JP4590651B2 - Replication control method, system and program - Google Patents

Description

  The present invention prevents concentration of search and download loads on a certain servant within a short period of time when performing so-called peer-to-peer (P2P) type communication on an IP (Internet Protocol) network. In order to prevent the congestion of the line and the servant by executing the replication of the file requested to be downloaded in advance (sending the file copy to another servant and storing it) to distribute the load. Is low, the number of replicas is reduced to equalize the load in the network, and in addition, a method of actively deleting unused replicas to prevent the existence of useless replicas, that is, `` Replica control using biological principles for the purpose of controlling the number of replicas Law, and devices comprising them, and to their program.

P2P type communication is roughly divided into the following two types.
Napster type: A type that has a dedicated server that collectively manages the location of the target information Gnutella type: A type that inquires the location of the target information one after another using a message game system. The former is a search request packet from each servant to the dedicated server. Because it concentrates, it tends to be a server bottleneck. The latter inquires the neighboring servant for the target information one after another up to the hop count range set in the TTL, and each servant is branched by the amount of “arm” extended to the neighboring servant. Packet is sent out.

  When such information retrieval packets are transmitted in parallel from a large number of servants, the number of packets in the network increases explosively, and congestion over a wide area in the network is expected. In addition, when performing P2P communication in an IP network, in order to improve the hit rate for a certain content (file), a copy (replica) of the content is sent to another servant, and the load of the servant There is a technique (replication) for suppressing an increase in the usage rate of a line (for example, Q. Lv, P. Cao, E. Cohen, K. Li, and S. Schenker, “Search and replication in unstructured peer-to-peer networks, "Processeds of 16th ACM International Conference on Supercomputing (ICS'02), June 2002. (see Non-Patent Document 1), Goto, Ata, Murata:" P2P Replication placement method for improving service stability in networks, “Science Technical Report NS2002-152, Oct. 2002. (see Non-Patent Document 2)” and Nogami, Uchida: “File Search and Replication in Pure P2P "Science University 2004 (announced in March 2004) (see Non-Patent Document 3)).

  Non-Patent Document 1 describes an explanation of peer-to-peer type communication, replication, and the like. Non-Patent Document 2 proposes a method for preferentially transferring to a peer of a higher degree by selecting a transfer destination of the P2P logical network based on a probability weighted according to the degree of the neighbor peer. ing. Further, in Non-Patent Document 3, replication is performed for files in a certain node in order to achieve an arrangement method in which the hit rate is maintained at a certain level or higher and distributed to a certain degree of spatially appropriate density. An arrangement method for determining a node has been proposed.

  In addition to such replication, attempts are being made to delete unnecessary replicas as soon as possible in order to efficiently use the memory in each servant (for example, Nogami, Kamiyama; "Adaptive control of replica lifetime in Pure P2P," Shingaku Sodai 2005 (announced in August 2005) (see Non-Patent Document 4), and Nogami, Kamiyama: "Load dynamics in Pure P2P communication "Smoothing method, apparatus equipped with the same, and program thereof" (Japanese Patent Application No. 2005-247227 and drawings) (see Non-Patent Document 5))

  Non-Patent Document 4 proposes a method of adaptively controlling the lifetime of the created replica in memory according to a query frequency (FQH) per unit time. Further, in Non-Patent Document 5, information is transmitted and received between the nuclear servant device and a nearby servant device, the value of the congestion level of a certain file is referred to, and the congestion level is determined from a predetermined first threshold. There has been proposed a method in which the nuclear servant performs replication of the file when it is high to increase the number of replicas, and when the congestion level is lower than a second threshold, the replica of the memory of the nuclear servant device is deleted.

