CN103888517B - Caching method for publishing and subscription system to achieve historical event subscription - Google Patents

Abstract

The invention discloses a caching method for realizing historical event subscription for a publishing and subscribing system, comprising the following steps: a routing table expansion step, adding caching routing information in the routing table; an event processing step, calculating a caching point reference through a hash function value; event publishing step: judge whether the current proxy node is used as a cache point and cache according to the cache point reference value, and update the cache routing information synchronously; cache subscription step; and cache acquisition step. The present invention has the following beneficial effects: the caching method provided by the present invention enables subscribers to subscribe to historical events; the present invention fully utilizes the storage space of all agents and their neighbor agents on the event distribution path, improves the utilization rate of storage resources, and enhances The caching capability of the publish-subscribe system is improved; the invention simplifies the caching redundancy and balances the memory access efficiency and the storage occupancy rate.

Description

Cache method for implementing historical event subscription for publish-subscribe system

technical field

The invention relates to a publish-subscribe system technology, in particular to a caching method for realizing historical event subscription.

Background technique

A publish-subscribe system consists of a distributed communication network, the event broker network, and a series of endpoints communicating with each other, namely subscribers and publishers. Subscribers and publishers are interconnected through an event proxy network, and an asynchronous publish-subscribe method is used to complete data exchange. Subscribers submit subscriptions by declaring a filter condition expression. When a publisher publishes an event that meets its subscription conditions, the event is routed to the subscriber through the proxy network. In terms of time, subscription and publication of content are asynchronous, and in terms of space, subscribers and publishers do not need to care about where each other is located. Therefore, the publish-subscribe system realizes the complete decoupling of the communication parties in terms of space, time and control flow .

In a traditional publish-subscribe system, event notifications are guaranteed to eventually reach each interested subscriber, but only if each subscriber is online and their subscriptions are known to the entire system. This prerequisite is a major limitation in dynamic network environments, since clients tend to join and leave the network frequently, and newly joined subscribers may request an event that was published before they joined the network. In the traditional publish-subscribe system, this kind of request cannot be responded, so how to use the event cache mechanism to enable new users to learn about historical events has become a major problem that the publish-subscribe system needs to solve.

Contents of the invention

In order to overcome the defect in the prior art that subscribers cannot subscribe to historical events, the present invention provides a caching method for realizing historical event subscription. Concrete technical scheme of the present invention is as follows:

A caching method for implementing historical event subscription for a publish-subscribe system, comprising the following steps:

Routing table expansion step: adding cached routing information in the routing table of each proxy node, the cached routing information is used to provide routing information that can be located to the cache point;

Event processing steps: when a proxy node sends out an event, calculate it through a hash function to obtain a cache point reference value; add a message header to the event, and record the cache point information and subscription path in the message header information; the cache point information includes the cache point reference value;

Event publishing step: according to the subscription path information recorded in the message header, publish the event to the corresponding subscription node; for each proxy node in the publishing path, judge whether the current proxy node is As a cache point and cache, at the same time, the current proxy node synchronously updates the cache routing information corresponding to the event;

Cache subscription step: a certain proxy node sends a cache request, and the cache request includes a subscription condition; the subscription condition is matched with the pre-stored filter conditions in the routing table of the current proxy node to determine a cache event; the cache event Refers to the event corresponding to the filter condition that successfully matches the subscription condition;

Cache acquisition step: If the routing table of the current proxy node contains the cache routing information corresponding to the cache event, then directly find the cache point corresponding to the cache event according to the cache routing information, forward the cache request to the cache point, and obtain the cache ; If the routing table of the current proxy node does not contain the cache routing information corresponding to the cache event, then according to the subscription routing information corresponding to the cache event, send the cache request to the next proxy node, and perform the cache acquisition step in turn.

As an optimization scheme, the hash function is as shown in formula (1):

The domain of the hash function is the value range (a, b) of the event e, and the value range is [0, MAX_HOP]; where MAX_HOP is the maximum distance from the agent node on the event distribution path to its root node.

The calculated cache point reference value is an integer value.

As an optimization scheme, according to the cache point reference value, it is judged whether the current proxy node is used as a cache point and the cache is specifically as follows:

After receiving the event, the current proxy node firstly subtracts 1 from the cache point reference value t, and then judges according to the cache point reference value. If t=0, the current proxy node is used as the cache point.

