KR100946696B1 - Streaming software service system using time-series technique and method thereof - Google Patents

Abstract

The present invention relates to a streaming software service system using a time-series technique and a method thereof. The present invention relates to a streaming pattern of a corresponding software in a streaming service of software such as software as a service (SaaS). Clustering is performed using a time-series technique, the resources of each software are estimated from each cluster, the resources are allocated, and the load of the streaming service is distributed using the resources. According to the present invention, since the load can be distributed by estimating the amount of resources required in the platform for preparing a large-capacity software streaming service, it is possible to build a pilot system for streaming service while efficiently utilizing a minimum of resources.

SaaS, Streaming Software, Time Series, Resource Allocation, Load Balancing

Description

Streaming software service system using time series technique and its method {STREAMING SOFTWARE SERVICE SYSTEM USING TIME-SERIES TECHNIQUE AND METHOD THEREOF}

The present invention relates to a streaming software service system and a method thereof, and more particularly, a time series technique for classifying and determining the characteristics of a streaming service software using a time series technique and dynamically distributing a service load by estimating the amount of resources required for the service. The present invention relates to a streaming software service system and a method thereof.

The present invention is derived from the research conducted as part of the IT new growth engine core technology development project of the Ministry of Information and Communication and the Ministry of Information and Communication Research and Development. [Task Management Number: 2005-S-007-02, Title: Open SW-based On-Demand Office Development of environmental provision technology].

Software streaming technology is a technology that executes a part of the execution code of application software existing in a server through a streaming technique on the client, and the client only needs to execute the software without downloading or installing all of the software to the storage device of the local system. Run by receiving from server.

In recent years, the usage of such software streaming services is increasing, and the number of streaming software such as software as a service (SaaS) is increasing.

In a system preparing such a large software streaming service, load balancing is required through efficient management of resources such as CPU, memory, and network bandwidth.

Conventional high-capacity software streaming service load balancing technology includes a load balancing system using a load balancing switch or a load balancing system using a virtual server system. Since these server load balancing methods only support simple load balancing algorithms, they cannot support flexible load balancing according to the application.

Also, due to the nature of the software streaming service, there are cases where load balancing is performed using the difference between the initial bandwidth of the software and the network bandwidth used at the time of user request. Therefore, the system cannot allocate the required resources dynamically in case the bandwidth demand increases due to the user's request.

The present invention is to solve the above-mentioned conventional problems, the present invention requires efficient management of resources such as CPU, memory, network bandwidth in a system for preparing a large software streaming service, and the administrator can manage the amount of resources according to the software streaming pattern If you can predict and allocate dynamically at the time of service, you can efficiently use the resources of the system, and since the pattern of the streaming software changes the page demand according to the change of time, the listed patterns are in the form of time-series data. It was developed based on what can be considered.

An object of the present invention is to classify and determine the characteristics of the streaming service software using a time series technique and to allocate and estimate the required resource amount during the service to efficiently service a large-capacity streaming service using minimum resources, and to dynamically distribute load. The present invention provides a streaming software service system using the time series technique and a method thereof.

Another object of the present invention is to find out which cluster the streaming pattern of the software belongs to through the clustering using the time series technique when new streaming software is added, and to stream the software using the time series technique to be assigned an appropriate service resource value. It is to provide a service system and a method thereof.

In order to achieve the object of the present invention as described above, a streaming software service system using a time series technique according to the present invention collects streaming patterns in a streaming service of a specific software, and normalizes the collected streaming patterns by a time series technique. A streaming software classifier that clusters the normalized streaming patterns into groups having similar characteristics to extract a representative pattern from the clusters; A resource allocator for allocating the amount of resources required according to the service situation to the extracted representative pattern to generate a load information table including a list of resource values; And when the client requests the software streaming service, the service load among the plurality of streaming service servers including the streaming software and the streaming service manager for providing the software streaming service by referring to the load information table generated by the resource allocator. Selecting a server, the load balancer for dynamically balancing the load of the streaming service according to the amount of resources allocated to the client to receive the streaming service from the server.

The streaming software service system using the time series technique according to the present invention is characterized in that the software streaming pattern uses a similar point to the time series as described above.