Q. Lv, P.M. Cao, E .; Cohen, K.M. Li, and S.L. Schenker, “Search and replication in unstructured peer-to-peer networks,” proceedings of 16th ACM International Conference on Supercomputing (ICS'02), June '200. Goto, Ata, Murata: "Replication placement method for improving service stability in P2P networks," IEICE Technical Report NS2002-152, Oct. 2002. Nogami, Uchida: “File Search & Replication in Pure P2P”, Shingaku Sodai 2004 (announced in March 2004) Nogami, Kamiyama; "Adaptive control of replica lifetime in Pure P2P," Shingaku Sodai 2005 (announced in August 2005) Nogami, Kamiyama: “Dynamic load smoothing method in Pure type P2P communication, apparatus including the same, and program thereof” (Japanese Patent Application No. 2005-247227 specification and drawings)

As described above, in the conventional technique, an evaluation function (a function of time) or a degree of congestion is defined, and the concept of controlling the number of replicas accordingly is used.
By the way, in the replication method, when multiple contents are targeted, the memory in each servant is limited, so if many content replicas are replicated many times, the memory capacity is insufficient. A replica cannot be created. In such a case, for a replica that is no longer referenced / searched by other servants, delete it to secure memory space, reserve memory space for the most recently required replica, Conversely, if there are more file search and transfer requests, it is desirable to create a replica by executing replication to distribute the load and support retention.

(the purpose)
The object of the present invention is, as described above, the “degree of request (for each file)” and “the frequency with which a file is referred / searched for time”, and the rating used when adding / deleting a replica. A replica control method using the principle of a living body that can control the number of replicas in consideration of, for example, changing the increase or decrease of the function according to the length of time elapsed from the past hit point, Another object of the present invention is to provide a device including the same and a program therefor.

In P2P (Peer-to-Peer) type communication, when certain file search and transfer requests are likely to be concentrated at a certain time, replication is performed, and “load distribution on a physical space (network)” is preferably performed. It should be done. However, performing replication uses memory in each servant and therefore uses extra resources.
Therefore, it is conceivable to control the number of replicas as many as necessary according to the number of requests, so that the optimum number can always be maintained by dynamically changing in time.
In the present invention, the rating function is used to change the increment and decrement of the value according to the length of time elapsed from the past hit point (this is an application of the biological principle). Use a different method.

  The replica control method using the principle of the living body of the present invention takes into consideration the control mechanism of the living body (that is, a mechanism in which the strength of the response to the stimulus decreases with time in the living body). By executing the above control, the number of replicas corresponding to the required number is dynamically changed according to the number of requests, and control is performed so that the optimum number can always be maintained.

  According to the present invention, when P2P type communication is performed, replication is performed on a target file in which inquiries and download requests in the network are likely to concentrate, and when the number of file transmission requests and search requests is still large. In addition, replication is executed to distribute the load. Conversely, when the number of file requests and search requests decreases, the file existence time is measured, and if the value falls below a certain value, the servant It is possible to maintain the optimum number of files necessary for a certain file in the network by applying the process of deleting the replica in the network as a control method using the biological control mechanism and executing this process. become.

The principles and embodiments of the present invention will be described below in detail with reference to the drawings.
(principle)
First, the definition of rating is as follows.
“Rating” (rating: meaning of evaluation amount, evaluation, score, etc.): In other words, a certain point (here, individual servant state) is given a “score”, and a certain event (here, “to servant”). This score increases or decreases depending on whether the file transmission request arrives or the time elapses, and if this value exceeds a certain threshold or falls below the threshold, what happens at that time? Or "things" (used here to perform replication or erase replicas).

(Processing 1) (Rating value increase / decrease method)
In the present invention, a certain servant S has a certain value V (referred to herein as a “rating”) for each file, the rating value V is used to increase or decrease the rating value according to the following procedure, and the other servant. In this method, replication is performed on a file or a file owned by itself is deleted (see claim 2).

(Step 1) rating that the file has of the servant, the file is increased by an amount △ V is the time t ₀ when the search request was received from the other servant.
(Procedure 2) The amount ΔV by which the rating is increased is determined by the time t ₁ when the previous search request arrives, the time t ₂ when the previous search request arrives,. Time t _m , and so on, which are determined by the elapsed time from the arrival time of the most recent m search requests, and the elapsed time (t ₀ -t ₁ ), (t ₀ -t ₂ ),. ₀ −t _m ) is a function. That is,
ΔV = Δ {(t ₀ -t ₁ ), (t ₀ -t ₂ ), ..., (t ₀ -t _m )}.
The value of the function V decreases with time according to a certain function.
When (Step 3) exceeds a threshold value theta ₁ with this rating function V, it decides to perform replication. Further, at a threshold θ less than ₂ with the value of the V, the file is deleted from the memory.