As an optimization scheme, the steps of expanding the routing table are specifically:

Add the cache_nexthops field in the routing table of each proxy node, so that the structure of the routing table is:

<filter|sub_nexthops|cache_nexthops>

Among them, filter refers to the filter condition, sub_nexthops records the next-hop agent that forwards the events that successfully match the filter to the direction of the subscription node, and cache_nexthops records the next-hop agent that forwards the cache request that successfully matches the filter to the direction of the cache point.

As an optimization scheme, in the event publishing step, for the current agent node in the publishing process of the event, the update method of its cached routing information is specifically:

If the current proxy node is behind the cache point, the next hop in the cache routing information is set as the predecessor proxy node;

If the current proxy node happens to be a cache point, the next hop in the cache routing information is set to the current proxy node itself;

If the current proxy node is in front of the cache point, the next hop in the cache routing information is set as the subsequent proxy node.

As an optimization scheme, in the cache obtaining step, the method of forwarding the cache request to the cache point to obtain the cache is as follows:

In the process of forwarding the cache request to the cache point, record the proxy nodes along the way in sequence, which is recorded as the cache request path; after the cache request reaches the cache point, the cache point sends the requested cache along the cache request path in reverse order to the proxy that issued the cache request node.

As an optimization scheme, in the event publishing step, if the current proxy node is used as a cache point and caches, it is judged: if the storage space of the current proxy node is exhausted, then ask whether the direct neighbors outside the event publishing path still have Available storage space; if the neighbor has available storage space, the event is forwarded to the neighbor for caching.

As an optimization solution, the cache point information also includes the cache point location, the number of cached copies, and the cache timestamp, and the initial values of the cache point location, the number of cached copies, and the cache timestamp are all empty; for each Proxy node, if the current proxy node is used as a cache point and caches, update the cache point location, cache copy number and cache timestamp recorded in the event message header, and update the cache point location and cache copy number stored in the current proxy node and cache timestamps.

As an optimization scheme, during the publishing process of the event, compare the cache timestamps of the proxy nodes along the way to obtain the location of the oldest cache. Go to the location of the oldest cache to delete the oldest cache.

As an optimization solution, in the event publishing step, for each proxy node, a redundant streamlining method is used to cache the event, and the redundant streamlining method is specifically:

Let the proxy node be b _i , which contains several child nodes b _ij , and let g _ij be the subtree corresponding to each child node b _ij ; according to the objective function Determines which subtrees of cache points to dispatch cached copies to, as shown in equation (2):

Among them, d _ij represents the maximum distance between the proxy node b _t and the end user of the subtree g _ij , r _ij represents the total number of user requests received by the proxy node b _t from the subtree g _ij , s _ij represents the total number of all nodes in the subtree g _ij The storage idle rate of , t is the reference value of the cache point; the objective function The numerator of represents the total cost for the user to obtain a copy of the event when it is placed on the subtree g _ij ; among them, the first item of the numerator is the cost when the user at the end of the subtree g _ij requests the copy, and the first item of the numerator The binomial represents the rest of the subtrees The total overhead when users on the network request this copy;

The proxy node b _t calculates a goal for each child node b _ij Select m=min{k, n _i } sub-trees g _t1 , g _t2 , ..., g _tm with the smallest value to assign copies, and correspondingly send copies to child nodes b _t1 , b _t2 , ..., b The replica number k of the event of _tm is modified according to formula (3):

Set the replica number k of events sent to other child nodes b _tt as k _i =0 (m<l≤n _t ), and return the corresponding t value to 0.

Compared with the prior art, the present invention has the following beneficial effects:

(1) The caching method provided by the present invention realizes the subscriber's subscription to historical events;

(2) The present invention makes full use of the storage space of all agents and their neighbor agents on the event distribution path, improves the utilization rate of storage resources, and enhances the caching capability of the publish-subscribe system;

(3) The present invention simplifies the buffer redundancy and balances the memory access efficiency and the storage occupancy rate.

Description of drawings

Fig. 1 is the general flowchart of the cache method provided by the present invention;

FIG. 2 is a schematic network diagram of the publish-subscribe system in Embodiment 1.

detailed description

The present invention will be described in detail below by means of embodiments in conjunction with the accompanying drawings.

Example 1:

As shown in Figure 1, a caching method for implementing historical event subscription for a publish-subscribe system includes the following steps:

Step S1, routing table expansion step: adding cached routing information in the routing table of each proxy node, the cached routing information is used to provide routing information that can locate the cached point.

