JP2006146951A - Contents dynamic mirroring system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To generate an optimum mirror server group according to an access-load status to an origin server by making an arbitrary node in a network adaptively serve as a mirror server. <P>SOLUTION: The arbitrary node in a network is given the function of adaptively serving as a mirror server (adaptive mirror node B1) a dynamic mirroring request from the origin server. When an origin server A1 performs dynamic mirroring of contents, an adaptive mirror node detecting means A14 detects an adaptive mirror node in the network, and a dynamic mirroring destination determination means A13 determines a mirroring destination. A dynamic mirroring request means A15 makes the dynamic mirroring request to the selected adaptive mirror node B1, and when the request is accepted, a dynamic mirroring means A16 transfers mirror contents. The adaptive mirror node stores the received mirror contents in a mirror contents storage B101, thus functions as the mirror server. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a dynamic content mirroring system, and in particular, distributes content for distribution such as WWW (World Wide Web) content and video content for the purpose of load distribution of an origin server or a mirror server on which content is already mirrored. The present invention relates to a content dynamic mirroring system for mirroring to a mirror server.

  Conventionally, this type of content dynamic mirroring system has been used to dynamically mirror content to other mirror servers in accordance with the load status of the origin server to achieve load distribution.

  For example, JP-A-11-85604 (Patent Document 1) describes a dynamic mirroring system for continuous media content.

  The content mirroring system described in this publication causes an access load when content requests are generated from a large number of terminal devices to a media server storing continuous media content such as video and audio. Identify the content, analyze the access situation from the terminal device to the identified content, predict the time exceeding the limit value of the access load that the media server can handle, and before reaching the limit value that the media server can handle, Dynamic content placement can be changed by creating a copy of the content or by rearranging the location of the content and other content.

Further, in the mirroring system disclosed in Japanese Patent Application Laid-Open No. 2001-69169 (Patent Document 2), it is detected whether the load of a specific server holding content exceeds a reference value, and the load exceeds the reference value. In this case, the content of the server is copied to another different server (mirror server), and an end host accessing via the network accesses a specific server or a server having a transfer path closest to the mirror server.
Japanese Patent Laid-Open No. 11-85604 JP 2001-69169 A

  As described above, the conventional techniques have the following problems.

  The first problem is that even if content is dynamically mirrored, the mirror server to be dynamically mirrored must be selected from a group of mirror servers that are fixedly arranged in advance. It is.

  Therefore, the effect of load distribution is limited by the number and position of mirror server groups that are fixedly arranged in advance, and there is a possibility that load distribution cannot be achieved efficiently. In order to solve this problem, for example, even if a large number of mirror servers are arranged everywhere in the network so that a load distribution effect can be generated flexibly in any access situation, its installation cost and management cost Is not realistic.

  The second problem is that dynamic mirroring considering the resource status in the network cannot be performed.

  For example, when trying to perform dynamic mirroring of 125 Mbytes content from an origin server, there are two candidate mirror servers, and each candidate mirror server is designated as A and B. Suppose you do. Candidate mirror servers A and B are selected to minimize the content transfer delay for client sites C and D, respectively. At this time, the available bandwidth from the origin server to the clients C and D is assumed to be 1 Mbit / sec. Furthermore, it is assumed that the available bandwidths from the candidate mirror server A to the client site C and from the candidate mirror server B to the client site D are 1 Mbit / sec and 100 Mbit / sec, respectively. In this case, when dynamic mirroring is performed on the candidate mirror server B, the available bandwidth at the time of content acquisition for the client D increases dramatically. On the other hand, even if dynamic mirroring is performed on the candidate mirror server A, the origin server The available bandwidth at the time of content acquisition for client C does not change.

  In another example, similarly, when dynamic mirroring of 125 Mbyte content from the origin server is performed to either of the candidate mirror servers A and B, the available bandwidth from the origin server to each candidate mirror server And the candidate mirror server A is 100 Mbit / sec, and the candidate mirror server B is 1 Mbit / sec. At this time, the content transfer delay required for performing dynamic mirroring on the candidate mirror server A is 10 seconds, whereas when dynamic mirroring is performed on the candidate mirror server B, 1000 seconds are required. If a reduction in the load on the origin server is required immediately, the difference in the transfer delay cannot be ignored, which causes the load distribution performance to be affected.

  From the above example, it can be seen that the performance improvement effect associated with dynamic mirroring can be enhanced by considering the resource status in the network when performing dynamic mirroring.

  The third problem is that, in dynamic mirroring, there is a possibility that a load is applied to the network due to a large amount of traffic accompanying the mirror content transfer from the origin server to the mirror server. In addition, when dynamic mirroring of the same content is simultaneously performed on a plurality of mirror servers, the amount of traffic increases.

  For example, if there is a link with a very small available bandwidth value on the transfer path from the origin server to the mirror server, not only will the transfer delay required for mirroring increase, but the large amount of traffic when transferring mirror content May interfere with other traffic passing through the link.

  When dynamic mirroring of the same content is simultaneously performed on a plurality of mirror servers, the same mirror content is transferred in parallel to each mirror server, so the load of transfer traffic on the network increases accordingly.

  The first object of the present invention is to solve the above-mentioned drawbacks of the prior art, and by making any node in the network adaptively become a mirror server, an optimal mirror server group can be obtained according to the access load situation with respect to the origin server. It is to provide a content dynamic mirroring system that can be generated.

  The second object of the present invention is to provide a content dynamic mirroring system that can improve the performance improvement effect associated with dynamic mirroring by solving the above-mentioned drawbacks of the prior art and considering the resource status in the network. There is.

  The third object of the present invention is to solve the above-mentioned drawbacks of the prior art and reduce the mirror content transfer delay and transfer traffic volume when performing dynamic mirroring from the origin server to one or more mirror servers. It is an object of the present invention to provide a content dynamic mirroring system capable of performing the above.

In order to achieve the above object, the present invention provides an origin server that performs dynamic mirroring for copying mirror content, which is a copy of all or part of content held by its own node, to another mirror server in the network, Server load measuring means for measuring a parameter indicating performance degradation of the origin server accompanying a content acquisition request from the server and determining whether to perform the dynamic mirroring based on the measurement result; and content held by the origin server Mirror content determination means for selecting which of the content as mirror content to perform the dynamic mirroring,
A network resource acquisition unit having a function of acquiring one or both of the in-network resource information and the mirror server performance information, and the mirror content according to a predetermined algorithm based on the information acquired by the network resource acquisition unit Mirror mirror destination determination means for determining to which mirror server to copy, and for performing the dynamic mirroring for the node that is the dynamic mirror destination determined by the dynamic mirror destination determination means Mirroring request means for issuing a dynamic mirroring request, and when the dynamic mirroring request is received, the mirror content for the node determined as a dynamic mirroring destination using the mirroring tree as a transfer path Mirror content transfer means for transferring Mirroring tree creating means having a function of calculating a ringing tree, and the mirroring tree created by the mirroring tree creating means is a mirror server selected as a relay hop in addition to the mirror server to which the mirror content is copied As a component of the mirroring tree.

The present invention of claim 2 is an origin server that performs dynamic mirroring that copies mirror content, which is a copy of all or part of the content held by the node, to another mirror server in the network. A server load measuring means for measuring a parameter indicating performance degradation of the origin server accompanying a content acquisition request and determining whether to perform the dynamic mirroring based on the measurement result;
Mirror content determination means for selecting which content of the content held by the origin server is to be used as the mirror content and performing the dynamic mirroring;
Network resource acquisition means having a function of acquiring either or both of the network resource information and the mirror server performance information;
Dynamic mirroring destination determining means for determining which mirror server to copy the mirror content according to a predetermined algorithm based on the information acquired by the network resource acquiring means;
Dynamic mirroring request means for issuing a dynamic mirroring request for performing dynamic mirroring to the node to be the dynamic mirroring destination determined by the dynamic mirroring destination determination means; and If accepted, mirror content transfer means for transferring the mirror content to the node determined as a dynamic mirror destination with the mirroring tree as a transfer path, and a mirroring tree having a function of calculating the mirroring tree Creating mirror, wherein the mirroring tree creating means includes the origin server and the origin server that is a component of the mirroring tree in a mirroring tree configured by only the origin server and the mirror server to which the mirror content is copied, and the mirror server Re-creating a mirroring tree in which the transfer path where the usable bandwidth of the transfer path between two servers is equal to or less than a predetermined threshold is replaced with a transfer path that relays using a mirror server other than the constituent elements of the mirroring tree It is characterized by.

  According to the third aspect of the present invention, the mirroring tree creation means includes resource information of the origin server and the mirror server that are constituent elements of the mirroring tree to be created, and the information between the origin server and the mirror server. The mirroring tree is calculated using either or both of resource information in a mirror content transfer path.

  According to a fourth aspect of the present invention, the dynamic mirroring destination determination means belongs to a client that makes a content acquisition request to the origin server obtained by the network resource acquisition means in determining the dynamic mirroring destination. The content transfer throughput (first content transfer throughput) per session from the origin server to the client site, measured for each client site that is a site, and the content acquisition delay for the client site are minimized. The content transfer throughput (second content transfer throughput) per session from the selected mirror server to the client site is used.

  According to the present invention of claim 5, in the determination of the dynamic mirroring destination, the dynamic mirroring destination determination unit determines the number of content acquisition requests from the client site per unit time × (the second content throughput / the first The higher the value of (one content throughput), the higher the priority is given.

  According to the present invention of claim 6, the dynamic mirroring destination determination means belongs to a client that makes a content acquisition request to the origin server obtained by the network resource acquisition means in the determination of the dynamic mirroring destination. Measured for each client site that is a site, the larger the available bandwidth of the transfer path from the mirror server that is a candidate for dynamic mirroring destination to the client site, or the candidate for dynamic mirroring destination The higher priority is given to the smaller one of the RTT from the mirror server to the client site or the transfer delay of the fixed-length data in the direction in which the client site is downstream. .