FIG. 1 is an explanatory diagram of a replication execution and replica deletion method according to the rating of the process 1 of the present invention.
As the search and hit its file generated at this time, and this time and t _0. At this time, the rating value is increased. The increased value ΔV is the time from the time t _{1 of} the previous search & hit to the current time from the time t _m of the previous search & hit. , (T ₀ -t ₁ ),..., (T ₀ -t _m ).
Here, the threshold θ ₁ is a threshold for executing replication, and the threshold θ ₂ is a threshold for deleting a file. In FIG. 1, the process proceeds from the right side to the left side.

(Process 2) (Replication control method considering the concept of "generation")
In the replication method of (Process 1), when the rating function exceeds a threshold value for executing replication, the probability of executing according to the following procedure is controlled by a function using the concept of “generation”. reference).
(Procedure 1) A servant that already holds a file K at a certain time is defined as “generation 0 related to this file K”. The servant that holds the replica of the file by replication from this generation 0 or through a series of procedures of “search, request, download” is created from “generation 1”, and similarly from “generation G”. A servant having a replica is called “generation G + 1 for this file”.
Here, if the result exceeds the threshold value theta ₁ rating value V in (process 1), the servant in the case of generation G with respect to the file, and performing the replication at a certain probability. This probability P is a decreasing function with respect to G, and replication is unlikely to occur every time the generation falls.
This produces an effect of suppressing unnecessary replica creation. Also, generation 0 is not deleted no matter how small the value of V is. This prevents a file in the network from disappearing.

(Process 3) (Conditional replication control method)
This is a replication method in which a replica is created by performing replication in a servant around itself when the following conditions are satisfied simultaneously (see claim 4).

(Condition 1) per time interval T _b, and counts the number of searches hit arriving at the servant for each file. The moment when a t _1, a range of the most recent: If _(n * T b n is an integer) becomes the moving average value [Phi (t ₁₎ is the threshold value theta ₁ or more search hits for in. Where moving average

がある閾値θ_１以上になった場合。ただし、Φ（ｔ_１，ｉ）は、時刻（ｔ_１−ｉ＊Ｔ_ｂ）から（ｔ_１−（ｉ−１）Ｔ_ｂ）までの区間における当該サーバント内の当該ファイルの検索ヒット数である。
すなわち、
Φ（ｔ_１）≧θ_１

When a certain threshold θ1 is greater than or equal to ₁ . Here, Φ (t ₁ , i) is the number of search hits for the file in the servant in the section from time (t ₁ −i * T _b ) to (t ₁ − (i−1) T _b ). .
That is,
Φ (t ₁ ) ≧ θ ₁

(Condition 2) When the search hit count of the file in the servant in the current section is equal to or greater than the value in the previous section,
Φ (t ₁ ) ≧ Φ (t ₁ −T _b )

(Condition 3) the ratio [psi (t ₁₎ and the number the file in the servant in this section is the download request [psi (t _1), also the number of times that the file is to search and hit Η and (t ₁₎ / Η (t ₁ ) is greater than a certain constant k.

FIG. 2 is an explanatory diagram for the above three conditions in the conditional replication control method.
That is,
(Condition 1) means that the moving average value of the number of search hits at time t ₁ exceeds the threshold value θ ₁ . In (Condition 2), the number of hits of the file is larger in the current section than in the previous section. That is, the condition is that the number of hits is increasing.
(Condition 3) is larger than a certain constant value k in the ratio between the “number of times download is requested” and the “number of searches and hits” in the section. That is, not only search and hit, but also the case where the actual download rate from the servant is large and it can be considered that the file possession rate other than the servant is low.
In FIG. 2, the solid line above the broken line indicating the threshold θ ₁ satisfies the condition 1, and the solid line above the next time satisfies the condition 2 that the number of hits increases from the previous time. In addition, the solid line in FIG. 2 represents the number of searches and hits (satisfaction of condition 3).
Accordingly, in FIG. 2, replication is executed at the end of the third and fifth downloads from the left.

