KR101638797B1 - Method for scheduling query - Google Patents

Abstract

A method of assigning a query, the method comprising: receiving query data from at least one client; calculating an Exponential Moving Average (EMA) of at least one application server to assign the received query data to at least one application server Determining a number of queries corresponding to a load, which is an amount of work assigned to at least one application server, assigning the received query data to one of application servers of at least one application server based on the calculated EMA and the load .

Description

{METHOD FOR SCHEDULING QUERY}

The present invention relates to a query allocation method.

In recent years, distributed networks have been widely used for storing, retrieving and processing large-scale data. Various researches have been conducted on how to allocate and distribute query data to a distributed network.

At this time, the query data is allocated in such a way as to reduce the load on each distributed resource. In this regard, Korean Unexamined Patent Publication No. 2003-0045987 (published Jun. 10, 2003) discloses a round-robin scheduling method.

However, since the round-robin scheduling method ignores the cache hit rate in the distributed network, the reuse rate of the data is low, and evenly distributing the number of jobs to the distributed network makes the distributed network It is not a method to utilize efficiently.

In an embodiment of the present invention, not only a server in which the distance between the center point of query data and the Exponential Moving Average (EMA) of a plurality of application servers is minimized, but also a query corresponding to a load allocated to a plurality of applications It is possible to provide a query allocation method capable of allocating query data to an application server so that query data is processed while simultaneously considering load balancing and cache hit rate. It should be understood, however, that the technical scope of the present invention is not limited to the above-described technical problems, and other technical problems may exist.

As a technical means for achieving the above-mentioned technical object, an embodiment of the present invention is a method for receiving query data from at least one client, comprising the steps of receiving at least one Calculating an Exponential Moving Average (EMA) of an application server; determining a number of queries corresponding to a load, which is an amount of work assigned to at least one application server, based on the calculated EMA and the load; And assigning the query data to at least one of the application servers.

According to any one of the above-mentioned objects of the present invention, it is possible to simultaneously process load balancing and cache hit ratio without causing an overload in the system in a distributed environment.

1 is a block diagram illustrating a query allocation system according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating a query allocation server shown in FIG. 1. FIG.
FIG. 3 is a diagram illustrating calculating the distance between an EMA point and a query in the query assignment server shown in FIG. 1; FIG.
FIG. 4 is a diagram illustrating calculation of the distance between an EMA point and a query in consideration of load in the query assignment server shown in FIG. 1; FIG.
5 is a diagram illustrating an embodiment for balancing the load of an application server in the query allocation server shown in FIG.
6 illustrates an embodiment of an algorithm for assigning queries based on EMA points and load in the query allocation server shown in FIG.
7 is a graph showing a performance evaluation graph of a query allocation method according to an embodiment of the present invention.
8 is a flowchart illustrating an example of a query allocation method according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings, which will be readily apparent to those skilled in the art. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In order to clearly illustrate the present invention, parts not related to the description are omitted, and similar parts are denoted by like reference characters throughout the specification.

Throughout the specification, when a part is referred to as being "connected" to another part, it includes not only "directly connected" but also "electrically connected" with another part in between . Also, when an element is referred to as "including" an element, it is to be understood that the element may include other elements as well as other elements, And does not preclude the presence or addition of one or more other features, integers, steps, operations, components, parts, or combinations thereof.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram illustrating a query allocation system according to an embodiment of the present invention. The query allocation system 1 according to an embodiment of the present invention may include at least one client 100, a query allocation server 300, and at least one application server 400. However, since the query assignment system 1 of FIG. 1 is only an embodiment of the present invention, the present invention is not limited to FIG.