  The present invention according to claim 7 is characterized in that the dynamic mirroring destination determining means determines the content transfer throughput per session from the origin server to the mirror server or the available bandwidth of the transfer path in determining the dynamic mirroring destination. The higher the value of is, the higher the priority is given.

  The present invention of claim 8 is a node of a dynamic mirroring system that performs dynamic mirroring that copies mirror content, which is a copy of all or part of content held by an origin server, to another mirror server in the network. When transferring the mirror content from the origin server to one or a plurality of the mirror servers, the mirror content according to a mirroring tree that is an arbitrary tree-like path constituted by the mirror server with the origin server at the top Relaying the mirror content received from the origin server or the mirror server upstream of the mirroring tree in the transfer of the mirror content using the mirroring tree as a transfer path from the origin server, The mirroring tree A mirror content relay means for transferring to the mirror server downstream of the mirror server, and when relaying the mirror content received from the origin server or mirror server upstream of the mirroring tree, The operation of storing or not storing content is selectively performed based on an instruction from an origin server.

  The present invention of claim 9 is a node of a dynamic mirroring system that performs dynamic mirroring that copies mirror content that is a copy of all or part of content held by an origin server to another mirror server in the network. When transferring the mirror content from the origin server to one or a plurality of the mirror servers, the mirror content according to a mirroring tree that is an arbitrary tree-like path constituted by the mirror server with the origin server at the top Relaying the mirror content received from the origin server or the mirror server upstream of the mirroring tree in the transfer of the mirror content using the mirroring tree as a transfer path from the origin server, The mirroring tree A mirror content relay means for transferring to the mirror server downstream of the mirror server, and when relaying the mirror content received from the origin server or mirror server upstream of the mirroring tree, It is characterized in that the operation of whether or not to store the content can be autonomously selected based on the state of the own node.

  In the first content dynamic mirroring system of the present invention, an adaptive mirror node (B1 in FIG. 1) function that can adaptively become a mirror server is added to any node arranged in the network. Then, a dynamic mirroring request from an arbitrary origin server is sent to a dynamic mirroring request unit (A15 in FIG. 1) on the origin server, a mirror content transfer unit (A16 in FIG. 1), and a dynamic mirror node on the adaptive mirror node. This is done via the mirroring request accepting means (B11 in FIG. 1), and the adaptive mirror node functions as a mirror server. The origin server can target any node in the network configured as an adaptive mirror node for dynamic mirroring, and can construct a mirror server network adaptively according to the access load status of the origin server. Become. Thereby, the first object of the present invention can be achieved.

  The second content dynamic mirroring system of the present invention is dynamic in consideration of the network resource status acquired through the network resource acquisition means (F11 in FIG. 11) during dynamic mirroring from the origin server to the adaptive mirror node. Determine the mirror destination. An example of a dynamic mirroring destination determination algorithm will be described in the following embodiments of the invention. Thereby, the second object of the present invention can be achieved.

  In the third dynamic content mirroring system of the present invention, the mirror content transfer path from the origin server to the adaptive mirror node is calculated by the mirroring tree creation means (G11 in FIG. 15). The mirror content is transferred on the calculated mirroring tree. By selecting the optimal mirroring tree, it is possible to reduce the mirror content transfer delay during dynamic mirroring, especially when transferring mirror content to multiple mirror destinations. It becomes possible to do. Thereby, the third object of the present invention can be achieved.

  As described above, according to the present invention, the following effects are achieved.

  The first effect is that an origin server uses an adaptive mirror node arranged in the network to adaptively construct a mirror server network according to the access load status of the origin server, thereby maximizing the effect of load distribution. be able to.

  The reason is that an arbitrary mirror node function such as a router or a proxy server arranged in the network can be made to function as a mirror server in response to a dynamic mirroring request from an arbitrary origin server. This is because the origin server can target any node in the network configured as an adaptive mirror node for dynamic mirroring.

  The second effect is that the client can always acquire content with low delay and high throughput.

  The reason for this is that if access to the origin server is concentrated and content acquisition delay and throughput performance decrease, content replication with a high frequency of acquisition requests is automatically copied to the nearest adaptive mirror node for the client. This is because the client can acquire the content from the nearest adaptive mirror node.

  The third effect is that it is possible to reduce the installation cost and the management cost required since the functions of the proxy server and the mirror server must be realized as independent nodes.

  The reason is that by implementing the adaptive mirror node function on the proxy server, it is possible to provide both the function provided by the proxy server and the function provided by the mirror server on the same node.

  The fourth effect is that it is possible to reduce the traffic to the origin server and the access load of the origin server due to the acquisition request for the content requiring authentication / encryption processing by the SSL protocol, execution of the CGI script, and the like.

  The reason is that the proxy server has only the content cache function in the past, so it cannot accept the acquisition request for the above content and only forward the request to the origin server. This is because the content can be handled by providing a function as a mirror server.

  The fifth effect is that when content mirroring is performed, conventionally, an appropriate mirror server must be selected on the client side, and a content acquisition request must be transmitted to the server corresponding to the destination IP address. However, mirror contents can be acquired by simply setting the proxy server on the client side without the necessity.

  The reason is that by setting the adaptive mirror proxy server as a proxy server in the subordinate client group, all content acquisition requests from the subordinate client group are received by this adaptive mirror proxy server, and the adaptive mirror proxy server This is because the content acquisition request is directly responded only to the acquisition request for the mirrored content, and the acquisition request for the other content is selectively processed as a normal proxy server.

  The sixth effect is that the load on the origin server that performs authentication and billing processing can be reduced.

  The reason for this is that authentication and billing processes can only be performed by the origin server, but authentication and billing processes are also performed at the same time when mirror content is transferred from the origin server to the adaptive mirror node due to dynamic mirroring. This is because by entrusting the origin mirror to the adaptive mirror node, the adaptive mirror node can perform authentication and accounting processing.

  The seventh effect is that the client can always acquire content that requires authentication and billing processing with low delay and high throughput.

  The reason for this is that when access to the origin server is concentrated and content acquisition delay and throughput performance drop, even content that requires authentication and billing processing is dynamically changed to the adaptive mirror node closest to the client. This is because the copy of the content is automatically copied to the nearest adaptive mirror for the client, and at the same time, authentication and billing processing are entrusted, so that the client can acquire the content from the nearest adaptive mirror node.

  As an eighth effect, when performing dynamic mirroring, the mirror server can be selected so as to maximize the performance improvement effect by performing dynamic mirroring.

  This is because the network resource status is taken into consideration during dynamic mirroring.

  The ninth effect is that, at the time of dynamic mirroring, it is possible to reduce the transfer delay of the mirror content and transfer the mirror content so as not to disturb other traffic on the transfer path.

  This is because a transfer path that uses another adaptive mirror node as a relay hop can be selected in the mirror content transfer path from the origin server to the adaptive mirror node that is the dynamic mirroring destination.

  The tenth effect is that the transfer delay and the amount of transfer traffic can be reduced in the transfer of mirror contents to a plurality of mirroring destinations.

  The reason is that when performing dynamic mirroring simultaneously on multiple mirroring destinations, the origin server creates a mirroring tree consisting of the adaptive mirror node that is the mirroring destination and the adaptive mirror node for relay, and the mirror content is created This is because it can be transferred along the mirroring tree.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  Referring to FIG. 1, the first embodiment of the present invention includes an origin server A1, an adaptive mirror node B1 to be subjected to content dynamic mirroring, and a client C1.

  The origin server A1 includes not only the origin server that owns the original content but also the mirror server in which the same content has already been mirrored, and hereinafter both are collectively referred to as the origin server. The origin server A1 has a function of actively performing content dynamic mirroring.

  The adaptive mirror node B1 is a device having a function that allows any node arranged in the network to adaptively become a mirror server, and is a feature of the configuration of the present embodiment. Examples of the adaptive mirror node include a router having an adaptive mirror node function, a layer 4 or layer 7 switch having an adaptive mirror node function, a proxy server having an adaptive mirror node function, and the like. Further, it may be a dedicated node having only the adaptive mirror node function and not having other functions. Considering an example in which an adaptive mirror node function is implemented in a router, when mirroring is being performed, the router functions not only as a router but also as a mirror server at the same time.

  The adaptive mirror node B1 functions as a mirror server in response to a dynamic mirroring request from an arbitrary origin server. From the viewpoint of the origin server A1 side, any node in the adaptive mirror node network can be the target of dynamic mirroring, and the mirror server network can be adaptively adapted according to the access load status in the origin server A1. Can be built.

  The client C1 may be not only a user terminal that makes a content acquisition request but also a proxy server that accommodates a plurality of clients and makes a content acquisition request on behalf.

  The origin server A1 includes a storage device A10 that stores content and content processing methods, processing programs, etc., server load measurement means A11, mirror content determination means A12, dynamic mirror destination determination means A13, and adaptive mirror node detection Means A14, dynamic mirroring request means A15, and mirror content transfer means A16 are included.

  The server load measuring means A11 measures the access load of the origin server A1, and determines whether or not load distribution should be performed by performing dynamic mirroring. Here, the access load is a parameter indicating the performance degradation of the origin server due to concentration of content acquisition requests such as the CPU load of the origin server A1, the number of content transfer sessions, and the content transfer throughput per session. Here, as an example of the content transfer session, there is a TCP session used at the time of content transfer by HTTP. Hereinafter, a TCP session will be described on behalf of a content transfer session.

  When performing dynamic mirroring to another mirror server, the mirror content determination unit A12 determines which content can be dynamically mirrored to maximize the load balancing effect, and determines the content to be dynamically mirrored. .

  The dynamic mirroring destination determination unit A13 determines one or more adaptive mirror nodes that can maximize the load distribution effect in the dynamic mirroring of the mirror content determined by the mirror content determination unit A12, and the access load status of the origin server A1. Determine based on. At this time, the adaptive mirror node existing in the network is detected by the adaptive mirror node detecting means A14.