FIG. 3 is an explanatory diagram of symbols used to represent each section.
The moving average value Φ _i = Φ _i (t ₁ , i) is represented, and in the servant in the section from the time (t ₁ −i * T _b ) to (t ₁ − (i−1) T _b ) The number of search hits for the file.
Also, ψ (t ₁ ) is the number of times that the file is requested to be downloaded in the section (t ₁ -T _b , t ₁ ).
Further, η (t ₁ ) is the number of times that the file is searched and hit in the section (t ₁ −T _b , t ₁ ).
K is a fixed value.

(Process 4) (Method for determining the number of deployments during replication)
When a certain servant A performs replication with respect to a certain file F, replicas are arranged for several servants around it. At that time, the number of the arrangements is expressed as “(a) number of searches and hits in a certain latest section [t ₁ , t ₁ + Δt] to the file F of the servant A, or (b) This process is determined by “the number of download requests” (see claim 5).

(Example)
Hereinafter, examples carried out based on the above principle will be described.
Example 1
For the purpose of appropriately maintaining the number of replicas and their location on the P2P network, control is performed according to the following procedures (1) to (3).
(1) The method of Non-Patent Document 3 is used as a replication method in order to deploy replicas on a P2P network so as to maintain an appropriate spread.
This method is a method of determining a servant to be replicated by the following procedure.

(Step 1) If the number of hits of the file in the latest T time at a certain time is greater than or equal to a certain threshold value, go to Step 2, otherwise end.
(Step 2) Attention is paid to nodes at k hops ahead (including intermediate routes) for all routes of the node Γ.
(Step 3) With respect to the node i, if Deg (i) ≧ m, it is determined as a target (replication) with probability P, and if Deg (i) <m, it is determined as non-target. (Where Deg (i) is the degree (number of arms) of node i, 1 ≦ i ≦ N _j , m: integer greater than or equal to 0, N _j ; total number of nodes at the j-th k-hop destination)
(Step 4) The Mi route (M _i = “X * Deg (i)]) corresponding to the ratio X of the nodes i is selected at random, and attention is paid to the k-hop destination node (where“ Z ”is Z Represents the largest integer that does not exceed.
(Step 5) Steps 3 to 4 are executed N _j times (i = 1 to N _j ).
(Step 6) Steps 3 to 5 are executed L times (j = 1 to L).

(2) The decision to execute replication is made when the following conditions are satisfied.
(Condition 1) Every 5 seconds, the number of searches that arrive at the servant and hit is counted for each file. Moving average value of the number of search hits in the most recent range (12 × 5 seconds = 60 seconds) when the current time is t ₁

がある閾値θ_１＝１０以上になった場合。ただし、Φ（ｔ_１，ｉ）は、時刻（ｔ_１−ｉ＊５）から（ｔ_１−（ｉ−１）＊Ｔ_ｂ）までの区間における当該サーバント内の当該ファイルの検索ヒット数である。
すなわち、
移動平均値Φ（ｔ_１）≧１０

When a certain threshold θ ₁ = 10 or more. Here, Φ (t ₁ , i) is the number of search hits of the file in the servant in the section from time (t ₁ −i * 5) to (t ₁ − (i−1) * T _b ). .
That is,
Moving average value Φ (t ₁ ) ≧ 10

(Condition 2) The number of search hits of the file in the servant in the current section is equal to or greater than the value in the previous section.
That is,
Φ (t ₁ ) ≧ Φ (t ₁ −5)

(Condition 3) the ratio [psi (t ₁₎ and the number the file in the servant is download requests in this interval [psi (t _1), also the number of times the file is searched and hit Η a (t ₁₎ / Η (t ₁ ) is greater than a certain constant k = 0.8.