At this time, the respective components of FIG. 1 may be connected through a network 200 in general. For example, as shown in FIG. 1, at least one client 100 and a query allocation server 300 may be connected through a network 200. In addition, the query assignment server 300 may be connected to at least one application server 400 through a network. At least one client 100 may be connected to at least one application server 400 through the network 200. [

Here, the network 200 refers to a connection structure capable of exchanging information between nodes such as terminals and servers. An example of the network 200 is an Internet, a LAN (Local Area Network) ), Wireless LAN (Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), 3G, 4G, LTE, Wi-Fi and the like. At least one client 100, a query assignment server 300, and at least one application server 400 shown in FIG. 1 are not construed as being limited to those shown in FIG.

At least one client 100 may send at least one query to at least one application server 400. At least one client 100 can receive response data that is the result of processing a query in at least one application server 400. [

The query assignment server 300 can receive query data from at least one client 100. [ The query allocation server 300 is further provided with an EMA (Exponential Moving Average) point of at least one application server 400 or a load corresponding to query data already allocated to at least one application server 400 Load), it is possible to allocate the query data received from at least one client 100.

At least one application server 400 may process query data received from at least one client 100 and distributed by the query assignment server 300. At least one application server 400 can transmit response data to the query, which is the processing result of the query data, to at least one client 100.

 FIG. 2 is a diagram for explaining a query assignment server shown in FIG. 1, FIG. 3 is a diagram for calculating a distance between an EMA point and a query in the query assignment server shown in FIG. 1, 1 illustrates calculating the distance between an EMA point and a query in consideration of load in a query assignment server shown in FIG.

2, the query assignment server 300 according to an exemplary embodiment of the present invention includes a receiving unit 410, a calculating unit 420, a determining unit 430, an assigning unit 440, an updating unit 450, . ≪ / RTI >

The receiving unit 410 can receive query data from at least one client 100. [

The calculation unit 420 may calculate an exponential moving average (EMA) of at least one application server 400 in order to allocate the received query data to at least one application server 400. [ Here, the calculation unit 420 can randomly assign an EMA to at least one application server 400, and calculate the distance between the multi-dimensional center point of the received query data and the EMA.

Hereinafter, the concept of the EMA will be briefly described.

The EMA is a statistical method for detecting long-term or short-term trend fluctuation. In the calculation unit 420, the weighted average of the received query data can be obtained by weighting the received query data. That is, the cached data can be expressed in multidimensional coordinates, and the multi-dimensional center point can be used to obtain the EMA value of the cache data. The EMA according to the variation of the time can be expressed by the following equation (1).

Here, t means time, EMA [t] means EMA which changes with time, and a means a weight having a value between 0 and 1. Here, the formulas and definitions for the EMA are obvious to those skilled in the art, and therefore, a description thereof will be omitted.

Returning to the description of the calculation unit 320 of FIG. 2, in the query allocation method according to the present invention, the calculation unit 320 determines whether at least one application server 400 having a distributed cache infrastructure has distributed queries An EMA point can be allocated so that the EMA point can be adapted even in a situation where a change occurs. That is, assuming that there are n application servers 400 and one query assignment server 300, the calculation unit 320 calculates EMA variables to be given to the n application servers 400 in order to initialize the EMA points Can be determined at random.

The query allocation method according to an embodiment of the present invention can consider both i) EMA and ii) load. First, i) a method of assigning a query using only EMA will be described, and a method considering both i) EMA and ii) load will be described hereinafter.

Here, the calculation unit 320 may allocate a query using only i) EMA. That is, when the query data is q, the calculation unit 320 calculates the center point of q and searches the application server 400 having the EMA point closest to the center point of q. Here, the distance between the query data and the EMA point can be defined as an Euclidian distance. At this time, i) a method of allocating a query using only EMA will be described with reference to FIG. Here, (a) EMA has each center point and each area, and when the query data (Query, q) is received, the application server having the EMA with the closest distance between the EMA point and the query data 400) is assigned to the query. (B) _Assume that the region of EMA _k is Bound (k, k + 1) and the region of EMA _{k +1} is Bound (k + 1, k). At this time, if a center point of the question data present in the area of the EMA _k, because the distance between the center point of the EMA _k and Q data, is shorter than the distance between the center point of the EMA _{k + 1} to the query data, query data is EMA _k . This will be compared to i) considering both EMA and ii) load as a result of i) considering only EMA.

