JP2004264956A - Method for managing cache, and cache server capable of using this method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To increase the efficiency of a proxy cache for an entire network. <P>SOLUTION: The metrix accumulating part 65 of an upper cache server 20 receives a report on the number of times that an object to be deleted at a lower cache server 22 is referred to, and adds to the number of times that the metrix management entry 55 of a cache buffer 52 is referred to. Based on the cumulative number of times that the object to be deleted is referred to, an evaluating part 66 determines whether or not the object is to be cached. Based on the determination result, a cache movement control part 62 receives data on the object from the lower cache server 22. A cache management part 60 stores the received data on the object in a cache object storage part 58 as a new cache object 56, and newly provides a cache entry 54 for management. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a cache technology, and more particularly to a cache management method in a network and a cache server that can use the method.
[0002]
[Prior art]
When providing WWW (World Wide Web) services on the Internet, there are limitations on the processing capacity of servers and routers, the bandwidth of communication lines, and the like. Therefore, a proxy cache server is provided on the network, and a client requests an access to a Web object. In general, the processing performance is secured by acting as a proxy and temporarily storing the accessed object. When a client accesses a server and acquires an object, the proxy cache server stores the cache data of the object. When a client makes a request for the same object again from the client, the proxy cache server sends the cache data to the client instead of sending an access request to the server. As a result, the load on the server or the router can be reduced, the network traffic can be reduced, the utilization efficiency of the network resources can be improved, and the response time of the service can be improved.
[0003]
In order to increase the use efficiency of the proxy cache, a large-scale recording device having a sufficient storage capacity for caching objects is required. With the widespread use of broadband networks, the use of large-volume content such as video and music is becoming more and more popular. Therefore, simply installing additional recording devices in the proxy server is not sufficient, and more efficient operation of the proxy cache is required. .
[0004]
The performance of a proxy cache is largely dependent on the cache entry replacement policy. To reserve memory space for caching a new object, the already cached object must be deleted. A replacement policy evaluates the metrics of a cache object in some way to determine which cache object is more efficient to delete. To improve the efficiency of the proxy cache, various replacement policies using metrics such as the last access time, access frequency, and object size have been studied (for example, see Non-Patent Document 1).
[0005]
[Non-patent document 1]
G. FIG. Barish and K.S. Obraczka, "World Wide Web caching: Trends and techniques," IEEE Commun. Mag. , Vol. 38, no. 5, pp. 178-184, May 2000.
[0006]
[Problems to be solved by the invention]
The proposed proxy cache replacement policy operates on a local proxy cache within a single proxy cache server, and the proxy cache server determines that its cache object is no longer useful. At that point, the cache object is immediately deleted from the cache memory. However, the utility value of the cache object should be evaluated in the entire network. To reduce object access time and reduce network traffic, object caches need to be properly distributed over the network, and relying solely on local replacement policies within individual proxy cache servers, Proxy cache performance cannot be fully exploited.
[0007]
The present invention has been made in view of such a situation, and an object of the present invention is to provide a cache management method capable of improving the use efficiency of a cache object and a cache server capable of using the method.
[0008]
[Means for Solving the Problems]
One embodiment of the present invention relates to a cache management method. The method includes, in a hierarchically structured network of cache servers, cumulatively evaluating metrics of objects cached in a plurality of lower-level cache servers in at least one upper-level cache server; When a cache object is deleted in one, it is determined whether the cache object to be deleted is transferred from a lower-level cache server to a higher-level cache server and cached according to the accumulated evaluated metrics. Here, on the network path of the object provided from the server to the client, the client side is set to the lower side and the server side is set to the upper side, and the hierarchical position of the cache server is determined by the relative positional relationship of the cache server on the network path. Can be determined.
[0009]
Only when the lower-level cache server deletes the cache object, the metrics of the cache object may be accumulated in the upper-level cache server. As a result, metrics can be accumulated in the upper-level cache server before the object is deleted in the lower-level cache server. Also, by limiting the accumulation of metrics to caches deleted by lower-level cache servers, it is possible to avoid the accumulation of metrics for a large number of objects cached by lower-level cache servers, and to reduce the traffic between cache servers. Can be minimized. In addition, since metrics are accumulated only at the timing when the cache object is deleted in the cache server in the lower hierarchy, it is possible to prevent the omission of accumulation of metrics and the occurrence of duplicate addition.