The streaming software service system using the time series technique according to the present invention collects streaming patterns in a streaming service of a specific software, normalizes the streaming patterns from the collected streaming patterns into a time series technique, and clustering techniques among the time series techniques. Clusters several normalized streaming patterns into groups with similar characteristics. Also, by extracting a representative pattern from the clusters, estimating the amount of resources required by the representative pattern, and applying the same to the software of the same clusters, it is possible to estimate the amount of resources required in the platform for preparing a large software streaming service.

In the streaming software service system using the time series technique according to the present invention, when software is added to a large-capacity streaming service server in a large amount, the streaming software service system may re-cluster the corresponding software to reallocate requirements for resources required for the service. In addition, when new streaming software is added, the similarity with the representative pattern representing the clusters is calculated to allocate the resource value to which the most similar representative pattern belongs to the new streaming software to estimate the server resource usage.

The load balancer included in the streaming software service system using the time series method according to the present invention can efficiently distribute the streaming service load based on the allocated resource information so that the server resource of the large-scale streaming service can be efficiently utilized.

In order to achieve the above object of the present invention, a streaming software service method using a time series technique according to the present invention includes: a first process of collecting streaming patterns in a streaming service of a specific software; A second step of normalizing the collected streaming pattern by a time series technique; Clustering the normalized streaming patterns into groups having similar characteristics using a time series technique; Extracting a representative pattern from the clustered patterns; Generating a load information table including a list of resource values by allocating an amount of resources required according to a service situation with respect to the extracted representative pattern; And requesting the software streaming service from the client, selecting a server having the least service load among the plurality of streaming service servers, and dynamically distributing the load of the streaming service according to the amount of resources allocated to the client to receive the streaming service from the server. Process; characterized by consisting of.

According to the streaming software service system and the method using the time series technique according to the present invention as described above, the load can be distributed by estimating the amount of resources required in the platform for preparing a large software streaming service, so that the minimum resources can be efficiently It can be used to build a pilot system for streaming services.

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

Referring to FIG. 1, the software streaming service system 120 according to the present invention is connected to the clients 110A and 110B through the communication network 100.

The clients 110A and 110B request the streaming software streaming service system 120 for a streaming service, and include a PC, a laptop, and the like.

The communication network 100 is a concept including a wired and wireless Internet network, and includes all communication lines and networks that can communicate with the software streaming service system 120.

The software streaming service system 120 includes a streaming software classifier 121, a resource allocator 122, a load balancer 123, and a streaming server 124.

The streaming software classifier 121 collects streaming patterns in a streaming service of a specific software, normalizes the collected streaming patterns with a clustering technique among time series techniques, and clusters the normalized streaming patterns into groups having similar characteristics. Extract the representative pattern from the clusters.

The streaming software classifier 121 clusters the normalized streaming patterns into groups having similar characteristics to extract representative patterns from the clusters, and assigns the same tag so that software belonging to the same cluster is assigned the same resource value.

The streaming software classifier 121 calculates which streaming pattern is similar to the representative patterns when new software is added, so that the software belongs to a specific cluster, and when it is necessary to perform new clustering by adding new software. Perform reclustering.

The resource allocator 122 generates a load information table including a list of resource values by allocating an amount of resources required according to a service situation with respect to the extracted representative pattern.

When the resource allocator 122 generates a load information table including a list of resource values, as shown in FIG. 3, the resource allocator 122 performs initial initialization of the resource values of the streaming software in response to the amount of resources according to the type of software and the service situation. After the initial execution, the screen is displayed to the user, and then allocated by dividing into 'busy' which requires a large number of pages and 'idle' which requires a small number of pages.

The resource allocator 122 looks at the streaming patterns of the representative patterns represented by the streaming software classifier 121 and not only the streaming services 'initial execution', 'busy', and 'idle' that recall the resource values of the streaming software, Resources may be allocated to the above-described subdivided situations. In this case, it is preferable that resource values of the software to which the representative patterns belong follow the representative pattern values.