Returning to FIG. 2, if the query data q is assigned to i, which is the application server 400, as a result of the calculation unit 320, the EMA of the application server 400, i, 2, and the EMA of the application server 400 which is i can be updated.

Here, EMAi * denotes the EMA of the i-th application server, a denotes a weight of 0 to 1, and q denotes a multi-dimensional center point of the query data.

On the other hand, in order to consider both i) EMA and ii) load, it is necessary to grasp the load corresponding to the query data allocated to each of at least one application server 400. [ Therefore, the holding unit 430 can grasp the number of queries corresponding to the load, which is the workload assigned to at least one application server 400. [

The assigning unit 440 can assign the received query data to one of the application servers 400 of at least one application server 400 based on the calculated EMA and the load. Here, the assigning unit 440 can consider both the EMA and the load through the following Equation (3). That is, the assigning unit 440 can use the result of multiplying the EMA by the load.

Here, Bound (i, j) denotes an EMA distance between the i-th application server and the j-th application server, Bound (j, i) denotes an EMA distance between the j- j denotes the number of queries corresponding to the load, which is the workload allocated to the j-th application server, and Load (i) denotes the number of queries corresponding to the load, which is the workload allocated to the i-th application server.

In summary, the assigning unit 440 assigns the total EMA calculated by multiplying the calculated EMA and the load to the at least one application 400, and calculates the distance between the multi-dimensional center point of the received query data and the total EMA And assigning the query data to any one of the application servers 400 having the shortest calculated distance. Here, the distance between the multi-dimensional center point and the EMA can be defined as an Euclidian distance. At this time, the total EMA is defined as the updated EMA.

A method of allocating a query in consideration of both i) EMA and ii) load will be described with reference to Fig. 4 in the allocating unit 440. Fig. 4 (a), EMA ₁ has a load of 15 and EMA ₂ has a load of 5, so that the area of EMA _{1 and} the area of EMA ₂ can be set to 5:15. Here, a circle formed around each EMA may be an Apollonius circle. That is, EMA1 is, than EMA ₂ because it is already the load is three times as much hanging, by setting the area of the EMA ₁ to be inversely proportional to the load ratio than the area of the EMA _2, to reduce the chance of a new query data is located in the region of EMA ₁ will be. (B) _Assume that the region of EMA _k is Bound (k, k + 1) and the region of EMA _{k +1} is Bound (k + 1, k). At this time, the load on each EMA is already taken into account in the area of each EMA. That is, the EMA _k having a load of 30 has an area of 1, and the EMA _{k +1} having a load of 10 has an area of 3. In this case, when the center point of the query data is within the area of EMA _{k +1} , the distance between EMA _k and the center point of the query data is shorter than the distance between EMA _{k + 1} and the center point of the query data, Area is set to EMA _k : EMA _{k +1} = 1: 3, and since the query data already exists in the area of EMA _{k +1} , new query data is assigned to EMA _{k +1} .

Returning to Fig. 2, the update unit 450 may update the EMA of the application server 400 to which the query data is allocated and the final EMA based on the load. At this time, EMA can be calculated by Equations (2) and (3). Here, the query data may be assigned to any one of the at least one application server 400 with the same probability. That is, the query data may be assigned to any one of the application servers 400, taking into consideration the similarity with the query data already allocated to at least one application server 400. [ This may be to increase cache hit rate. When query data is received, the query data may be stored in place of the oldest query data stored in the queue of the query assignment server 400. [

The matters not described with respect to the query allocation methods of FIGS. 2 to 4 can be easily deduced from the same or explanations as those described above with reference to FIG. 1, and the description will be omitted.