  An example of an algorithm for determining an adaptive mirror node that is a dynamic mirroring destination capable of maximizing the load balancing effect in the dynamic mirroring destination determination unit A13 will be described. First, the number of content acquisition requests per unit time for content to be subjected to dynamic mirroring is measured for each management domain in the network where the client exists. Then, an adaptive mirror node existing in the management domain having a value of the content acquisition request number equal to or greater than a certain threshold is determined as a dynamic mirroring destination. An example of such a management domain is AS (Autonomous System) on the Internet.

  The adaptive mirror node detection means A14 has a function of detecting an adaptive mirror node existing in the network. At this time, detecting all the adaptive mirror nodes existing in the network has a large overhead. Therefore, adaptive mirror node detection should be done with as little overhead as possible.

  As an example, a method is conceivable in which a representative management server arranged for each management domain in the network is inquired about an adaptive mirror node list existing in the management domain. The representative management server manages the adaptive mirror nodes in each management domain, and grasps the position, resource status, and the like of each adaptive mirror node. The origin server A1 can reduce the detection overhead by inquiring the adaptive mirror node list only with the representative management server of the management domain in which mirror contents are newly placed. An example of such a management domain is AS on the Internet.

  Another example of a method for detecting an adaptive mirror node in the network will be described. It is assumed that an adaptive mirror node exists on the passage route of the content acquisition request from the client C1 to the origin server A1. A router, a layer 4 / layer 7 switch, a proxy server, etc. can be considered as an adaptive mirror node existing on such a passing route. At this time, information that an adaptive mirror node exists on the passage route of the content acquisition request is added to the content acquisition request from the client C1. The origin server A1 that has received the content acquisition request to which the information is added identifies that there is an adaptive mirror node on the passage path of the content acquisition request from the client C1 to the origin server A1. The origin server A1 can determine that dynamic mirroring should be performed on the adaptive mirror node in order to reduce the access load due to the increase in content acquisition requests from the client C1.

  The dynamic mirroring request unit A15 transmits a dynamic mirroring request to the adaptive mirror node determined as the target of the dynamic mirroring request by the dynamic mirroring destination determination unit A13. Further, when a response from the adaptive mirror node to the dynamic mirroring request is received and accepted, the mirror content transfer means A16 actually performs dynamic mirroring. If rejected, the dynamic mirroring is given up or the dynamic mirroring is performed. Search for other adaptive mirror nodes to power.

  The mirror content transfer means A16 reads the mirror content determined from the storage device A10 and transmits it to the adaptive mirror node that has accepted the dynamic mirroring request. The mirror content transmitted at this time includes not only objects such as text, images, and moving images but also various processing programs and information associated with the content. Examples of such processing programs and information include search programs, CGI (Common Gateway Interface) scripts, and authentication / encryption information such as SSL (Secure Socket Layer). It also has a function of compressing the mirror content to be transmitted in order to reduce transfer traffic, and a function of encrypting the mirror content to be transmitted in order to ensure security during transfer.

  The adaptive mirror node B1 includes a storage device B10, a dynamic mirroring request receiving unit B11, a request data receiving unit B12, a response content creating unit B13, and a content response unit B14.

  The storage device B10 includes a mirror content storage unit B101.

  The mirror content storage unit B101 stores the dynamically mirrored content. What is stored includes not only objects such as text, images, and moving images, but also various processing programs and information associated with the contents.

  The dynamic mirroring request accepting unit B11 performs processing of content dynamic mirroring request from the origin server A1 and processing of receiving mirror content transferred from the origin server.

  The request data receiving unit B12 receives the content acquisition request from the client C1, and passes the request content to the response content creating unit B13.

  The response content creation unit B13 reads the requested content from the mirror content storage unit B101 and creates response content. At this time, if it is necessary to execute various processing programs associated with the content or read out the authentication information, the processing is appropriately performed. Then, the created response content is passed to the content response means B14.

  The content response unit B14 transmits the response content created by the response content creation unit B13 to the client C1 that has transmitted the content acquisition request.

  FIG. 2 is a diagram showing in detail the configuration of the dynamic mirroring request accepting unit B11. The dynamic mirroring request receiving unit B11 includes a data transmitting / receiving unit B111 with the origin server, a dynamic mirroring request receiving determining unit B112, a mirror content receiving unit B113, and a decompression / decryption unit B114.

  Data transmission / reception means B111 with the origin server receives the data from the origin server A1 and identifies its contents. If the content of the received data is a dynamic mirroring request, the dynamic mirroring request acceptance determining means B112 is inquired whether the dynamic mirroring request can be accepted, and the result is returned to the origin server A1. Further, when the mirror content received from the origin server A1 is received, the received data is passed to the mirror content receiving means B113.

  The dynamic mirroring request acceptance determination unit B112 determines whether or not the dynamic mirroring request from the origin server A1 can be accepted.

  The mirror content receiving means B113 stores the mirror content received from the origin server A1 in the mirror content storage unit B101. If the received mirror content is compressed or encrypted, the decompression / decryption means B114 performs necessary decompression / decryption processing and then stores it in the mirror content storage unit B101.

  The decompression / decryption means B114 has a function of decompressing / decrypting the mirror content received from the origin server A1. In an environment where the mirror content received from the origin server A1 is not compressed / encrypted, the decompression / decryption means B114 is not necessarily required.

  Next, with reference to FIG. 3, in this Embodiment, the operation | movement in the origin server A1 at the time of dynamic mirroring is demonstrated in detail.

  In the origin server A1, when the dynamic mirroring function is enabled, the server load measuring unit A11 periodically monitors the CPU load of the server, and determines whether the server CPU load is equal to or higher than a preset threshold value. Check (steps S101 and S102 in FIG. 3). Examples of targets to be monitored here include the number of content acquisition requests per unit time, content transfer throughput, and the like in addition to the CPU load.

  If the server load is equal to or greater than the threshold value in step S102, the mirror content determination unit A12 determines the content to be dynamically mirrored (step S103). Here, the content to be dynamically mirrored should be selected with a large granularity such as a page unit or a website unit, rather than being selected with a fine granularity of an object unit. Examples of criteria for determining content to be dynamically mirrored include: a. “The number of content acquisition requests from a client per unit time exceeds a threshold value”, b. “Those satisfying criterion a and the content size exceeds the threshold value”, c. “Maximize content transfer throughput for clients as a result of load balancing by dynamic mirroring according to content type and size”.

  If the server load is equal to or less than the threshold value in step S102, it is considered that load distribution is not necessary, and dynamic mirroring is not performed.

  After step S103, the dynamic mirroring destination determination unit A13 determines one or a plurality of adaptive mirror nodes that maximize the load distribution effect for each content to be subjected to dynamic mirroring (step S104). , S105). In the determination of the adaptive mirror node, the detection result by the adaptive mirror node detection means A14 is used. In step S104, if it is determined that the overhead of dynamic mirroring is larger than the load distribution effect by dynamic mirroring, execution of dynamic mirroring is stopped and it is determined that there is no target adaptive mirror node. To do.

  Next, if there is an adaptive mirror node to be subjected to dynamic mirroring in steps S104 and S105, the dynamic mirroring request means A15 transmits a dynamic mirroring request message to each adaptive mirror node (step S106). ). The dynamic mirroring request message includes information about the content such as content type and size. If there is no target adaptive mirror node in step S105, dynamic mirroring is not performed.

  When an acceptance response to the dynamic mirroring request message is received from the adaptive mirror node (step S107), the mirror content transfer unit A16 reads the mirror content from the storage device A10, and receives the dynamic mirroring from the adaptive mirror node. Transmit (step S108). At this time, if necessary, the mirror contents are transmitted after being compressed or encrypted.

  In step S108, in addition to the method of directly transmitting the mirror content from the storage device A10 in the origin server A1 to the adaptive mirror node B1, another mirror server holding the same content is detected, and the mirror server with the smallest transfer delay is detected. A method of transmitting from the adaptive mirror node B1 to the adaptive mirror node B1 can also be considered. Also, transfer delay is reduced by dividing one mirror content into a plurality of parts, transmitting each part divided from a plurality of mirror servers holding the same content in parallel, and assembling at the adaptive mirror node B1 A method to do this is also conceivable.

  FIG. 4 shows in detail the configuration of the mirror content transfer means A16 for realizing the method of performing dynamic mirroring from the mirror server with the smallest transfer delay in step S108 of FIG. The mirror content transfer unit A16 includes a communication unit A161 with another mirror server, a source determination unit A162, and a communication unit A163 with an adaptive mirror node.

  The communication means A161 with other mirror servers has a function of communicating with other mirror servers, inquires about the contents and network resources to be held with respect to the other mirror servers, and further instructs mirror content transfer. The source determination unit A162 determines the source with the smallest transfer delay based on the network resources from the own node or another mirror server to the adaptive mirror node B1. The communication means A163 with the adaptive mirror node has a function of communicating with the adaptive mirror node B1, and transfers mirror contents to the adaptive mirror node B1.

  Next, detailed operations when the mirror content transfer means A16 shown in FIG. 4 performs dynamic mirroring from the mirror server with the smallest transfer delay in step S108 will be described with reference to FIG.

  First, a mirror server that holds the same content as the content to be dynamically mirrored is detected (step S801). When a mirror server holding the same content can be detected in step S801 (step S802), network resources from each detected mirror server to the adaptive mirror node to which the content is transferred, and from the next node to the adaptive mirror node Network resources are acquired (step S803). Next, using the network resource information acquired in step S803, a transmission source with the smallest transfer delay when transmitting mirror content to the adaptive mirror node is determined (step S804).

  When the transmission source is the own node (step S805), the mirror content is transmitted from the own node to the adaptive mirror node (step S806). Even if there is no mirror server that holds the same content in step S802, step S806 is executed. In step S805, if the transmission source is not the local node but another mirror server, the mirror server serving as the transmission source is instructed to transfer the mirror content (step S807), and the mirror content is transmitted from the mirror server. By receiving a message that the transfer is completed, the mirror content transfer completion is confirmed (step S808).