[0010]
In the cache server of the higher hierarchy, the cache object may be managed according to a replacement policy based on the metrics evaluated cumulatively. By using the accumulated metrics, it is possible to implement a cache replacement policy in consideration of the access state to the object among a plurality of clients to which the object is provided via the higher-level cache server.
[0011]
Another embodiment of the present invention relates to a cache server. This cache server is a cache server that is relatively lower in the hierarchy in a hierarchically structured cache server network, and includes a cache storage unit that caches and holds an object provided to a client; A cache management unit that manages cache objects in accordance with a replacement policy based on metrics; a metrics notification unit that notifies a cache server in an upper layer only of metrics of cache objects deleted in accordance with the replacement policy; A transfer control unit that, when a transmission request is issued from the cache server, moves the cache object to be deleted to the cache server in the upper hierarchy.
[0012]
Yet another embodiment of the present invention also relates to a cache server. This cache server is a cache server that is relatively higher in a hierarchically configured network of cache servers, and includes a cache storage unit that caches and holds objects provided to clients, and a lower layer. A metrics accumulator that accumulates metrics of cache objects notified from a plurality of cache servers and calculates total metrics for each object; and a cache object held in the cache storage unit according to a replacement policy based on the total metrics. A cache management unit that manages a cache object to be deleted, and when a notification of a cache object to be deleted is received from a lower-level cache server, the cache object to be deleted based on the total metrics. An evaluator for deciding whether to retain the object, and when the evaluator decides to retain the cached object to be deleted, the cached object to be deleted is acquired from a cache server in a lower hierarchy, and A movement control unit for causing the cache management unit to newly manage the object.
[0013]
The metric accumulation unit may receive a notification of the metric of the cache object when the lower-level cache server deletes the cache object according to the replacement policy of the cache server, and accumulate the metric in the total metric.
[0014]
Yet another embodiment of the present invention relates to a computer program. The program includes a step of accumulating metrics of a cache object notified from each of a plurality of lower-level cache servers when a cache object is deleted and calculating a total metric for each object; Receiving the notification of the cache object to be deleted, determining whether or not to retain the cache object to be deleted based on the total metrics. Obtaining the deleted cache object from a cache server; and managing the obtained cache object according to a replacement policy based on the total metrics. Cause the computer to execute.
[0015]
It is to be noted that any combination of the above-described components and any conversion of the expression of the present invention between a method, an apparatus, a server, a system, a computer program, a recording medium, and the like are also effective as aspects of the present invention.
[0016]
BEST MODE FOR CARRYING OUT THE INVENTION
FIG. 1 is a configuration diagram of an entire network including a cache server according to an embodiment. Clients 10a to 10c (hereinafter, collectively referred to as client 10) are provided via the Internet via Web pages and images provided by servers 14a and 14b (hereinafter, collectively referred to as server 14). Access objects such as audio.
[0017]
The plurality of cache servers 12a to 12e arranged on the Internet have a function as a proxy server for proxying an object access request from the client 10 to the server 14, and relay the object data acquired from the server 14 to the client 10 And caches the object's data temporarily for future use.
[0018]
These cache servers 12a to 12e form a hierarchical structure on a network path from the servers 14a and 14b to the clients 10a to 10c, and cache data of the objects # 1 to # 4 accessed by the clients 10a to 10c. For example, when the client 10a sends an access request for the object # 2 to the server 14a, first, the cache server 12b located closest to the client 10a substitutes for the access request for the object # 2 and ranks higher than the cache server 12b. The object # 2 is requested to a certain cache server 12a. That is, the lower cache server 12b and the upper cache server 12a have a client-server relationship.
[0019]
Further, if there is a cache of the object # 2, the upper cache server 12a returns the cache data to the lower cache server 12b. Otherwise, the upper cache server 12a itself becomes the client 14a as a client. To the object # 2. The upper cache server 12a receives the data of the object # 2 from the server 14a, caches it as the cache object 16a, and returns the data to the lower cache server 12b.