(3) The replica is deleted when a certain rating function has V at that time point equal to or smaller than the threshold θ ₂ = 0.0001.

(Example 2)
At some point, a servant that already holds a file K is set to generation 0. The servant that has retained the replica of the file by replication from generation 0 or by a series of procedures of “search, request, download” is generation 1, and so on, generations 2, 3, 4,. And Here, of the file of the servant Les - when computing the value V exceeds a threshold value theta _1, if the servant is generation G with respect to the file, the probability P = exp (-αG) (wherein, α is a replication with a weighting factor (constant).

(Example 3)
This is a method in which a certain servant determines the number of replicas created when executing replication for a certain file F under the following conditions.
In other words, at a certain time _{t 1,
0 ≦ U (t 1) ≦} d 1 if _{R 1} or _{d 1 ≦ U (t 1)} ≦ d 2 if _{R 2} or _{d 2 ≦ U (t 1)} ≦ d 3 If _{R 3} or ...
d _k ≦ U _(t 1) if _{R k + 1} or or _{0 ≦ V (t 1) ≦} e 1 if _{Q 1} or _{e 1 ≦ V (t 1)} ≦ e 2 if _{Q 2} pieces _{e 2} ≦ V ( If t ₁ ) ≦ e ₃ then ₃ Q ...
If e _k ≦ V (t ₁ ), Q _{k + 1} pieces are created.

Either one of these values is used, or both are calculated and the larger value is selected, and replicas corresponding to that number are created.
However,
U (t ₁ ): Number of searches and hits in the latest section [t ₁ −Δt, t ₁ ] of the servant A to the file F,
V (t ₁ ): Same as above, the number of download requests for the file F,
R _i , Q _i (i = 1, 2,..., K + 1): constants.

Example 4
Hereinafter, functions inside the servant according to the fourth embodiment of the present invention will be described.
FIG. 4 is a diagram illustrating a hardware configuration example of the servant.
In FIG. 4, reference numeral 1 denotes a servant (terminal device), which is a so-called terminal having a normal PC function, but appears as a servant as a P2PP application layer. 2 is a normal timer. Reference numeral 3 denotes an information calculator, which includes a rating information calculator. The rating / generation / replica placement number information calculation unit 4 has a function of calculating the rating / generation / replica placement number for each replica file. An information management / transfer unit 5 includes a replica search information management unit 6, a hit information management unit 7, and a request information transfer unit 8. The replica search information management unit 6 has a function of managing replica search information from other servants. The hit information management unit 7 has a function of managing information for which the search has hit the servant 1.

The request information transfer unit 8 has a function of transferring a hit replica to another servant.
Reference numeral 9 denotes a memory unit. A memory capacity control / control unit 10 and a memory unit 11 are included therein.
The memory capacity control / control unit 10 has a control function for managing the remaining capacity of the memory, deleting unnecessary information, and only writing new information. The memory unit 11 is a place for storing / accumulating various types of replication information, hit information, information about held files, IP addresses of servants, and the like.
A replication execution unit 12 includes a replication information analysis unit 13 and a replication / deletion action determination / execution unit 14 inside.
The replication information analysis unit 13 has a function of analyzing which file is replicated at what timing. The replication / deletion action determination / execution unit 14 has a function of determining and executing an action to be taken such as execution of replication or unnecessary replica deletion.

Reference numeral 15 denotes a PAD (packet assembling / decomposing unit) having a function of converting various types of information into IP packets, and conversely, decomposing IP packets into various types of information.
A packet transmission unit 16 has a function of transmitting an IP packet to another servant.
A packet receiving unit 17 has a function of receiving an arrival packet from another servant in the network. Reference numeral 18 denotes an IP network (Internet) to which the servant is connected.

Note that the servant device of the present invention is not limited to the configuration shown in FIG. 4 because it is sufficient to arrange means having the following functions. That is,
1) means for managing the number of search request hits in itself, 2) means for calculating the rating value and generation for each file owned by itself, and 3) increasing the rating value of each replica file at present. Means for calculating the amount, 4) comparing the calculated rating value with the first threshold value, and taking into account the “generation” of the file, and 5) the calculated It is an apparatus comprising means for deleting the replica by comparing the rating value with the second threshold, and 6) means for determining the number of replication arrangements based on the number of hits and the number of downloads in a certain time interval.