  FIG. 6 shows an example of a method of performing dynamic mirroring from the mirror server with the smallest transfer delay in step S108. An origin server 101 and adaptive mirror nodes 102, 103, and 104 are connected to the backbone network 1 via links 201, 202, 203, and 204, respectively. Now, assume that the content in the adaptive mirror node 104 is to be dynamically mirrored from the origin server 101, and both the adaptive mirror nodes 102 and 103 already have the same content. At this time, Table 301 shows the available bandwidth from each origin server and adaptive mirror node 104 to adaptive mirror node 104. When this available bandwidth information is given, the origin server 101 determines that dynamic mirroring can be performed with less transfer delay if the content is transferred from the adaptive mirror node 103, and the adaptive mirror node 103 Instructs the adaptive mirror node 104 to transmit the content. The adaptive mirror node 103 instructed to transfer the content transfers the content to the adaptive mirror node, confirms the completion of the transfer, and then responds to the origin server 101 that the content has been transferred normally.

  If a rejection response to the dynamic mirroring request message is received in step S107, the process returns to step S104, and the adaptive mirror node that has made a rejection response is excluded and the adaptive mirror node is determined again.

  Next, with reference to FIG. 7, in this Embodiment, the operation | movement in the adaptive mirror node B1 at the time of dynamic mirroring is demonstrated in detail.

  When the adaptive mirror node B1 receives the dynamic mirroring request message from the origin server A1 (step S201 in FIG. 7), the dynamic mirroring request accepting unit B11 determines whether or not the request can be accepted (step S202). Examples of criteria for determining whether or not a request can be accepted include: a. “Is there enough resources left to store mirror content in storage device B10 in adaptive mirror node B1,” b. “If there is a processing program associated with the mirror content, is the environment capable of executing the processing program?” C. “Is there enough network resources from the mirror server to the client?”

  If it is determined in step S202 that the request can be accepted, an acceptance response is transmitted to the origin server A1 (step S203). If it is determined that the request cannot be accepted, a rejection response is transmitted (step S204).

  When the acceptance response is transmitted in step 203, the mirror server B1 receives the mirror content from the origin server A1, and stores it in the mirror content storage unit B101 in the storage device B10 (step S206). At this time, if the received mirror content is compressed or encrypted, it is decompressed or decrypted and stored.

  The operation of the mirror server B1 when receiving a content acquisition request from the client is the same as the operation of the conventional mirror server.

  Next, the effect of this embodiment will be described.

  In the prior art, when the origin server performs dynamic mirroring to distribute the access load, the dynamic mirroring destination can be selected only from those that are fixedly arranged in advance as a mirror server. For this reason, the access load distribution effect is limited by the number and position of mirror servers arranged in advance.

  In the present embodiment, an arbitrary mirror node function such as a router or a proxy server arranged in the network is provided with an adaptive mirror node function, and a dynamic mirroring request from an arbitrary origin server is received to function as a mirror server. The origin server can target any node in the network configured as an adaptive mirror node for dynamic mirroring, and can construct a mirror server network adaptively according to the access load status of the origin server. Become.

  From the client's point of view, when the access load is concentrated on the origin server, the mirror content is automatically copied to the nearest adaptive mirror node, and content can always be acquired with low latency and high throughput. become.

  Next, a second embodiment of the present invention will be described in detail with reference to the drawings.

  The second embodiment of the present invention corresponds to the case where the proxy server is changed to an adaptive mirror node in the first embodiment.

  Referring to FIG. 8, the second embodiment of the present invention is different in that the adaptive mirror node B1 in the first embodiment shown in FIG. 1 is an adaptive mirror proxy server D1.

  The adaptive mirror proxy server is a proxy server having an adaptive mirror node function, and includes a storage device D10, dynamic mirroring request reception means B11, response content creation means B13, content response means B14, and request data. It includes a receiving unit D11, an operation determination unit D12, a cache hit determination unit D13, a content acquisition request transfer destination determination unit D14, and a content acquisition request transfer unit D15.

  The storage device D10 is different from the storage device B10 in that it includes a cache storage unit D101. The cache storage unit D101 is used to store cache content when the adaptive mirror proxy server has a content cache function as a proxy server.

  The request data receiving unit D11 receives the content acquisition request from the client C1 and passes it to the operation determining unit D12.

  The operation determination unit D12 identifies the content requested by the content acquisition request. Then, according to the request contents, it is determined whether to operate as a mirror server or a proxy server. Simply, if it is an acquisition request for content mirrored in the adaptive mirror proxy server D1, it is determined to operate as a mirror server, otherwise it is determined to operate as a proxy server. In addition, even when mirroring, a method of authenticating a client and performing an operation as a mirror server only for an authorized client may be considered.

  The cache hit determination means D13 determines whether or not the content corresponding to the content acquisition request from the client exists in the cache storage unit D101, and if it exists, passes the corresponding content to the content response means B14. If it does not exist, the content acquisition request is transferred to the content acquisition request transfer destination determination means D14.

  The content acquisition request transfer destination determination unit D14 determines the origin server or mirror server to which the content acquisition request is transferred when the adaptive mirror proxy server D1 does not mirror or cache the content corresponding to the content acquisition request. To do.

  The content acquisition request transfer unit D15 transfers the content acquisition request to the origin server or mirror server determined by the content acquisition request transfer destination determination unit D14.

  Next, with reference to FIG. 9, in this Embodiment, the operation | movement in the adaptive mirror proxy server D1 at the time of content acquisition request reception is demonstrated in detail.

  In the adaptive mirror proxy server D1, when the request data receiving unit D11 receives a content acquisition request from the client C1 (step S301 in FIG. 9), the operation determining unit D12 operates on the content mirrored in the adaptive mirror proxy server D1. Whether the request is an acquisition request is identified (step S302).

  In step S302, if the content corresponding to the content acquisition request is mirrored in the adaptive mirror proxy server D1, the response content creation unit B13 reads the content from the mirror content storage unit B101 and creates the response content (step S302). S303). At this time, if the requested content requires execution of some processing program, authentication processing, or the like, the execution and processing are appropriately performed to create response content. The response content created according to the above procedure is transferred to the content response means B14 and responds to the client C1 (step S305).

  In step S302, if the content corresponding to the content acquisition request is not mirrored in the adaptive mirror proxy server D1, the content corresponding to the content acquisition request is present in the cache storage unit D101 by the cache hit determination unit D13. Whether or not (step S304).

  If the content exists in step S304 (hits the cache), the corresponding content is read from the cache storage unit D101, the content is transferred to the content response means B14, and the client C1 is responded (steps S306 and S305).

  If the content does not exist in step S304 (no hit in the cache), the content acquisition request transfer destination determination unit D14 determines the origin server or mirror server to which the content acquisition request is transferred (step S307).

  When the adaptive mirror proxy server D1 has means for detecting a plurality of candidate transfer destinations, the content acquisition request transfer destination determination means D14 selects one from these candidate transfer destinations. Examples of selection criteria at this time include those with the lowest load and those with the largest available bandwidth in the downstream direction (from the server to the client).

  The content acquisition request transfer unit D15 transfers the content acquisition request to the content acquisition request transfer destination determined in step S307 (step S308).

  Next, the effect of this embodiment will be described.

  In this embodiment, an adaptive mirror node function is implemented on the proxy server. Therefore, it is possible to provide both the function provided by the proxy server and the function provided by the mirror server on the same node, and it was necessary to install each function as a separate node. Costs and management costs can be reduced.

  From the standpoint of a proxy server, since it has only a conventional content cache function, it cannot accept an acquisition request for content that requires authentication / encryption processing by the SSL protocol, execution of a CGI script, etc. There was no choice but to transfer. By implementing the function of a mirror server by implementing the adaptive mirror node function, it becomes possible to accept an acquisition request for the above content, and to reduce the traffic to the origin server and the access load of the origin server.

  Furthermore, in this embodiment, by setting the adaptive mirror proxy server as a proxy server in the subordinate client group, all content acquisition requests from the subordinate client group are received by this adaptive mirror proxy server. The adaptive mirror proxy server looks at each content acquisition request and directly responds only to the acquisition request for the mirrored content, and the acquisition request for the other content selectively performs processing as a normal proxy server. Conventionally, when content mirroring is performed, an appropriate mirror server must be selected on the client side and a content acquisition request must be transmitted to the server corresponding to the destination IP address. In this embodiment, There is no need for this, and mirror content can be acquired simply by setting a proxy server on the client side.

  Next, a third embodiment of the present invention will be described in detail with reference to the drawings.

  In the third embodiment of the present invention, authentication and accounting processing is required for data communication between the adaptive mirror node and the client in the first embodiment, and the adaptive mirror node authenticates and charges. This corresponds to the case of having a processing function.

  Referring to FIG. 10, the third embodiment of the present invention is different from the first embodiment shown in FIG. 1 in that an adaptive mirror node E1 is used instead of the adaptive mirror node B1. .

  The adaptive mirror node E1 includes response content creation means B13, content response means B14, storage device E10, dynamic mirroring request reception means E11, request data reception unit E12, and authentication / billing information processing means E13.

  The storage device E10 differs from the storage device B10 in that it includes an authentication / billing information management unit E101. The authentication billing information management unit E101 manages information related to authentication and billing. Examples of information to be managed include an authentication ID of each client, a password, and a download price for each content. Further, there is dynamic information such as a charging status for each client.