[0020]
The lower cache server 12b receives the data of the object # 2 from the upper cache server 12a, caches the data as the cache object 16b, and returns the data to the client 10a.
[0021]
Thereafter, when another client 10c sends an access request for the same object # 2 to the server 14a, the cache server 12e in the first hierarchy closest to the client 10c substitutes for the access request for the object # 2, and The object # 2 is requested to the cache server 12c in the upper hierarchy. If there is no cache for the object # 2, the cache server 12c in the second hierarchy sends an access request for the object # 2 to the cache server 12a in the next higher hierarchy. Since the third-tier cache server 12a already has the cache object 16a of the object # 2, the data is returned to the second-tier cache server 12c. The second-tier cache server 12c caches the data of the object # 2 as the cache object 16c, and returns the data to the first-tier cache server 12e. The cache server 12e of the first hierarchy caches the received data of the object # 2 as the cache object 16e and returns the data to the client 10c.
[0022]
In this way, the cache object 16e cached in the first-tier cache server 12e is managed according to the cache replacement policy in the cache server 12e, and is replaced and deleted when another object is newly cached. Sometimes. Each of the cache servers 12a to 12e manages a cache object according to a replacement policy based on metrics to be described below.
[0023]
FIG. 2 is a diagram for explaining a method of evaluating cache object metrics by the cache server according to the embodiment. In the example shown in the figure, the object access requests of the clients 10a to 10c to the servers 14a and 14b are represented by the lower cache servers 22a to 22c, respectively, and the lower cache servers 22a to 22c request the object access to the upper cache server 20. Is made. The upper cache server 20 receives and caches the objects from the servers 14a and 14b, operates as a server for the lower cache servers 22a to 22c, and returns the objects to the lower cache servers 22a to 22c. In the figure, the numbers attached to the right shoulders of the caches of the objects # 1 to # 4 indicate the number of local references of the respective cache objects in the cache servers 20, 22a to 22c. 22c manage cache objects in accordance with a replacement policy using the number of times of reference as a metric.
[0024]
Although the two lower cache servers 22a and 22b store the cache objects 18a and 18b of the object # 2, respectively, the number of references to the cache objects 18a and 18b of the object # 2 is four times, and the other cache objects Less than the number of references. However, when viewed from the whole network, the number of references to the cache of the object # 2 is eight in total, which is larger than the number of references to other cache objects.
[0025]
Here, it is assumed that the two lower cache servers 22a and 22b delete the cache objects 18a and 18b with a small number of references, respectively, in order to secure a space for caching a new object. The upper cache server 20 receives the notification of the number of references to the cache objects 18a and 18b of the object # 2 from each of the two lower cache servers 22a and 22b, accumulates the number of references, and the total number of references is eight. Is stored as the total metrics of object # 2.
[0026]
Since the upper cache server 20 manages the cache object 19 of the object # 2 according to the replacement policy based on the total metrics, the cache objects 18a and 18b of the object # 2 are evaluated locally by the lower cache servers 22a and 22b. , The upper level cache server 20 holds the cache object 19 of the object # 2. However, in the upper cache server 20, if there is another object having a large number of references, the cache object 19 of the object # 2 may be deleted soon. In this way, the cache of the object # 2 is gradually deleted from the lower tier to the upper tier cache server, and the cache of the object # 2 is just generated from the upper tier to the lower tier cache server. Contrast with propagation.
[0027]
Similarly, when the cache object 17c of the object # 3 is deleted according to the replacement policy in the lower-level cache server 22c acting on behalf of another client 10c, the reference number of the cache object 17c is six times, and the other client 10a The reference count of the cache object 17a of the object # 3 stored in the lower cache server 22a acting as a proxy is accumulated five times in the upper cache server 20, and it is assumed that the total metric has a reference count of 11 times. It is managed in the server 20.
[0028]
FIG. 3 is a configuration diagram of the lower cache server 22, and FIG. 4 is a configuration diagram of the upper cache server 20. These configurations can be realized in hardware by a CPU, a memory, or other LSI of any computer, and are realized in software by a program having a cache management function loaded in the memory. Here, the functional blocks realized by their cooperation are drawn. Therefore, it is understood by those skilled in the art that these functional blocks can be realized in various forms by hardware only, software only, or a combination thereof.
