data center, and / or the protocol type of the packet constituting the load traffic Assigns to the load traffic currently transmitted to the data center,
In the determination step, when it is determined that the resource usage status of the current data center is greater than a predetermined second threshold, the allocation step is transmitted to the current data center based on the resource usage status of the lower layer data center. The load balancing method according to claim 7, wherein the load traffic is allocated to the lower layer data center or the upper layer data center. A hierarchical data center system configured by connecting a plurality of data centers by a tiering method and including a plurality of load balancing devices respectively corresponding to the data centers,
Each load balancer is configured to use the current data center based on a resource usage state between a current data center corresponding to the load balancer and a lower data center connected to the current data center and positioned below the current data center. A hierarchical data center system characterized by allocating load traffic transmitted to a network.

Images