In the traditional publish-subscribe system, the routing table of each proxy node already contains subscription routing information, which is used to locate the subscribers corresponding to the event; The table is extended.

The existing routing table structure is <filter|sub_nexthops>, where filter is the filter condition, and sub_nexthops records the next-hop proxy node that forwards the event that matches the filter successfully to the direction of the subscription node; the sub_nexthops field corresponds to the subscription routing information .

Add the cache_nexthops field in the existing routing table structure, so that the structure of the routing table is:

<filter|sub_nexthops|cache_nexthops>

Among them, cache_nexthops records the next-hop agent that the cache request that successfully matches the filter is forwarded to the cache point, and the cache_nexthops field corresponds to the cache routing information. The cache_nexthops field is constructed hop-by-hop along the distribution path during the event publishing process, and its function is to provide routing information for the user's cache request so that it can find the cache point, just as sub_nexthops helps the event find subscribers.

Step S2, event processing step: when a proxy node sends out an event, it calculates through a hash function to obtain a cache point reference value; attaches a message header to the event, and the cache point information is recorded in the message header and subscription path information; the cache point information includes the cache point reference value.

This step is for the publisher to implement caching before sending events to subscribers. Assuming that the value range of event e is e, value ∈ (a, b), the publisher decides which nodes on the path to cache the event through the following hash function, as shown in formula (1):

Wherein, the definition domain of the hash function is the value range (a, b) of the event e, and the value range is [0, MAX_HOP], where MAX_HOP is the maximum distance from the agent node on the event distribution path to its root node ( can be obtained by sending a discovery message in advance), counted in hops. When there is no distribution path for event e, it is mapped to the 0th hop, which is the root node.

A cache point reference value is calculated, and the cache point reference value is an integer value. The result of hash mapping is that the reference value of the cache point is not the proxy identifier of the cache point, but a hop variable t similar to a subtraction counter, which records the distance between the cache point of the event and the current proxy node. After the hash operation is performed, it is not immediately possible to determine which proxy node is the cache point. During the event publishing process, each proxy node on the publishing path needs to determine whether it is a cache point according to the reference value of the cache point. Record the calculated cache point reference value in the message header of the event and publish it together with the event. The subscription path information records the path information that the event needs to pass through from the publisher to the subscriber.

Step S3, event publishing step: according to the subscription path information recorded in the message header, publish the event to the corresponding subscription node; for each proxy node in the publishing path, judge the current Whether the proxy node acts as a cache point and caches, and at the same time, the current proxy node synchronously updates the cache routing information corresponding to the event.

In this step, according to the cache point reference value, it is judged whether the current proxy node is used as a cache point and the cache is specifically as follows:

After the current proxy node receives the event, it first subtracts 1 from the cache point reference value t, and then judges according to its value. If t=0, the current proxy node is used as the cache point.

When caching, the storage space of the current proxy node may be full. In order to make full use of the storage space in the network, the following judgment should be made before caching: if the storage space of the current proxy node is exhausted, ask the Whether the direct neighbor outside the event publishing path still has available storage space; if the neighbor has available storage space, the event will be forwarded to the neighbor for caching. Adopting this method can improve the utilization rate of the storage space of the whole system.

After the current proxy node completes the judgment of the cache point, it needs to synchronously update the cache routing information corresponding to the event in the routing table of the current proxy node. The update method of the cached routing information is specifically, including three situations: and correspondingly construct the cached route:

Case 1: If t<0, that is, the forwarding of the event has passed the cache point, and the current proxy node is located downstream of the cache point, then set the next hop in the cache routing information as the predecessor proxy node;

Case 2: if t=0, that is, the current proxy node is just a cache point, then the next hop in the cache routing information is set as the current proxy node itself;

Case 3: If t>0, that is, the forwarding of the event has not yet reached the cache point, and the current proxy node is in front of the cache point, then the next hop in the cache routing information is set as the subsequent proxy node.

In the process of publishing the event, each proxy node performs the operation of judging the cache point and updating the cache routing information. When the event is sent to the subscriber, all cache points have been determined and cached, and all proxy nodes in the publishing path have updated the cache routing information.