[0029]
The configuration of the lower cache server 22 will be described with reference to FIG. The lower cache server 22 is provided with a cache buffer 32, and stores entries 34 of objects to be cached (hereinafter, simply referred to as cache entries 34) in a list structure or the like. The cache entry 34 includes a pointer to the cache data 36 of the object (hereinafter, simply referred to as the cache object 36), and the cache object 36 itself is stored in the cache object storage unit 38 which is a large-capacity storage device.
[0030]
FIG. 5 shows the data structure of the cache entry 34. The cache entry 34 includes a URL 70 indicating a storage location of the object in the server 14, a reference count 72 of the cache in the lower cache server 22, a cache last access time 74, a cache access frequency 76, a cache object size 78, And a pointer 80 to the cache object.
[0031]
Referring back to FIG. 3, the search unit 30 searches the list structure of the cache entry 34 of the cache buffer 32 for the object requested by the client 10, and checks whether the object is cached. The URL 70 in FIG. 5 is used as a search key for an object. When a hit occurs in the cache, the search unit 30 retrieves the cache object 36 indicated by the pointer 80 of the hit cache entry 34 from the cache object storage unit 38 and transmits it to the client 10.
[0032]
The cache management unit 40 manages the cache entries 34 stored in the cache buffer 32 in accordance with the replacement policy based on the metrics with reference to the metrics assigned to each cache entry 34 stored in the cache buffer 32. . For example, the reference count 72 in FIG. 5 is used as the metric, and an LFU (Least Frequently Used) scheme for flushing the cache entry 34 with the smallest reference count 72 first is used as the replacement policy.
[0033]
When space for caching a new object is required in the cache object storage unit 38, the cache management unit 40 selects the cache entry 34 having the smallest metric and deletes the cache entry 36 from the cache object storage unit 38. To determine. The cache management unit 40 reads the metrics of the cache object 36 to be deleted from the cache buffer 32 and supplies the metrics to the metrics notification unit 44.
[0034]
The metrics notification unit 44 notifies the higher-level cache server 20 of the metrics of the cache object 36 to be deleted. The cache migration control unit 42 reads the cache object 36 from the cache object storage unit 38 and transfers it to the upper cache server 20 only when receiving a transmission request of the cache object 36 to be deleted from the upper cache server 20.
[0035]
The cache management unit 40 deletes the cache entry 34 of the cache object 36 from the cache buffer 32 in response to the completion of the transfer control of the cache object 36 to be deleted by the cache movement control unit 42, and caches the cache object 36. It is deleted from the object storage unit 38.
[0036]
The configuration of the upper cache server 20 will be described with reference to FIG. The search unit 50, the cache management unit 60, and the cache object storage unit 58 have the same configuration as the search unit 30, the cache management unit 40, and the cache object storage unit 38 of the lower cache server 22 in FIG. There are differences in the entries that are entered.
[0037]
The cache buffer 52 has a metric management entry 55 in addition to the cache entry 54 provided for the cache object 56 in the cache object storage unit 58. The metric management entry 55 is used for cumulative calculation of metrics of objects cached by the lower cache server 22, and the cache object storage unit 58 does not store cache data of the corresponding object. .
[0038]
The cache entry 54 has the same data structure as the cache entry 34 of the lower cache server 22 described with reference to FIG. 5, but the reference count 72 is obtained by cumulatively adding the reference count of the cache object notified from the lower cache server 22. They differ in that they are used as metrics.
[0039]
FIG. 6 shows the data structure of the metrics management entry 55. The metrics management entry 55 stores the URL 82 of the object, the cumulative reference count 84, and the object size 86. When the object specified by the URL 82 of the metric management entry 55 is to be cached also in the upper cache server 20, the metric management entry 55 is replaced with the cache entry 34 described in FIG. Thus, the cache entry 34 is managed in the cache buffer 52 according to the replacement policy.
[0040]
Therefore, the object cached in the lower cache server 22 is managed in the cache buffer 52 of the upper cache server 20 as the cache entry 54, the metric management entry 55, or both. Either not applicable. When managed as the cache entry 54, the cache object 56 is cached in the cache object storage unit 58. However, while managed as the metrics management entry 55, the cache object 56 corresponds to the cache object storage unit 58. No cache data exists.