  In addition to the function of the dynamic mirroring request accepting unit B11 in FIG. 1, the dynamic mirroring request accepting unit E11 performs procedures necessary for the adaptive mirror node E1 to perform authentication and billing processing instead of the origin server A1. In this procedure, the origin server A1 and the adaptive mirror node E1 trust each other to perform authentication and charging processing entrustment, and the adaptive mirror node E1 among the authentication and charging information of the origin server A1 includes the client C1. Information required for authentication and billing of the content acquisition request received from the server is stored in the authentication billing information management unit E101, and then the information stored in the authentication billing information management unit E101 with the origin server A1 Synchronization procedure. Further, when the management of the authentication billing information is centralized in the origin server A1, the authentication billing information managed in the origin server A1 during the authentication and billing process performed by the authentication billing information processing means E13 in the adaptive mirror node E1. It also includes procedures for reading and writing. Furthermore, exchange between the origin server A1 and the adaptive mirror node E1 in reading and writing of authentication / billing information managed in the origin server A1 at the time of database synchronization, authentication, and billing processing with the origin server A1 The data to be processed may be compressed and encrypted.

  The request data receiving unit E12 receives the content acquisition request from the client C1, and performs the authentication process via the authentication / billing information processing unit E13. If the content acquisition is permitted as a result of the authentication, the requested content is passed to the content creating means B13 to instruct the creation of the content.

  The authentication billing information processing means E13 has a function of authenticating the content acquisition request received by the request data receiving unit E12 and charging for the authenticated request. At this time, information such as the requested client ID and password is read out from the authentication / billing information management unit E101, and charging information generated when the content is acquired is written into the authentication / billing information management unit E101.

  The above example is a node in which authentication and charging processing functions are introduced for the adaptive mirror node B1, but in addition, authentication and charging processing for the adaptive mirror proxy server D1 according to the second embodiment. The configuration of the node in which the function is introduced can be considered similarly.

  Next, the effect of this embodiment will be described.

  In the present embodiment, authentication and billing processing can be performed in the adaptive mirror node. Conventionally, such processing can be performed only by the origin server, but the load on the origin server can be reduced by entrusting the authentication and accounting processing to the adaptive mirror node dynamically. In the case of content with authentication and billing processing, compared to the case of normal content, much more processing is required for the request response from the client. It can be said that it is larger than the case of mirroring.

  Next, a fourth embodiment of the present invention will be described in detail with reference to the drawings.

  The fourth embodiment of the present invention is different from the first embodiment in that the dynamic mirroring destination is determined in consideration of the network resource status at the time of dynamic mirroring from the origin server.

  Referring to FIG. 11, the fourth embodiment of the present invention has a network resource acquisition means F11 in place of the origin server A1 in the first embodiment shown in FIG. The difference is that the origin server F1 includes a dynamic mirroring destination determination unit A13a that determines a dynamic mirroring destination based on the network resource information acquired by the acquisition unit F11. Although the adaptive mirror node B1 is illustrated as a destination where dynamic mirroring is performed, in the present embodiment, a conventional fixed mirror server may be used. In the present embodiment, the dynamic mirroring destination is hereinafter collectively referred to as a mirror server.

  The network resource acquisition unit F11 has a function of acquiring network resource information related to dynamic mirroring, such as available bandwidth and delay to other mirror servers, and access load status of other mirror servers. In addition to the above, the network resource information may include information related to server performance such as the CPU load of other mirror servers and the number of TCP sessions. As a method for acquiring the above network resource and server performance information, it is collected by a method of dynamically acquiring a measurement packet by flowing it to each mirror server, or a routing protocol that supports resource information exchange of each node / link in the network. A method using the information obtained can be considered.

  The dynamic mirroring destination determination unit A13a of the present embodiment includes a network resource base calculation unit A130 that performs a calculation according to an algorithm described later based on the resource information obtained by the network resource acquisition unit F11. To determine the optimal dynamic mirroring destination.

  Conventionally, network resource information has not been considered as a priority when determining a mirror server to which dynamic mirroring is performed. In a typical example, unit time from a client site supported by the mirror server is typical. The number of content acquisition requests per unit was used. The number of content acquisition requests is the number of acquisition requests for content to be dynamically mirrored. By performing dynamic mirroring of the content, the number of content acquisition requests received at the origin server, that is, the decrease in the access load of the origin server is reduced. The goal is to maximize.

  This embodiment is different in that priority is determined in consideration of network resources related to each mirror server in addition to the number of content acquisition requests. Hereinafter, a plurality of examples of network resources to be considered in addition to the number of content acquisition requests when determining the priority of the dynamic mirroring destination will be given.

  (1) Minimize the TCP throughput per session from the origin server to each client site and the content acquisition delay for the client site, measured for each site to which the client making a content acquisition request to the origin server belongs TCP throughput per session from the selected mirror server to the client site.

  (2) In the above (1), an available bandwidth is used instead of the TCP throughput per session.

  (3) In (1) above, RTT (Round Trip Time) is used instead of TCP throughput per session.

  (4) In the above (1), the fixed-length data transfer delay time (in the direction where the client site is downstream) instead of the TCP throughput per session.

  (5) TCP throughput per session from the origin server to the mirror server when content is dynamically mirrored from the origin server to a mirror server.

  (6) In the above (5), a usable bandwidth is used instead of the TCP throughput per session.

  Hereinafter, some implementation examples of the dynamic mirroring destination determination algorithm using the above (1) to (6) will be described in detail with reference to FIGS.

  First, an example in which the dynamic mirroring destination is determined by the algorithm in the case of using the above (1) will be described below with reference to FIG. This is algorithm 1.

  The origin server F1 monitors the number of content acquisition requests per unit time from each domain in which the client exists. Here, AS is given as an example of the domain. Hereinafter, the case where the domain is AS will be described as an example.

  Now, it is assumed that the origin server performs dynamic mirroring of contents held by several mirror servers in order to reduce its own load. Here, it is assumed that the number of mirror servers to be subjected to dynamic mirroring is given.

  First, the dynamic mirroring destination determination unit A13a extracts an AS whose content acquisition request number per unit time exceeds a threshold value from all ASs that have clients that make content acquisition requests to the origin server (step in FIG. 12). S401).

  Next, for each extracted AS (step S402), one mirror server in the AS is selected (step S403). Here, as a method of detecting a mirror server in the AS and selecting one of them, the representative management server that centrally manages the mirror servers in each AS is inquired to obtain the most CPU load. A method of selecting a low value is conceivable.

  For each extracted AS, the TCP throughput per session from the origin server to the client in the AS is measured, and this is defined as TCPBW1 (step S404). Further, the TCP throughput per session from the mirror server selected in the AS to the client in the AS is measured, and this is set as TCPBW2 (step S405). Here, the client in the AS may be a number of arbitrarily selected terminals, a fixed measurement point, or the like. Next, X = N × (TCPBW2 / TCPBW1) is calculated by the network resource base calculation unit A130 (step S406). Here, N is the number of content acquisition requests per unit time from clients existing in the AS.

  The above X is calculated for all the extracted ASs (step S407), and the mirror servers in the AS having the largest value of X are selected in advance as the targets for dynamic mirroring in a given number (step S407). S408, S409).

  An example of an algorithm in the case of using the above (2) to (4) can be considered similarly. A detailed procedure can be considered in the same manner as in Algorithm 1, and is omitted here.

  Next, an example in which the dynamic mirroring destination is determined by the algorithm in the case of using the above (6) will be described below with reference to FIG. This is algorithm 2.

  Here, it is assumed that there are a plurality of candidate mirror servers that are content transfer destinations when dynamic mirroring is performed, and dynamic mirroring is performed only on a limited number of mirror servers. That is, the number of dynamic mirroring is limited. Such a situation is sufficient when the cost of dynamic mirroring per one time (for example, disk usage) is given and the total cost is limited. In this example, it is assumed that all contents are directly transferred from the origin server.

  First, for each mirror server that is a candidate for a dynamic mirroring destination (step S501), the available bandwidth from the origin server to the mirror server is measured or acquired, and this is set as AvBW (step S502). Further, the network resource base calculation means A130 calculates S / AvBW using the size S of the content that is to be dynamically mirrored for the mirror server (step S503). This is the expected content transfer delay during dynamic mirroring. After the above calculation is performed for all candidate mirror servers (step S504), dynamic mirroring is performed up to the limit number in order from the smallest S / AvBW value (steps S505 and S506).

  Next, an example in which a dynamic mirroring destination is determined by an algorithm using a combination of the network resources (1) and (6) will be described below with reference to FIG. This is algorithm 3.

  Here, it is assumed that there are a plurality of candidate mirror servers that are destinations of content when performing dynamic mirroring, and dynamic mirroring is performed only on a limited number of mirror servers. First, TCPBW2 / TCPBW1 and AvBW / S are calculated for each candidate mirror server in the same manner as in Algorithms 1 and 2 (step S601 in FIG. 14). Here, each parameter acquisition method and calculation method are as described in Algorithms 1 and 3, respectively.

After the calculation is completed for all candidate mirror servers, the reciprocal of the minimum value of TCPBW2 / TCPBW1 is k1, and the reciprocal of the minimum value of AvBW / S is k2 (step S602). Next, for all candidate mirror servers,
Y = α * k1 * TCPBW2 / TCPBW1 + (1-α) * k2 * AvBW / S
Is calculated (step S603). Here, α is a predetermined coefficient of 0 ≦ α ≦ 1. In order from the largest Y, the dynamic mirroring is performed up to the limit number (steps S604 and S605).

  Further, consider another example of the algorithm that uses a combination of the network resources of (1) and (6). This is algorithm 4.

  Here, it is assumed that there are a plurality of candidate mirror servers that are destinations of content when performing dynamic mirroring, and dynamic mirroring is performed only on a limited number of mirror servers.

  First, the selection range is narrowed down to candidate mirror servers whose values of TCPBW2 / TCPBW1 are equal to or greater than a threshold value among all candidate mirror servers. Then, dynamic mirroring is performed from the narrowed candidate mirror servers up to the limited number of times in ascending order of the S / AvBW value by algorithm 3.

  Next, the effect of this embodiment will be described.