Step S4, cache subscription step: a proxy node sends a cache request, and the cache request includes a subscription condition; matches the subscription condition with the filter condition pre-stored in the routing table of the current proxy node, and determines the cache event; The above cached events refer to the events corresponding to the filter conditions that successfully match the subscription conditions.

In this step, after the subscriber accesses the publish-subscribe system, if he wants to subscribe to the historical event, he must subscribe to the required content from the cached event. Similar to the subscription request sent when subscribing to events, subscribing to caches is achieved by sending a cache request. Let the cache request be Requset(Δf), where Δf is the subscription condition.

Step S5, cache acquisition step: if the routing table of the current proxy node contains the cache routing information corresponding to the cache event, then directly find the cache point corresponding to the cache event according to the cache routing information, and forward the cache request to the cache point , to obtain the cache; if the routing table of the current proxy node does not contain the cache routing information corresponding to the cache event, then according to the subscription routing information corresponding to the cache event, the cache request is sent to the next proxy node, and the cache acquisition is performed in turn step.

Specifically, in this step, obtaining the cache includes two situations:

(1) The routing table contains cached routing information corresponding to cached events, that is, users who subscribe to historical events appear on the distribution path of their interested events. At this time, the cache request can directly find the event cache point according to cache_nexthops. In the process of forwarding the cache request to the cache point hop by hop, the proxy nodes along the way are recorded in sequence, which is recorded as the cache request path; after the cache request reaches the cache point, the cache point sends the requested cache to the issuing cache along the reverse order of the cache request path The requested proxy node.

(2) The routing table does not contain the cached routing information corresponding to the cached event, that is, the user subscribing to the historical event is not on the distribution path of the event of interest. According to the sub_nexthops corresponding to the cache event, the cache request is sent to the next proxy node. The next proxy node repeats the above operations until an event cache point is found.

An example is given below to describe the caching method provided in this embodiment in detail. Figure 2 shows a network schematic diagram of the publish-subscribe system, where A-E represent five proxy nodes, P represents the proxy node where the publisher is located, and S ₁ Represents the proxy node where the user (subscriber) resides.

Assuming that the filter condition of user S ₁ is f ₁ , the publisher P publishes an event e ₁ that meets the filter condition f ₁ and selects agent D as the cache point, and e ₁ carries the cache point information along the distribution path A→B→D→ E is matched and forwarded by the publisher to the subscriber hop by hop. Proxy A successfully matches the event with f ₁ , reads the cache point information in the event message header and finds that its cache point is located downstream of itself, then adds the successfully matched subscription route next hop node (proxy B) to the cache_nexthops field, indicating Future received cache lookup requests will be forwarded to B. B will also perform similar operations and use the successor node D as the next hop for cache lookup. After receiving the event, D finds that it is selected as the cache point, then stores the event locally and adds itself to the next hop of the cache route. Proxy E knows that e ₁ is cached upstream in the event distribution path according to the cache point information in the event message header, so it uses the predecessor node D as the next hop of the cache route. Thus, the construction of the cached routing information is completed hop by hop.

Suppose a new user S ₂ connects to proxy E and requests event e _1. Since new user S ₂ appears on the publishing path of event e ₁ , E searches the routing table and matches f ₁ successfully, and forwards the request to the cache according to cache_nexthops The next hop (agent D) to find the event cache is the case (1) mentioned in step S5.

If S ₂ appears at agent C, since the new user S ₂ is not on the release path of event e ₁ , C does not have cache routing information (NIL) in the successfully matched f ₁ entry, then it will cache the request according to sub_nexthops Forward to the subscriber, that is, forward to agent B, and B happens to be on the distribution path of e ₁ , and the request can be forwarded to D according to the cache route cache_nexthops, so that the cache is finally found, that is, the first step mentioned in step S5 (2) Situations. During the forwarding process of the Request message, the IDs of the nodes along the route will be recorded sequentially. When the user's cache request finally reaches the proxy D, it contains a node link: S ₂ →C→B, which is the cache request path. Proxy D encapsulates the event content requested by the user and this link into a response message Response(e ₁ , S ₂ →C→B), and sends it back to the user along the reverse direction of the link to complete the cache request response.

Example 2:

The difference between this embodiment and Embodiment 1 is that the cache point information in the event message header also includes the cache point location, the number of cache copies, and the cache timestamp, and each event cache corresponds to a cache timestamp.