[0041]
Referring to FIG. 4 again, when the search unit 50 searches for an object requested from the lower-level cache server 22, only the cache entry 54 in the cache buffer 52 is searched. And retrieves the corresponding cache object 56 from the subordinate cache server 22. Further, the cache management unit 60 manages only the cache entry 54 in the cache buffer 52 according to the replacement policy based on the total metrics.
[0042]
When receiving the notification of the metric of the object to be deleted from the lower-level cache server 22, if the object is cached, the metric accumulation unit 65 accumulatively adds the notified metric to the reference count 72 of the cache entry 54. If the object is not cached, the metric notified to the cumulative reference count 84 of the metric management entry 55 is cumulatively added.
[0043]
The evaluation unit 66 receives the notification of the object to be deleted from the lower-level cache server 22, searches the cache entry 54 or the metrics management entry 55 in the cache buffer 52 using the URL of the object as a search key, and Gets the cumulative reference count of. If the object to be deleted exists as the metric management entry 55, since the data of the object is not cached in the upper cache server 20, a new cache entry is created based on the cumulative reference count of the object to be deleted. It is determined whether or not to cache as 54. For example, if the cumulative reference count is larger than the reference count of the other cache entry 54, the object to be deleted is accepted, otherwise, it is not accepted. If there is enough free space in the cache object storage unit 58, the object to be deleted is accepted without evaluating the cumulative reference count.
[0044]
The evaluation unit 66 notifies the cache migration control unit 62 of the determination result. When the cache migration control unit 62 receives the object to be deleted, it sends a transmission request for the object to the lower cache server 22 and the lower cache server 22 Receive object data from.
[0045]
The object data received by the cache movement control unit 62 is stored as a new cache object 56 in the cache object storage unit 58, and the cache management unit 60 replaces the metrics management entry 55 for the cache object 56 with the cache entry 54. In the cache buffer 52 according to the replacement policy.
[0046]
FIGS. 7 to 9 are flowcharts for explaining a cache management procedure performed by the lower cache server 22 and the upper cache server 20 having the above configurations.
[0047]
In FIG. 7, the cache management unit 40 of the lower cache server 22 determines the cache object 36 to be deleted based on the replacement policy (S10). The cache management unit 40 temporarily locks the cache object to be deleted (S12). The metrics notification unit 44 notifies the upper cache server 20 of the URL 70 of the cache entry 34 corresponding to the cache object 36 to be deleted and the reference count 72, which is a local metric (S14). For this notification, an ICP (Internet Cache Protocol) used for communication between the cache servers configured in a hierarchy is used, and the lower-order cache server 22 uses the URL 70 and the reference count 72 in the ICP_MOVE_QUERY message obtained by expanding the standard ICP_QUERY. Is transmitted to the host cache server 20 (S16).
[0048]
Upon receiving the ICP_MOVE_QUERY message, the metrics accumulation unit 65 of the upper cache server 20 searches the cache entry 54 or the metrics management entry 55 in the cache buffer 52 based on the URL 70 included in the ICP_MOVE_QUERY message, and The reference count 72, which is the local metric notified from the lower-level cache server 22, is cumulatively added to the reference count 72 or the cumulative reference count 84 of the metric management entry 55 (S18). If there is no corresponding entry in the cache buffer 52, a new metric management entry 55 is created based on the URL 70 and the reference count 72 notified from the lower cache server 22.
[0049]
When the object to be deleted notified from the lower cache server 22 is not cached, that is, the entry of the object to be deleted is stored in the cache buffer 52 as the metric management entry 55, If there is, based on the accumulated reference count 84 of the metric management entry 55, it is determined whether the object to be deleted is accepted and newly cached (S20). FIG. 8 shows the flow of processing when an object is received (Y in S20), and FIG. 9 shows the flow of processing when an object is not received (N in S20).