  In the present embodiment, the performance improvement effect associated with dynamic mirroring can be enhanced by determining the dynamic mirroring destination in consideration of the resource status in the network during dynamic mirroring. In each of the algorithms 1 to 4 described above, specific performance improvement effects are shown below.

  In algorithm 1, dynamic mirroring is preferentially performed on a mirror server in which X = N × (TCPBW2 / TCPBW1) becomes large. It can be considered that the larger X is, the more the content acquisition delay is improved by the dynamic mirroring to the mirror server in the AS and the number of content acquisition requests that can benefit from the improvement effect. Therefore, in addition to the load reduction effect of the origin server, it is possible to improve the performance at the time of client content acquisition.

  In algorithm 2, the one with a small mirror content transfer delay from the origin server to the mirror server during dynamic mirroring is preferentially selected as the mirroring destination. Therefore, it is possible to complete the transfer of the mirror content accompanying dynamic mirroring at high speed, and to immediately reduce the load on the origin server. In addition, since a link with a low available bandwidth is not used as much as possible for transferring mirror content, the possibility of hindering other traffic flowing on such a link with a low available bandwidth by a large amount of mirror content transfer is reduced. With the above points, the performance of the content dynamic mirroring system can be improved.

  In algorithm 3, TCPBW2 / TCPBW1 in algorithm 1 and AvBW / S in algorithm 2 are each normalized by a minimum value, and an arbitrary weight is added to each of them to compare and select a mirroring destination. Therefore, when both the improvement effect of the content acquisition delay in the client and the effect of the algorithm 2 are desired, the optimum dynamic mirroring destination can be selected by performing weighting according to the respective importance levels.

  In the algorithm 4, first, candidate mirror servers are narrowed down only to a certain level that has an effect of improving the content acquisition delay in the client. Among them, the mirror content transfer delay from the origin server to the mirror server is preferentially selected. Select the mirror destination. Therefore, it is possible to obtain the effect of the algorithm 2 while maintaining the improvement effect of the content acquisition delay in the client at a certain value or more, and both the performance improvement effect on the client side and the performance improvement effect on the origin server side can be achieved.

  Next, a fifth embodiment of the present invention will be described in detail with reference to the drawings.

  Referring to FIG. 15, the present embodiment is different in that the origin server G1 has a mirroring tree creation means G11 in addition to the configuration of the origin server F1 in the fourth embodiment shown in FIG. Further, referring to FIG. 16, the dynamic mirroring request accepting means H11 in the adaptive mirror node in the adaptive mirror node H1 has mirror contents in addition to the configuration of the dynamic mirroring request accepting means B11 in the adaptive mirror node B1 in FIG. The difference is that the relay means H111 is provided. The adaptive mirror node H1 corresponds to the adaptive mirror node B1, E1, or the adaptive mirror proxy server D1, in which the dynamic mirroring request receiving unit B1 is replaced with the dynamic mirroring request receiving unit H1.

  The mirroring tree creation means G11 calculates a mirroring tree that is the optimum path for transferring mirror contents to the dynamic mirroring destination determined by the dynamic mirroring destination determination means A13.

  The mirror content relay unit H111 relays the mirror content transferred from the origin server G1 upstream of the mirroring tree or another adaptive mirror node H1, and transfers the mirror content to the adaptive mirror node H1 downstream of the mirroring tree. It has the function to do. At this time, if the adaptive mirror node H1 that relays the mirror content needs to be a mirror server of the mirror content, the mirror content is simultaneously stored in the storage device at the time of relay.

  In the following description, it is assumed that the target of dynamic mirroring from the origin server G1 is the adaptive mirror node H1, but in the present embodiment, the mirror content relay means in the conventional fixed mirror server It may be a mirror server having the same function as that realized by H111.

  Next, the operation of the present embodiment will be described in detail with reference to the flowchart of FIG. 17 and FIG.

  The dynamic mirroring destination determination means A13 in the origin server A1 determines one or a plurality of adaptive mirror nodes that are the dynamic mirroring destinations of the mirror content (step S701 in FIG. 17). An example of the determination algorithm at this time is the one described in the fourth embodiment of the present invention.

  Next, the mirroring tree creation unit G11 calculates a mirroring tree that is a transfer path of mirror contents for the adaptive mirror node that is the dynamic mirroring destination determined in step S701 (step S702). The mirroring tree created here is composed only of adaptive mirror nodes that are dynamic mirroring destinations. To create a mirroring tree, the resource information of the origin server and adaptive mirror node itself that is a component of the mirroring tree, and the resource information in the transfer path of the mirror content between the origin server and adaptive mirror node that is a component of the mirroring tree A method using either one or both of the above is conceivable.

  An example of the calculation algorithm of the mirroring tree in step S702 will be described with reference to FIG. The origin server 113 wants to transfer the mirror content to both the adaptive mirror nodes 114 and 115. Here, the mirroring tree is calculated by the Dijkstra algorithm using the reciprocal of the available bandwidth in the transfer path between the origin server and the adaptive mirror node and between the adaptive mirror nodes as the dynamic mirroring destination as a cost. The mirroring tree created by this algorithm includes only the origin server and the adaptive mirror node as its components. The transfer path by the mirroring tree maximizes the available bandwidth at the time of mirror content transfer. The available bandwidth in each transfer path is shown in Table 303. As a result of the calculation, a mirroring tree that has the origin server 113 as the apex, passes through the adaptive mirror node 114 via the transfer path 217 (113 → 114), and further reaches the adaptive mirror node 115 via the transfer path 220 (114 → 115). 401 is created.

  Then, it is determined whether it is necessary to additionally select the adaptive mirror node for relay (step S703). Examples of this criterion include: a. “In the mirroring tree created in step S702, the available bandwidth that is a bottleneck is equal to or less than a threshold value”, or b. “There exists an adaptive mirror node for relaying that can reduce the total amount of traffic in mirror content transfer”. The number of relay adaptive mirror nodes to be selected may be limited. The reason for this is that selecting a large number of adaptive mirror nodes for relay not only increases the calculation cost of the mirroring tree performed in step S705, but also increases the total traffic in mirror content transfer by passing through a large number of adaptive mirror nodes for relay. This is because the amount tends to increase.

  If it is determined in step S703 that an additional adaptive mirror node for relay needs to be selected, an optimal adaptive mirror node for relay is selected from the adaptive mirror nodes arranged in the network (step S704). . If it is determined in step S703 that it is not necessary to additionally select the adaptive mirror node for relay, the mirroring tree calculated in step 702 is determined as the mirror content transfer path.

  An example of the relay adaptive mirror node selection algorithm in step S704 will be described. Here, it is assumed that the criterion a in step S703 is used for the mirroring tree 401 in FIG. 18, and the threshold of the usable bandwidth that becomes a bottleneck is 100 kbps. Since the available bandwidth that becomes the bottleneck link (transfer path 217) of the mirroring tree 401 is 80 kbps, it is determined that an adaptive mirror node for relay needs to be additionally selected. Further, it is assumed that the number of relay adaptive mirror nodes to be selected is limited to one. In the example of this algorithm, a transfer path whose usable bandwidth is equal to or less than a threshold value is bypassed through the relay adaptive mirror node via one hop.

  This adaptive mirror node for relay exists in a domain adjacent to the domain (for example, AS in the Internet) in which the upstream origin server or adaptive mirror node exists in the transfer path whose usable bandwidth is equal to or less than the threshold. The mirror node is selected to maximize the available bandwidth when relaying. Here, the relay adaptive mirror node that maximizes the available bandwidth from the origin server 113 to the adaptive mirror node 114 is selected. However, if there is no relay adaptive mirror node whose usable bandwidth is larger than the transfer path 217 (113 → 114), the relay adaptive mirror node is not selected. Here, it is assumed that the adaptive mirror node 116 that relays between the origin server 113 and the adaptive mirror node 114 via the transfer paths 219 (113 → 116) and 222 (116 → 114) is selected as the adaptive mirror node for relay. To do. When the adaptive mirror node 116 is used as an adaptive mirror node for relay, the available bandwidth between the origin server and the adaptive mirror node is 150 kbps.

  After the relay adaptive mirror node is selected in step S704, a mirroring tree is created again (step S705). In the mirroring tree created here, the selected relay mirror node is included in its constituent elements in addition to the adaptive mirror node as the dynamic mirroring destination. The calculation algorithm of the mirroring tree used here may be the same as or different from the calculation algorithm used in step 702. As an example of a different algorithm, in the mirroring tree calculated in step S702, a mirroring tree is created in which the transfer path portion in which the available bandwidth is a bottleneck is replaced with a transfer path that passes through the relay adaptive mirror node. and so on. If it is determined in step S703 that an adaptive mirror node for relay needs to be additionally selected, the mirroring tree calculated in step S705 is determined as a mirror content transfer path.

  In the example of FIG. 18, the transfer path 217 (113 → 114) between the origin server 113 and the adaptive mirror node 114 forwards the adaptive mirror node 116 with respect to the mirroring tree 401, and 219 (113 → 116). , 222 (116 → 114), and the mirroring tree 402 is determined as the mirror content transfer path.

  Next, a mirror content transfer request is transmitted to each adaptive mirror node on the mirroring tree determined as the mirror content transfer path (step S706). Here, the information included in the mirror content transfer request includes the mirroring tree path information in addition to the information described in step S106 of FIG. 3, and further includes the mirror content of the mirror content for the relay adaptive mirror node. Information indicating whether mirroring processing is requested or relay processing is requested is included.

  When the origin server A1 receives the acceptance response from all the adaptive mirror nodes to which the mirror content transfer request is transmitted in step S706, the mirror content is transferred onto the mirroring tree (steps S707 and S708).