5 is a diagram illustrating an embodiment for balancing the load of an application server in the query allocation server shown in FIG. Referring to FIG. 5, a query allocation method according to an embodiment of the present invention is expressed in terms of min-max optimization. Here, assuming that a linear query space and a uniform quality distribution are satisfied, it is assumed that the second application server 400 whose EMA is E2 has a maximum load. That is, it is assumed that the load 2 loaded on the second application server 400 is larger than the load 1 loaded on the first application server 400 or the load 3 loaded on the third application server 400 do. It is assumed that m1 is closer to E2 than E1 and m2 is closer to E2 than E3. That is, since r2 <r1 and r3 <r4, r2 + r3 <r1 + r4, and the probability that the query data is allocated to the second application server 400 is that the query data is allocated to the first application server 400 The probability that the query data is allocated to the third application server 400 is lower than the probability. Therefore, the load on the second application server 400 can be reduced.

The matters not described in relation to the query allocation method of FIG. 5 can be easily deduced from the same or described contents of the query allocation method described above with reference to FIGS. 1 to 4, and the description thereof will be omitted.

6 illustrates an embodiment of an algorithm for assigning queries based on EMA points and load in the query allocation server shown in FIG. The algorithm according to an embodiment of the present invention is not limited to the algorithm of Fig. 6, but may be applied to at least one application server in consideration of i) EMA or ii) load according to the present invention, It will be apparent to those skilled in the art that any algorithm can be used to distribute and allocate.

The matters not described with respect to the query allocation method of FIG. 6 can be easily deduced from the same or explanations of the query allocation method described above with reference to FIGS. 1 to 5, and the description thereof will be omitted.

FIG. 7 and FIG. 8 are graphs showing performance evaluation graphs of a query allocation method according to an embodiment of the present invention. 7 and 8, in order to measure the performance of the query allocation method according to an embodiment of the present invention, a query set having various query distributions (uniform, normal, Zipf) is used. Here, the experiments of FIGS. 7 and 8 use a simulator for performance evaluation of 41 Linux cluster machines and a plurality of machines. Here, the experimental machine is equipped with dual QuadCore Xeon E5506 2.13 GHz CPUs, each node uses 12 GB of memory, and each node is connected to Gigabit switched Ethernet.

Referring to FIG. 7, as the number of application servers increases, (a) query response time, (b) cache hit rate, and (c) load balancing change. At this time, the cache hit rate means the probability that the data used by processing the previous query data is reused.

Here, the query response time means the time from when the query allocation server receives the query data, the query data is allocated to the application server, and the application server completes the processing of the query data allocated thereto. Here, each application server has an LRU cache capable of storing 100 query results.

(a) As the application server increases, the query response time decreases. That is, a plurality of application servers can provide more cache space. Here, the method according to an embodiment of the present invention is represented by BEMA (Balancing EMA). As the number of application servers increases, the processing time of the BEMA of the present invention is the fastest. Also, (b) the BEMA of the present invention shows that the cache hit ratio increases the most as the number of application servers increases. (C) It can be seen that the BEMA of the present invention has the best load distribution ratio regardless of whether the number of applications is small or not.

8 is a flowchart illustrating an example of a query allocation method according to an embodiment of the present invention. Referring to FIG. 8, the query assignment server receives query data from at least one client (S7100).

Then, the query allocation server calculates an Exponential Moving Average (EMA) of at least one application server in order to allocate the received query data to at least one application server (S7200).

In addition, the query assignment server determines the number of queries corresponding to the load, which is the workload allocated to at least one application server (S7300).

Finally, the query assignment server assigns the received query data to any one of the application servers based on the calculated EMA and the load (S7400).

8, matters not described with respect to the query allocation method according to an embodiment of the present invention can be easily derived from the same or described contents of the query allocation method described above with reference to FIGS. 1 to 7 The description will be omitted.