[0050]
With reference to FIG. 8, a process in a case where the upper cache server 20 receives the object to be deleted in the lower cache server 22 will be described. The cache migration control unit 62 of the upper cache server 20 requests the lower cache server 22 to transmit the object to be deleted (S22). This transmission request is made by an HTTP (Hypertext Transport Protocol) request using the URL 70 of the object to be deleted notified from the lower cache server 22 (S24). The cache migration control unit 42 of the lower cache server 22 transfers the cache object 36 to be deleted to the upper cache server 20 (S26). This transfer is performed by HTTP reply (S28). The cache management unit 40 of the lower cache server 22 unlocks the cache object 36 to be deleted, deletes the data from the cache object storage unit 38, and deletes the corresponding cache entry 34 from the cache buffer 32 (S34). ).
[0051]
The cache migration control unit 62 of the upper cache server 20 receives the object data transferred by the HTTP reply (S30). The cache migration control unit 62 stores the received data of the object as a new cache object 56 in the cache object storage unit 58, and the cache management unit 60 stores the metrics management entry 55 corresponding to the object in the new cache entry 54. And is managed in the cache buffer 52 (S32).
[0052]
With reference to FIG. 9, a description will be given of a process in a case where the upper cache server 20 does not accept an object to be deleted in the lower cache server 22. The cache migration control unit 62 of the upper cache server 20 transmits an ICP_MOVE_NG message to the lower cache server 22 to notify that the object to be deleted cannot be received (S36). The cache management unit 40 of the lower cache server 22 unlocks the cache object 36 to be deleted, deletes the data from the cache object storage unit 38, and deletes the corresponding cache entry 34 from the cache buffer 32 (S38). ).
[0053]
As described above, according to the embodiment of the present invention, the metric of the cache locally evaluated by the lower-level proxy cache is integrated and reevaluated by the upper-level proxy cache. It is possible to realize a cache replacement policy that takes into account. Thereby, the hit rate of the cache can be increased, and the response time at the time of accessing the object can be improved.
[0054]
Further, according to the embodiment, the cache of objects frequently used by clients in a relatively limited area is provided in a lower-level proxy cache close to the client in the area, and is used from a set of clients in a relatively large area. The cache of the object to be moved can be moved to a higher-level proxy cache, so that the network load can be distributed by appropriately distributing the cache.
[0055]
In addition, even if the lower-level proxy caches belong to different management domains and it is difficult to cooperate between the proxy caches, providing a higher-level proxy cache across the management domains enables the upper-level proxy caches. It will be possible to perform an integrated reevaluation of metrics.
[0056]
The present invention has been described based on the embodiments. These embodiments are exemplifications, and it is understood by those skilled in the art that various modifications can be made to the combination of each component and each processing process, and such modifications are also within the scope of the present invention. By the way. Hereinafter, such a modified example will be described.
[0057]
In the above description, the reference count is used as the metric indicating the utility value of the cache object, and the LFU method is adopted as the replacement policy. However, the present invention can naturally be implemented using metrics other than the reference count. . Further, a replacement policy may be realized by combining other metrics with the number of times of reference. For example, the last access time and access frequency are evaluated as metrics in addition to the number of references, and the cache entry with the longest elapsed time since the last reference is preferentially deleted, or the cache life is considered based on the access frequency. You may do so. Alternatively, it may be determined whether to delete the cache based on the object size. For example, a cache having a large object size takes a long access time, so that it may not be deleted as much as possible, or may be deleted preferentially to save storage space.
[0058]
Also, in the lower cache server, the LRU (Least Recently Used) method using only the last access time as a metric may be adopted as the replacement policy only by counting the number of times of reference. In this case, the lower cache server notifies the upper cache server of the reference count counted for the cache replaced by the LRU method, and the upper cache server manages the cache with a replacement policy based on the accumulated reference count. May be. As described above, the replacement policy may be different between the upper and lower cache servers.
[0059]
In the above description, the configuration and operation are described separately for the upper cache server and the lower cache server. However, if there is a higher cache server with respect to the upper cache server, the higher cache server The server operates as a lower cache server. Therefore, the upper cache server may have the configuration of the lower cache server so that the functions of any cache server can be relatively executed.