  In the example of FIG. 18, when mirror content is transferred onto the mirroring tree 402, the origin server transmits the mirror content to the adaptive mirror node 116 via the transfer path 219 (113 → 116). The adaptive mirror node 116 relays the received mirror content and transmits it to the adaptive mirror node 114 via the transfer path 222 (116 → 114). Here, the adaptive mirror node 116 may store the mirror content once received in the storage device, and may transmit to the adaptive mirror node 114 after receiving all the mirror content, or relay in real time without storing it. Also good. The latter is preferable in terms of reducing transfer delay of mirror contents. The adaptive mirror node 114 stores the received mirror content in the storage device, and at the same time, relays to the adaptive mirror node 115 via the transfer path 220 (114 → 115). The adaptive mirror node 115 stores the received mirror content in the storage device, and when reception of the mirror content is completed, the adaptive mirror node 115 transmits a message indicating that all transfer of the mirror content is completed to the origin server 113.

  Here, the case where the mirror content is transferred onto the mirroring tree 402 has been described, but the case where the mirroring tree is in the form of a multicast tree will also be briefly described with reference to FIG. Here, it is assumed that the origin server 113 transfers mirror content to the adaptive mirror nodes 114, 115, and 117 using the mirroring tree 403. First, the origin server 113 transmits mirror content to the adaptive mirror node 114 via the transfer path 217 (113 → 114). The adaptive mirror node 114 stores the received mirror content in the storage device, and at the same time, the transfer path 220 (114 → 115) and the transfer path 223 to the adaptive mirror nodes 115 and 117, which are downstream nodes in the mirroring tree 403, respectively. The mirror content is relayed via (114 → 117).

  When a rejection response is received from one or more adaptive mirror nodes among all the adaptive mirror nodes to which the mirror content transfer request is transmitted in step S706, the process returns to step S701, and the adaptive mirror nodes that are candidates for the dynamic mirroring destination are selected. Reselection is made (step S707). In addition, a method may be considered in which only the adaptive mirror node that has received a rejection response is removed and the mirroring tree is recalculated in step S703.

  In the present embodiment described above, the mirroring tree is created with the origin server as the apex, but in addition to the origin server, other mirror servers that have already mirrored the same content as the mirror content May be the vertex. When dynamic mirroring is performed simultaneously on multiple adaptive mirror nodes, the origin server or mirror server that is the top of the mirroring tree is selected for each adaptive mirror node so that the transfer delay of the mirror content is minimized, and thus selected. Alternatively, a plurality of mirroring trees having the origin server or the mirror server as vertices may be created and transferred.

  Next, the effect of this embodiment will be described.

  Conventionally, when dynamic mirroring is performed from an origin server to one or more mirror servers, the mirror content is transferred directly from the origin server to the mirror server, so a large amount of traffic is involved in the mirror content transfer. As a result, there is a risk of applying a load to the net. In particular, if there is a link with a very small available bandwidth value on the transfer path from the origin server to the mirror server, not only will the transfer delay required for mirroring increase, but a large amount of traffic during mirror content transfer will May interfere with other traffic passing through the link. Also, when the same content is dynamically mirrored to a plurality of mirror servers at the same time, the same mirror content is transferred in parallel to each mirror server, so the load of transfer traffic on the network is large.

  In the present embodiment, it is possible to select a transfer path that uses another adaptive mirror node as a relay hop in the mirror content transfer path from the origin server to the adaptive mirror node that is the dynamic mirroring destination. Therefore, the mirror content transfer delay can be reduced and the mirror content can be transferred without disturbing other traffic on the transfer path.

In particular, when performing dynamic mirroring simultaneously on a plurality of mirroring destinations, the origin server creates a mirroring tree with an adaptive mirror node as a mirroring destination, and the mirror content is transferred along the created mirroring tree. At this time, other adaptive mirror nodes other than the adaptive mirror node to be mirrored may be added to the mirroring tree for relay. Therefore, in the transfer of mirror content to a plurality of mirroring destinations, it is possible to reduce the transfer delay and the amount of transfer traffic.
(Example)
Next, a first embodiment of the present invention will be described with reference to the drawings. This first example corresponds to the first and second embodiments of the present invention.

  Referring to FIG. 19, the present embodiment is composed of four ASs AS 2 to 5, and has intra-AS networks 6 to 9, respectively. The ASs are connected by links 213 to 216. An origin server 105 exists in AS2, and adaptive mirror proxy servers 106-108 exist in AS3-5, respectively. Each AS has clients 109 to 112, respectively. Here, not a single client machine but a group of clients existing in each AS is shown.

  In the present embodiment, for the sake of simplicity, only contents (hereinafter referred to as “news contents”) whose URLs begin with “http://www.biglobe.com/news/” are subject to dynamic mirroring. And

  The adaptive mirror proxy servers 106 to 108 are set as proxy servers for the clients 109 to 112 in each AS, respectively. All content acquisition requests (HTTP requests) transmitted by each client are once received by the proxy server in each AS, and if the content corresponding to the request is cached, the content is directly transmitted to the client. If the content corresponding to the request is not cached, the request is transmitted to the origin server as a proxy.

  The origin server 105 counts the number of requests of new content per unit time for each AS in which the client exists, and when the value exceeds a threshold, the new content is dynamically mirrored to the AS. Here, the threshold value is set to 50000 requests / hour.

  Since the number of requests from the client 111 already exceeds the threshold value in the adaptive mirror proxy server 107, it is assumed that the news content has already been dynamically mirrored and is operating as a mirror server. All requests for news content in the client 111 are transmitted to the adaptive mirror proxy server 107. The adaptive mirror proxy server 107 identifies that the request is for mirror content and responds directly without sending the request to the origin server 105.

  Now, it is assumed that the number of requests for the new contents from the client 110 in the origin server 105 exceeds the threshold value of 50000 requests / hour.

  The origin server 105 searches for an adaptive mirror node in the AS 3 and tries to perform dynamic mirroring to an appropriate adaptive mirror node.

  Here, the HTTP request sent from the adaptive mirror proxy server 106 to the proxy of the client 110 to the origin server 105 includes information indicating that it has the function of the adaptive mirror node in addition to the proxy server function. Shall. Based on the information, the origin server 105 recognizes that the adaptive mirror proxy server 106 exists in the AS 3, and decides to perform dynamic mirroring of news content on the adaptive mirror proxy server 106.

  The origin server 105 transmits a dynamic content mirroring request to the adaptive mirror proxy server 106. The adaptive mirror proxy server 106 receives this request and transmits an acceptance response to the origin server 105 if it can be accepted.

  Next, the origin server 105 checks the available bandwidth from each of the other adaptive mirror nodes that already have news content to the adaptive mirror proxy server 106. If there is a larger bandwidth than the available bandwidth from the origin server 105, the mirror content is transmitted from the adaptive mirror node with the largest available bandwidth.

  Here, the available bandwidth from the origin server 105 to the adaptive mirror proxy server 106 is compared with the available bandwidth from the adaptive mirror proxy server 107 to the adaptive mirror proxy server 106, and the latter is assumed to be larger.

  The origin server 105 transmits an instruction to the adaptive mirror proxy server 107 to transmit the news content to the adaptive mirror proxy server 106.

  When the adaptive mirror proxy server 107 receives this instruction, it transmits the news content to the adaptive mirror proxy server 106.

  When the adaptive mirror proxy server 106 receives all the new contents, the adaptive mirror proxy server 106 functions as a mirror server for the new contents, and an HTTP request for the new contents received from the client 110 is directly returned.

  Next, a second embodiment will be described with reference to the drawings. The second example corresponds to the fourth embodiment, and particularly corresponds to the case where the dynamic mirroring destination is determined using the algorithm 1. In the present embodiment, the adaptive mirror proxy servers 106 to 108 may be read as normal mirror servers.

  In this embodiment, the network configuration shown in FIG. 19 is also used. Here, it is assumed that there is news content in the origin server 105, and the news content is not mirrored in any of the adaptive mirror proxy servers 106-108.

  The origin server 105 counts the number of requests per unit time of the news content for each AS where the client exists, and when the value exceeds a threshold, the new content can be dynamically mirrored to the mirror server in the AS. it can. Here, the threshold value is set to 50000 requests / hour.

  In AS3-5, there are clients 110-112 which are clients corresponding to each, but now the number of requests N from clients 110-112 is 100,000 requests / hour, 200,000 requests / hour, 300,000 requests, respectively. / Hours, and all exceed the threshold value of 50000 requests / hour. In each AS, the origin server 105 recognizes that the adaptive mirror proxy servers 106 to 108 exist as mirror servers, as in the first embodiment. Therefore, the origin server 105 can perform dynamic mirroring of news contents for each of the adaptive mirror proxy servers 106 to 108 existing in each AS. Here, the use of the disk space in the adaptive mirror proxy server is used. For this reason, it is assumed that only two dynamic mirroring destinations are restricted. Therefore, the number is reduced to two using the algorithm 4 in the fourth embodiment.

  First, for all AS, TCPBW1, which is the TCP throughput per session from the origin server 105 to the client in the AS, and from the adaptive mirror proxy server in the AS to the client in the AS TCPBW2, which is the TCP throughput per session, is measured. Here, it is assumed that the measurement point set in the AS is used for the client in the AS. The measurement results are shown in Table 302 in FIG.

  From the values shown in Table 302 and the number of requests per unit time from each AS, X = N × (TCPBW2 / TCPBW1) is calculated for AS3 to 5, respectively. As a result, X = 1.5 million, 4 million, and 3 million, respectively, for AS3-5. The value of X increases in the order of AS4, 5, 3. Therefore, the adaptive mirror proxy servers 107 and 108 are determined as the dynamic mirroring destinations of the news content.

  The content dynamic mirroring system described above is not only realized in hardware, but is also implemented in software by a program for realizing each function described above, which is recorded on a magnetic disk, semiconductor memory or other recording medium of a computer system. It can also be realized. In the content dynamic mirroring system according to each of the embodiments described above, the dynamic mirroring programs P10 and P20 of the origin server A1 and the adaptive mirror node B1 are read from the recording medium into the data processing unit (CPU) of the computer system, and data processing is performed. By controlling the operation of the unit, the functions of the respective means described above are realized and dynamic mirroring is executed. That is, in the origin server and the adaptive mirror node, the processes of FIGS. 3, 5, 7, 9, 12, 12, 13, and 17 are executed.