Before the event is released, the initial values of the location of the cache point, the number of cache copies, and the cache timestamp are all empty. During the event publishing process, for each proxy node in the publishing path, if the current proxy node is used as a cache point and cached, update the cache point location, the number of cache copies, and the cache timestamp recorded in the message header of the event, and update The location of the cache point, the number of cache copies, and the cache timestamp stored in the current proxy node.

In practical applications, the storage space in the publish-subscribe network may be exhausted. Therefore, at this time, some old caches need to be deleted to release storage space for new caches. In this embodiment, the following method is specifically adopted: during the publishing process of the event, compare the cache timestamps of the proxy nodes along the way to obtain the location of the oldest cache. If the space is exhausted, go back to the location of the oldest cache and delete the oldest cache.

Example 3:

In order to streamline cache redundancy and balance between memory access efficiency and storage occupancy, this embodiment adds a cache streamlining method on the basis of Embodiment 1. The specific method is as follows:

In the event publishing step, for each proxy node, a redundancy reduction method is used to cache the event, and the redundancy reduction method is specifically:

make Represents the collection of all agent nodes in the event distribution tree, Represents an agent node in the event distribution tree, where, Represents the collection of all child nodes of node b _i , where, Each proxy node contains several child nodes b _ij , and g _ij is each child node A corresponding subtree, where j=1, 2,..., n _t ; Indicates the set of subtrees of b _t composed of n _i subtrees.

According to the objective function Determines which subtrees of cache points to dispatch cached copies to, as shown in equation (1):

Among them, d _ij represents the maximum distance between the proxy node b _t and the end user of the subtree g _ij , r _ij represents the total number of user requests received by the proxy node b _i from the subtree g _ij , s _ij represents the total number of all nodes in the subtree g _ij The storage idle rate of , t is the reference value of the cache point; the objective function The numerator of the event represents the total cost of getting a copy of the event when the user places it on the subtree g _ij ; the first item of the numerator is the cost when the user at the end of the subtree g _ij requests the copy, and the first item of the numerator The binomial represents the rest of the subtrees The total overhead when users on the server request this copy.

The distances d _tj -t and d _ti +t here are only approximate values, and each proxy node cannot know the exact topology of the subtree, so it is approximately considered that the distance from each user to the cache point is the same. The smaller the total overhead of the cache request, or the larger the free rate of the total storage space on the subtree, the smaller the value of the objective function.

The proxy node b _t calculates a goal for each child node b _ij Select m=min{k, n _t } subtrees g _t1 , g _t2 , ..., g _tm with the smallest value to assign copies, and correspondingly send copies to child nodes b _t1 , b _t2 , ..., b The replica number k of the event of _tm is modified according to formula (2):

Set the replica number k of events sent to other child nodes b _tt as k _i =0 (m<l≤n _t ), and return the corresponding t value to 0.

Each proxy node serving as a cache point performs the above processing, so that each proxy node completes the storage of the cache copy in a distributed manner.

The above disclosure is only a specific embodiment of the present application, but the present application is not limited thereto. Any changes conceivable by those skilled in the art should fall within the protection scope of the present application.

Claims (10)