[0060]
In a hierarchically configured proxy cache network, a lower cache server generally corresponds to a child proxy cache, and an upper cache server corresponds to a parent proxy cache. However, a child proxy cache belonging to a different parent proxy cache may be set as a lower cache server, and an upper cache server may be provided to control them. As described above, the upper and lower cache servers may be configured without being limited to the parent-child relationship in the proxy cache network.
[0061]
【The invention's effect】
According to the present invention, a cache can be efficiently managed on a network.
[Brief description of the drawings]
FIG. 1 is a configuration diagram of an entire network including a cache server according to an embodiment.
FIG. 2 is a diagram illustrating a method of evaluating cache object metrics by a cache server according to an embodiment.
FIG. 3 is a configuration diagram of a lower cache server of FIG. 2;
FIG. 4 is a configuration diagram of an upper cache server of FIG. 2;
FIG. 5 is an explanatory diagram of a data structure of a cache entry in FIG. 3;
FIG. 6 is an explanatory diagram of a data structure of a metric management entry of FIG. 4;
FIG. 7 is a flowchart illustrating a cache management procedure by the lower cache server and the upper cache server in FIG. 2;
8 is a flowchart illustrating a process when an upper-level cache server in FIG. 2 receives an object to be deleted in the lower-level cache server in FIG. 2;
9 is a flowchart illustrating a process when an object deleted in the lower cache server in FIG. 2 is not accepted by the upper cache server in FIG. 2;
[Explanation of symbols]
10 clients, 12 cache servers, 14 servers, 20 upper cache servers, 22 lower cache servers, 30, 50 search units, 32, 52 cache buffers, 34, 54 cache entries, 55 metrics management entries, 36, 56 cache objects, 38, 58 cache object storage unit, 40, 60 cache management unit, 42, 62 cache movement control unit, 44 metrics notification unit, 65 metrics accumulation unit, 66 evaluation unit.

Claims (8)

In a hierarchically structured network of cache servers, metrics of objects cached in a plurality of lower-level cache servers are cumulatively evaluated in at least one upper-level cache server, and cached in one of the lower-level cache servers. When an object is deleted, it is determined whether the cache object to be deleted is transferred from a lower-level cache server to a higher-level cache server and cached according to the accumulated metrics. Cache management method. 2. The cache management method according to claim 1, wherein the metrics of the cache object are accumulated in the upper-level cache server only when the lower-level cache server deletes the cache object. 3. The cache management method according to claim 1, wherein a cache object is managed in a cache server of a higher hierarchy according to a replacement policy based on the accumulated evaluated metrics. A cache storage unit that caches and holds an object provided to a client;
A cache management unit that manages cache objects according to a replacement policy based on predetermined metrics,
Only a metric of the cache object deleted according to the replacement policy, a metric notifying unit that notifies a cache server of a higher hierarchy,
A cache control unit that, when a transmission request is issued from the upper-layer cache server, moves a cache object to be deleted to the upper-layer cache server. A cache storage unit that caches and holds an object provided to a client;
A metric accumulating unit that accumulates metrics of cache objects notified from a plurality of cache servers in a lower hierarchy and calculates a total metric for each object;
A cache management unit that manages a cache object held in the cache storage unit according to a replacement policy based on the total metrics;
An evaluator configured to determine whether to retain the deleted cache object based on the total metrics when a notification of the cache object to be deleted is received from the lower-level cache server,
When the evaluator determines that the cache object to be deleted is to be retained, the migration control that obtains the cache object to be deleted from a lower-level cache server and newly manages the cache object by the cache management unit And a cache server. The metric accumulation unit, when a lower-level cache server deletes a cache object in accordance with a replacement policy of the cache server, receives a notification of the metric of the cache object and accumulates the metric in the total metric. Item 6. The cache server according to item 5. The cache server according to claim 4, wherein the metric is a reference count of a cache object. Calculating a total metric for each object by accumulating metrics of the cache object notified at the time of deleting the cache object from each of the plurality of cache servers in the lower hierarchy;
Receiving a notification of a cache object to be deleted from a lower tier cache server, and determining whether to retain the deleted cache object based on the total metrics;
Obtaining the deleted cache object from a lower-level cache server if it is determined to retain the deleted cache object;
And a step of managing the acquired cache object in accordance with a replacement policy based on the total metrics.

Images