  Although the present invention has been described with reference to the preferred embodiments and examples, the present invention is not necessarily limited to the above-described embodiments and examples, and various modifications can be made within the scope of the technical idea. Can be implemented.

It is a block diagram which shows the structure of the 1st Embodiment of this invention. It is a block diagram which shows in detail the structure of the dynamic mirroring request | requirement means in the 1st Embodiment of this invention. It is a flowchart which shows operation | movement of the origin server of the 1st Embodiment of this invention. It is a block diagram which shows the detailed structure of the mirror content transfer means of the origin server of the 1st Embodiment of this invention. It is a flowchart explaining the detailed operation | movement of the system which performs dynamic mirroring from the mirror server with the smallest transfer delay of the 1st Embodiment of this invention. It is a figure which shows the example of the system which performs dynamic mirroring from the mirror server with the least transfer overhead in the 1st Embodiment of this invention. It is a flowchart which shows operation | movement of the mirror server of the 1st Embodiment of this invention. It is a block diagram which shows the structure of the 2nd Embodiment of this invention. It is a flowchart which shows operation | movement of the adaptive mirror proxy server of the 2nd Embodiment of this invention. It is a block diagram which shows the structure of the 3rd Embodiment of this invention. It is a block diagram which shows the structure of the 4th Embodiment of this invention. It is a flowchart which shows the operation | movement of the algorithm 1 in the 4th Embodiment of this invention. It is a flowchart which shows the operation | movement of the algorithm 2 in the 4th Embodiment of this invention. It is a flowchart which shows the operation | movement of the algorithm 3 in the 4th Embodiment of this invention. It is a block diagram which shows the structure of the origin server of the 5th Embodiment of this invention. It is a block diagram which shows the structure of the adaptive mirror node of the 5th Embodiment of this invention. It is a flowchart which shows operation | movement of the origin server of the 5th Embodiment of this invention. It is a figure which shows the example of creation of the mirroring tree in the 5th Embodiment of this invention. It is a network block diagram in the 1st Example to which this invention is applied. It is a figure which shows the measurement result of the TCP throughput per session in the 2nd Example to which this invention is applied.

Explanation of symbols

A1 Origin server A10 Storage device A11 Server load measurement means A12 Mirror content determination means A13 Dynamic mirroring destination determination means A14 Adaptive mirror node detection means A15 Dynamic mirroring request means A16 Mirror content transfer means A161 Communication means with other mirror servers A162 Source Determination means A163 Communication means with adaptive mirror node B1 Mirror server B10 Storage device B11 Dynamic mirroring request reception means B12 Request data reception section B13 Response content creation means B14 Content response means B101 Mirror content storage section B111 Data transmission / reception means B112 Dynamic mirroring Request acceptance determination means B113 Mirror content reception means B114 Decompression / decryption means C1 Client D1 Adaptive mirror proxy server D 0 storage device D11 request data reception unit D12 operation determination unit D13 cache hit determination unit D14 content acquisition request transfer destination determination unit D15 content acquisition request transfer unit D101 cache storage unit E1 adaptive mirror node E10 storage device E11 dynamic mirroring request reception unit E12 Request data receiving unit E13 Authentication billing information processing unit E101 Authentication billing information management unit F1 Origin server F11 Network resource acquisition unit G1 Origin server G11 Mirroring tree creation unit H1 Adaptive mirror node H11 Dynamic mirroring request reception unit H111 Mirror content relay unit 1 Backbone Network 2-5 AS
6-9 AS intra-network 101, 105 Origin server 106-108 Adaptive mirror proxy server 109-112 Client 113 Origin server 114-117 Adaptive mirror node 201-216 Link 217-223 Transfer path 301 Indicates available bandwidth between servers Table 302 Table showing TCP throughput between each server / client 303 Table showing available bandwidth in each transfer path 401 to 403 Mirroring tree

Claims (9)

An origin server that performs dynamic mirroring to copy mirror content that is a copy of all or part of the content held by its own node to another mirror server in the network,
Server load measurement means for measuring a parameter indicating performance degradation of the origin server accompanying a content acquisition request from a client, and determining whether to perform the dynamic mirroring based on the measurement result;
Mirror content determination means for selecting which content of the content held by the origin server is to be used as the mirror content and performing the dynamic mirroring;
Network resource acquisition means having a function of acquiring either or both of the network resource information and the mirror server performance information;
Dynamic mirroring destination determining means for determining which mirror server to copy the mirror content according to a predetermined algorithm based on the information acquired by the network resource acquiring means;
Dynamic mirroring requesting means for issuing a dynamic mirroring request for performing the dynamic mirroring to the node to be the dynamic mirroring destination determined by the dynamic mirroring destination determining means;
Mirror content transfer means for transferring the mirror content to the node determined as a dynamic mirroring destination using the mirroring tree as a transfer path when the dynamic mirroring request is received;
Mirroring tree creating means having a function of calculating the mirroring tree,
The mirroring tree created by the mirroring tree creation means is:
In addition to the mirror server to which the mirror content is copied,
An origin server comprising a mirror server selected as a relay hop as a component of the mirroring tree. An origin server that performs dynamic mirroring to copy mirror content that is a copy of all or part of the content held by its own node to another mirror server in the network,
Server load measurement means for measuring a parameter indicating performance degradation of the origin server accompanying a content acquisition request from a client, and determining whether to perform the dynamic mirroring based on the measurement result;
Mirror content determination means for selecting which content of the content held by the origin server is to be used as the mirror content and performing the dynamic mirroring;
Network resource acquisition means having a function of acquiring either or both of the network resource information and the mirror server performance information;
Dynamic mirroring destination determining means for determining which mirror server to copy the mirror content according to a predetermined algorithm based on the information acquired by the network resource acquiring means;
Dynamic mirroring requesting means for issuing a dynamic mirroring request for performing the dynamic mirroring to the node to be the dynamic mirroring destination determined by the dynamic mirroring destination determining means;
Mirror content transfer means for transferring the mirror content to the node determined as a dynamic mirroring destination using the mirroring tree as a transfer path when the dynamic mirroring request is received;
Mirroring tree creating means having a function of calculating the mirroring tree,
The mirroring tree creation means includes:
In a mirroring tree composed only of the origin server and the mirror server to which the mirror content is copied,
The transfer path in which the available bandwidth of the transfer path between the origin server and the mirror server that is a component of the mirroring tree is a predetermined threshold or less,
An origin server that re-creates a mirroring tree that is replaced with a transfer path that relays using a mirror server other than the constituent elements of the mirroring tree. The mirroring tree creation means includes:
Resource information of the origin server and the mirror server that are constituent elements of the mirroring tree to be created;
The origin according to claim 1 or 2, wherein the mirroring tree is calculated using either or both of resource information in a transfer path of the mirror content between the origin server and the mirror server. server. The dynamic mirroring destination determination means determines the dynamic mirroring destination,
Measured for each client site, which is a site to which a client making a content acquisition request to the origin server, obtained by the network resource acquisition means,
A first content transfer throughput per session from the origin server to the client site;
The second content transfer throughput per session from the mirror server selected to minimize the content acquisition delay for the client site to the client site is used. The origin server according to any one of the above. The dynamic mirroring destination determination means determines the dynamic mirroring destination,
5. The higher priority is given to a larger value of the number of content acquisition requests from the client site per unit time × (the second content throughput / the first content throughput). Origin server. The dynamic mirroring destination determination means determines the dynamic mirroring destination,
From the mirror server, which is a candidate for a dynamic mirroring destination, measured for each client site to which the client making a content acquisition request to the origin server belongs, obtained by the network resource acquisition unit, from the mirror server The larger the available bandwidth of the transfer path to the site is, or the RTT from the mirror server that is a candidate for the dynamic mirroring destination to the client site, or the client site of certain fixed-length data becomes downstream. The origin server according to any one of claims 1 to 3, wherein a higher priority is given to a smaller value of any one of the direction transfer delays. The dynamic mirroring destination determination means determines the dynamic mirroring destination,
The higher priority is given to the content transfer throughput per session from the origin server to the mirror server, or as the value of the usable bandwidth of the transfer path is larger. The origin server according to item 1. A node of a dynamic mirroring system that performs dynamic mirroring that copies mirror content that is a copy of all or part of content held by an origin server to another mirror server in the network,
When transferring the mirror content from the origin server to one or more mirror servers,
With the origin server as a vertex, the mirror content is transferred according to a mirroring tree that is an arbitrary tree-like path constituted by the mirror server,
In the transfer of the mirror content from the origin server using the mirroring tree as a transfer path,
Relay the mirror content received from the origin server or the mirror server upstream of the mirroring tree;
Mirror content relay means for transferring to the mirror server downstream of the mirroring tree;
Based on an instruction from the origin server, whether to store the mirror content in the mirror server when relaying the mirror content received from the origin server or the mirror server upstream of the mirroring tree A node that is selectively performed. A node of a dynamic mirroring system that performs dynamic mirroring that copies mirror content that is a copy of all or part of content held by an origin server to another mirror server in the network,
When transferring the mirror content from the origin server to one or more of the mirror servers, the mirror content of the mirror content is determined according to a mirroring tree that is an arbitrary tree-like path constituted by the mirror server, with the origin server serving as a vertex. Received the transfer
In the transfer of the mirror content from the origin server using the mirroring tree as a transfer path,
Relay the mirror content received from the origin server or the mirror server upstream of the mirroring tree;
Mirror content relay means for transferring to the mirror server downstream of the mirroring tree;
When relaying the mirror content received from the origin server or the mirror server upstream of the mirroring tree, the operation of whether or not to store the mirror content in the mirror server is autonomous based on the state of the own node. A node that can be selected dynamically.

Images