1. A caching method for realizing historical event subscription for a publish-subscribe system, characterized in that, comprising the steps: Routing table expansion step: adding cached routing information in the routing table of each proxy node, the cached routing information is used to provide routing information that can be located to the cache point; Event processing steps: when a proxy node sends out an event, calculate it through a hash function to obtain a cache point reference value; add a message header to the event, and record the cache point information and subscription path in the message header information; the cache point information includes the cache point reference value; Event publishing step: according to the subscription path information recorded in the message header, publish the event to the corresponding subscription node; for each proxy node in the publishing path, judge whether the current proxy node is As a cache point and cache, at the same time, the current proxy node synchronously updates the cache routing information corresponding to the event; Cache subscription step: a certain proxy node sends a cache request, and the cache request includes a subscription condition; the subscription condition is matched with the pre-stored filter conditions in the routing table of the current proxy node to determine a cache event; the cache event Refers to the event corresponding to the filter condition that successfully matches the subscription condition; Cache acquisition step: If the routing table of the current proxy node contains the cache routing information corresponding to the cache event, then directly find the cache point corresponding to the cache event according to the cache routing information, forward the cache request to the cache point, and obtain the cache ; If the routing table of the current proxy node does not contain the cache routing information corresponding to the cache event, then according to the subscription routing information corresponding to the cache event, send the cache request to the next proxy node, and perform the cache acquisition step in turn. 2. The caching method according to claim 1, wherein the hash function is as shown in formula (1): The definition domain of the hash function is the value range (a, b) of the event e, and the value range is [0, MAX_HOP]; where MAX_HOP is the maximum distance from the agent node on the event distribution path to its root node. 3. The caching method according to claim 2, wherein judging whether the current agent node is used as a caching point according to the caching point reference value and caching is specifically: After receiving the event, the current proxy node firstly subtracts 1 from the cache point reference value t, and then judges according to the cache point reference value. If t=0, the current proxy node is used as the cache point. 4. The caching method according to claim 1, wherein the routing table expansion step is specifically: Add the cache_nexthops field in the routing table of each proxy node, so that the structure of the routing table is: <filter|sub_nexthops|cache_nexthops> Among them, filter refers to the filter condition, sub_nexthops records the next-hop agent that forwards the events that successfully match the filter to the direction of the subscription node, and cache_nexthops records the next-hop agent that forwards the cache request that successfully matches the filter to the direction of the cache point. 5. The caching method according to claim 4, characterized in that, in the event publishing step, for the current agent node in the publishing process of the event, the update method of its cache routing information is specifically: If the current proxy node is behind the cache point, the next hop in the cache routing information is set as the predecessor proxy node; If the current proxy node happens to be a cache point, the next hop in the cache routing information is set to the current proxy node itself; If the current proxy node is in front of the cache point, the next hop in the cache routing information is set as the subsequent proxy node. 6. The cache method according to claim 5, wherein in the cache acquisition step, the method of forwarding the cache request to the cache point to obtain the cache is specifically: In the process of forwarding the cache request to the cache point, record the proxy nodes along the way in sequence, which is recorded as the cache request path; after the cache request reaches the cache point, the cache point sends the requested cache along the cache request path in reverse order to the proxy that issued the cache request node. 7. The caching method according to claim 1, wherein in the event publishing step, if the current proxy node is used as a cache point and cached, it is judged: if the storage space of the current proxy node is exhausted, then query Whether the direct neighbor outside the event distribution path still has available storage space; if the neighbor has available storage space, the event will be forwarded to the neighbor for caching. 8. The caching method according to claim 1, wherein the cache point information further comprises a cache point location, a cache copy number, and a cache timestamp, and the cache point location, the cache copy number, and the cache timestamp are initially The values are all empty; for each proxy node in the publishing path, if the current proxy node is used as a cache point and cached, update the cache point location, the number of cache copies, and the cache timestamp recorded in the message header of the event, and update the current The location of the cache point, the number of cache copies, and the cache timestamp stored in the proxy node. 9. The caching method according to claim 8, characterized in that, during the publishing process of the event, the cache timestamps of the proxy nodes are compared along the way to obtain the position of the oldest cache, when the event arrives at the subscribing node, if the entire If the storage space on the publishing path is exhausted, go back to the location of the oldest cache and delete the oldest cache. 10. The caching method according to claim 1, wherein in the event publishing step, for each proxy node, a redundancy reduction method is used to cache the event, and the redundancy reduction method is specifically: Let the proxy node be b _i , which contains several child nodes b _ij , and let g _ij be the subtree corresponding to each child node b _ij ; according to the objective function Determines which subtrees of cache points to dispatch cached copies to, as shown in equation (2): Among them, d _ij represents the maximum distance between the proxy node _bi and the end user of the subtree g _ij , r _ij represents the total number of user requests received by the proxy node bi from the subtree g _ij , s _ij represents the total _number of all nodes in the subtree g _ij The storage idle rate of , t is the reference value of the cache point; the objective function The numerator of represents the total cost for the user to obtain a copy of the event when it is placed on the subtree g _ij ; among them, the first item of the numerator is the cost when the user at the end of the subtree g _ij requests the copy, and the first item of the numerator The binomial represents the rest of the subtrees The total overhead when users on the network request this copy; The proxy node b _i calculates a goal for each child node b _ij Select m=min{k, n _i } sub-trees g _i1 , g _i2 , ..., g _im with the smallest value to assign copies, and correspondingly send copies to child nodes b _l1 , b _l2 , ..., b The replica number k of events of _lm is modified according to formula (3): Set the replica number k of events sent to other child nodes b _ii as k _l =0 (m<l≤n _i ), and return the corresponding t value to 0.