The order between the above-described steps (S7100 to S7400) is merely an example, but is not limited thereto. That is, the order between the above-described steps S7100 to S7400 may be mutually varied, and some of the steps may be executed or deleted at the same time.

The method of assigning a query according to the embodiment described with reference to FIG. 8 may also be implemented in the form of a recording medium including instructions executable by a computer such as an application or a program module executed by a computer. Computer readable media can be any available media that can be accessed by a computer and includes both volatile and nonvolatile media, removable and non-removable media. In addition, the computer-readable medium can include both computer storage media and communication media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Communication media typically includes any information delivery media, including computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave, or other transport mechanism.

The query allocation method according to an embodiment of the present invention may be executed by an application installed in a terminal (which may include a program included in a platform or an operating system basically installed in the terminal) (I.e., a program) installed directly on the master terminal through an application providing server such as an application store server, an application, or a web server associated with the service. In this sense, the query allocation method according to an embodiment of the present invention described above is implemented as an application (i.e., a program) installed basically in a terminal or directly installed by a user and stored in a computer readable recording medium Lt; / RTI &gt;

It will be understood by those skilled in the art that the foregoing description of the present invention is for illustrative purposes only and that those of ordinary skill in the art can readily understand that various changes and modifications may be made without departing from the spirit or essential characteristics of the present invention. will be. It is therefore to be understood that the above-described embodiments are illustrative in all aspects and not restrictive. For example, each component described as a single entity may be distributed and implemented, and components described as being distributed may also be implemented in a combined form.

The scope of the present invention is defined by the appended claims rather than the detailed description and all changes or modifications derived from the meaning and scope of the claims and their equivalents are to be construed as being included within the scope of the present invention do.

Claims (10)

A query allocation method executed in a query assignment server,
Receiving query data from at least one client;
Calculating Exponential Moving Average (EMA) of the at least one application server to assign the received query data to at least one application server;
Determining a number of queries corresponding to a load, which is an amount of work allocated to the at least one application server;
Assigning the received query data to any one of the application servers based on the calculated EMA and the load;
, &Lt; / RTI &
The step of calculating the EMA comprises:
Randomly assigning an EMA to the at least one application server;
Calculating a distance between the multi-dimensional center point of the received query data and the EMA;
Wherein the query is a query.
delete The method according to claim 1,
Wherein the step of assigning the received query data to one of the at least one application server based on the calculated EMA and the load includes:
Assigning a total EMA calculated by multiplying the calculated EMA to a load to the at least one application,
Calculating a distance between the multi-dimensional center point of the received query data and the granted total EMA,
And allocating the query data to any one of the application servers having the shortest calculated distance.
The method of claim 3,
Wherein the distance between the multi-dimensional center point and the EMA is defined as an Euclidean distance.
The method according to claim 1,
Wherein the query data is assigned to one of the at least one application server with the same probability.
The method according to claim 1,
Wherein the query data is allocated to any one of the application servers in consideration of the degree of similarity with query data already allocated to the at least one application server.
The method according to claim 1,
Updating the final EMA based on the EMA and the load of the application server to which the query data is allocated
Further comprising the steps of:
The method according to claim 1,
Storing the query data in place of the oldest query data stored in a queue of the query assignment server when the query data is received;
Further comprising the steps of:
The method according to claim 1,
Wherein the step of assigning the received query data to one of the at least one application server based on the calculated EMA and the load includes:
Wherein the query data is allocated to one of the application servers using the following equation:

Here, EMAi denotes EMA of the i-th application server, α denotes a weight of 0 to 1, q denotes a multi-dimensional center point of query data, Bound (i, j) (j, i) denotes the EMA distance between the j-th application server and Bound (j, i) denotes the EMA distance between the j-th application server and the i-th application server, (I) denotes the number of queries corresponding to the load, which is the workload allocated to the i-th application server.
10. A processor-readable recording medium having recorded thereon a program for causing a processor to execute the method according to any one of claims 1 to 9.

Images