  By programming the procedure described in each embodiment and storing it in a recording medium such as a CD-ROM, attaching this medium to a servant processor, installing the program in the processor, and executing it The present invention can be easily realized. It is also possible to generalize this program by downloading the program from the processor to another terminal via the Internet.

It is explanatory drawing regarding the increase in a rating value regarding the execution of replication by the rating of this invention, and the replica deletion method. It is explanatory drawing of three types of conditions in the replication control method with a condition regarding this invention. It is explanatory drawing of the area in the conditional replication control method, and the symbol used for it. It is an internal block diagram of a servant for implement | achieving the function of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Servant 2 Timer 3 Information calculation part 4 Rating / generation / replica arrangement number / information calculation part 5 Information management / transfer part 6 Replica search information management part 7 Hit information management part 8 Request information transfer part 9 Memory part 10 Memory capacity management / Control unit 11 Memory 12 Replication execution unit 13 Replication information analysis unit 14 Replication / deletion action determination / execution unit 15 PAD
16 packet transmitter 17 packet receiver 18 IP network (Internet)

Claims (6)

A method for controlling replication in a servant device that performs P2P communication over an IP network by programmed computer processing by a processor provided therein,
As a programmed computer processing procedure by the processor of the servant device,
To another servant device for a file stored in the storage device in which the number of search requests in a predetermined time interval measured by a timer provided in the servant device is equal to or greater than a predetermined first threshold value. A first procedure for performing the replication of at a predetermined probability P;
A second procedure for deleting from the storage device a file whose number of search requests in the time interval is less than or equal to a predetermined second threshold and whose lifetime is less than or equal to a predetermined third threshold;
A replication control method comprising: a third procedure for changing the probability P of executing the replication according to a generation defined below with respect to the file of the servant device that executes the replication.

Definition of generation: A servant device that holds a file that already exists at a certain point in time is defined as “generation 0 related to the file”, and a series of procedures “retrieval, request, download” by replication from the generation 0 Thus, the servant device that holds the replica of the file is called “generation 1”, and similarly, the servant device having the replica created from “generation G” is called “generation G + 1 for the file”. The replication control method according to claim 1 ,
The third procedure is
Probability P of generation G servant devices holding files replicated directly from generation 0 servant devices holding files already existing at a certain point in time or via one or more other servant devices. Including a procedure for reducing the amount by a predetermined amount from the replication source servant device,
The second procedure is as follows:
A replication control method comprising a procedure of not deleting a file of the generation 0 servant device. A replication control method according to claim 1 or 2 , wherein
The first procedure is as follows:
A procedure for counting the number of search hits for each file at a predetermined time interval Tb;
A procedure for calculating the moving average value Φ of the number of search hits counted for each file,
For each file, the procedure to compare the number of search hits counted with the number of search hits previously counted,
For each file, a procedure for calculating a ratio Ψ / H between the number of search hits H counted and the number of times Ψ requested to download the file;
The calculated moving average value Φ is not less than a predetermined threshold value θ1,
The number of search hits counted above is equal to or greater than the number of search hits previously counted,
And, when the calculated ratio Ψ / H is larger than a predetermined constant k,
A replication control method comprising a procedure for executing replication for the file. The replication control method according to any one of claims 1 to 3 , wherein
The first procedure is as follows:
Including a procedure for determining the number of servant devices in which a replica of a file to be replicated is arranged according to either a predetermined condition according to the number of search hits for the file or a predetermined condition according to the number of download requests A replication control method characterized by the above. A replication control system provided in a servant device that performs P2P type communication on an IP network by programmed computer processing by a processor provided therein,
As means for executing programmed computer processing by the processor,
A replication control system comprising means for executing each procedure in the replication control method according to any one of claims 1 to 4 . The program for making a computer perform each procedure in the replication control method in any one of Claims 1-4 .

Images