The resource value list means resource values allocated by the resource allocator 122 for each software. The resource value list contains the type values of the software cluster and related resource values as a list. Knowing which cluster type a particular software is can tell the resource value of that software.

When the load balancer 123 requests the software streaming service from the clients 110A and 110B, the load balancer 123 refers to the streaming software 126 for providing the software streaming service by referring to the load information table generated by the resource allocator 122. Among the plurality of streaming service servers 124 including the streaming service manager 125, the server with the least service load is selected to dynamically stream the client 110A and 110B according to the amount of resources allocated to receive the streaming service from the server. Distribute the load of services.

The load values of the load information table referenced by the load balancer 123 are the sum of resource values of software currently being serviced in the corresponding server. The load balancer 123 receives the situation from the streaming service manager 125 when the service situation changes, and dynamically changes the load state of the server by using the resource values required through the resource value list.

The load balancer 123 knows through which tag the software to be serviced belongs, through the tag, and dynamically changes the load state of the server using the resource value allocated according to the type and the service situation.

The load balancer 121 uses the load information table including the resource value list to find out the resource values required by the service target software requested by the clients 110A and 110B. The streaming software 126 is serviced under the management of the streaming service manager 124 of 124.

The streaming service manager 124 notifies the load balancer 121 when the service situation changes, so that the system resources can be efficiently managed.

The streaming software service system using the time series technique according to the present invention configured as described above operates as follows.

First, referring to FIG. 2, a process of performing a streaming software service method using a time series method according to the present invention will be described.

In the present invention, first, streaming patterns are collected in a streaming service of software (S301), normalized the collected streaming patterns by a time series technique (S302), and clustering various normalized streaming patterns by using a time series technique (S303). ). In addition, the representative pattern is extracted from the clustered patterns (S304), the software type is set for the representative pattern, and the resource value required according to the service situation is estimated (S305). The estimated values are stored in a load information table format including the resource list value of FIG. 3, and each streaming software is tagged to determine which software type belongs to the streaming software (S306). Load balancing is performed through the process of referring to FIG. 5 described below using the resource value list generated or modified during the streaming service (S307).

On the other hand, in the streaming software service method using the time series technique as described above, when new software is added to the service system 120 according to the present invention, a resource allocation process will be described as follows.

If new normalized software is added, the streaming pattern of the corresponding software is input to the streaming software classifier 121 (S401). If the cumulative value of the added software is lower than the preset reclustering setting value (S402), the similarity with the representative patterns of the clusters is calculated (S403), and the same tag as the representative pattern is assigned so that the resource value of the most similar representative pattern is assigned ( In step S404, the added software cumulative value is updated (S405), and the resource value list is updated (S406).

On the other hand, if the cumulative value of the added software exceeds the preset reclustering setting value (S402), the entire streaming pattern is reclustered (S407), and the streaming software may be well classified by performing from step S303 of FIG. Make sure

Finally, the operation of the streaming software service system using the time series method according to the present invention is described through the load balancing process described above with reference to FIGS. 3 and 4 and the resource allocation process according to the addition of the software.

Referring to FIG. 5, first, when a streaming service is requested from a client 110A or 110B for a specific software (S501), the load balancer 123 loads the streaming servers 124 in the streaming software service system 120. By viewing the status value, the most suitable streaming server 1124 is allocated to serve (S502).

Subsequently, the load balancer 123 modifies the effective resource value of the assigned server according to the 'initial execution' resource value of the corresponding software (S503), and after the initial execution is completed (S504), the load balancer 123 is The valid resource value of the allocated server is modified according to the 'idle' resource value of the corresponding software (S505).

At this time, if the service is busy and 'busy' (S506) to modify the effective resource value of the assigned server according to the 'busy' resource value (S509), the step S506 is performed until the service is terminated, actually Whenever the service status changes to 'busy' and 'idle', the effective resource value of the server is dynamically updated, so that resource management can be efficiently performed.

When the service is terminated, the streaming service manager 125 in the streaming server 124 notifies the load balancer 123 (S507), so that the load balancer 123 is currently streaming server 124 The resource allocated to the server (resource value of the software) is returned to the server (modifies the effective resource value of the server) (S508).

The streaming software service system and the method using the time series technique according to the present invention described above are not limited to the service according to the above embodiment, and the present invention without departing from the gist of the present invention as claimed in the following claims. Anyone with ordinary knowledge in this field will have the technical spirit of the present invention to the extent that it can be variously modified and implemented.

1 is a block diagram of a streaming software service system using a time series technique in accordance with the present invention.

2 is a flowchart of a streaming software service method using a time series technique according to the present invention.

3 is an embodiment of a load information table of a resource value list set according to a software type and a service situation by a streaming software classifier according to the present invention.

4 is a flowchart illustrating a resource allocation process when new software is added to a streaming software service system using a time series technique according to the present invention.

5 is a flowchart illustrating an operation process of a streaming software service system using a time series technique according to the present invention.

<Explanation of symbols for the main parts of the drawings>

100: network

110A, 110B: Client

120: Software Streaming Service System

121: load balancer

122: Streaming Software Classifier

123 resource allocator

124: streaming server

125: Streaming Service Manager

126: Streaming Software

Claims (8)

A streaming software classifier that collects streaming patterns in a streaming service of the software, normalizes the collected streaming patterns by a time series technique, and clusters the normalized streaming patterns into groups having similar characteristics to extract a representative pattern from the clusters; A resource allocator for allocating the amount of resources required according to the service situation to the extracted representative pattern to generate a load information table including a list of resource values; And When the client requests the software streaming service, the server having the least service load among the plurality of streaming service servers including the streaming software and the streaming service manager for providing the software streaming service by referring to the load information table generated by the resource allocator. Selecting a load balancer for dynamically balancing the load of the streaming service according to the amount of resources allocated to the client to receive the streaming service from the server; Streaming software service system using a time series technique comprising a. The streaming software classifier of claim 1, wherein the streaming software classifier clusters the normalized streaming patterns into groups having similar characteristics to extract a representative pattern from the clusters, and the software belonging to the same cluster assigns the same tag to be assigned the same resource value. Streaming software service system using a time series technique, characterized in that. The method of claim 1, wherein the streaming software classifier calculates which streaming pattern is similar to the representative patterns when new software is added, so that the software belongs to a specific cluster, and new clustering needs to be performed by adding new software. Streaming software service system using a time series technique, characterized in that reclustering is performed if present. The resource allocator of claim 1, wherein the resource allocator generates a load information table including a list of resource values, after initializing the resource values of the streaming software according to the type of software and the amount of resources according to service conditions. Streaming software service system using a time series method, characterized in that the screen is displayed to the user, and then divided into 'busy' requesting a large number of pages and 'idle' requesting a small number of pages. Collecting streaming patterns in a streaming service of the software; A second step of normalizing the collected streaming pattern by a time series technique; Clustering the normalized streaming patterns into groups having similar characteristics using a time series technique; Extracting a representative pattern from the clustered patterns; Generating a load information table including a list of resource values by allocating an amount of resources required according to a service situation with respect to the extracted representative pattern; And When the client requests the software streaming service, the sixth process of selecting the server with the least service load among the plurality of streaming service servers and dynamically balancing the load of the streaming service according to the amount of resources allocated for the client to receive the streaming service from the server. ; Streaming software service method using a time series technique, characterized in that consisting of. The method of claim 5, wherein in the fourth process, the normalized streaming patterns are clustered into groups having similar characteristics to extract representative patterns from the clusters, and software belonging to the same cluster is assigned the same tag to be assigned the same resource value. Streaming software service method using a time series technique, characterized in that. The method of claim 5, wherein in the third step, when the new software is added, the streaming pattern is similar to any of the representative patterns, so that the software belongs to a specific cluster, and new clustering is required by adding new software. Streaming software service method using a time series technique, characterized in that to perform reclustering if there is. The method of claim 5, wherein in the fifth process, when generating a load information table including a list of resource values, the resource value of the streaming software is initialized and initially executed according to the type of software and the amount of resources according to service conditions. After showing the screen to the user, the streaming software service method using a time series technique characterized in that the allocation by dividing into 'busy' requesting a large number of pages and 'idle' requesting a small number of pages when the user requests .

Images