KR101023585B1 - Method for managing a data according to the frequency of client requests in object-based storage system - Google Patents

Abstract

The present invention relates to a data management method according to a client request frequency in an object-based storage system. When the request frequency of a client increases, the chunk is divided and distributed, and when the client request frequency decreases, the divided chunk is moved to some servers. In this paper, we will provide a data management method based on client request frequency in the object-based storage system.

In addition, the present invention is to provide a data management method according to the client request frequency in the object-based storage system, which is to distribute and distribute the chunks when the request frequency of the client increases, and to delete the divided chunks when the client request frequency is reduced do.

To this end, the present invention provides a data management method, comprising: determining a corresponding management range according to a management type for a data storage unit; A measuring step of measuring a client request frequency for the data storage unit; And a management step of managing the data storage unit according to a management type corresponding to the management range to which the measured client request frequency value belongs, wherein the management range includes a response IOPS calculated by the data server for user satisfaction. Input Output Per Second) value is a range formed between the first response value and the second response value, which is a response IOPS value calculated by multiplying the first response value by a predetermined fraction.

Object-based storage system, client request frequency, data management, chunking, partitioning, distributed deployment, replication

Description

METHODS FOR MANAGING A DATA ACCORDING TO THE FREQUENCY OF CLIENT REQUESTS IN OBJECT-BASED STORAGE SYSTEM}

The present invention relates to a data management method according to a client request frequency in an object-based storage system. More particularly, the present invention relates to a data management method according to a client request frequency. The present invention relates to a data management method according to the frequency of client requests in an object-based storage system, which is moved to some servers and stored with the remaining partition chunks.

In addition, according to another embodiment of the present invention, the data according to the client request frequency in the object-based storage system that duplicates and distributes the chunk when the client's request frequency increases, and deletes the divided chunk when the client's request frequency decreases. It is about a management method.

As the massive amount of data generated by various network-based applications such as e-mail and e-commerce grows rapidly, companies need to use storage systems to continuously accumulate as much data as possible. We are working hard to build. This is because data management to support an organization's critical business processes is seen as a central component of an organization's capabilities. However, the storage system technology used in the enterprise faces various problems. In other words, due to the heterogeneous Internet environment that companies already have, it is very difficult for companies to manage and utilize the information that exists in various storage repositories.

Accordingly, as a technology related to a storage system for efficiently processing a large amount of data, 'DAS (Direct Attached Storage)', 'NAS (Network Attached Storage)', and 'SAN (Storage Area Network)' have been proposed. Specifically, DAS is a structure in which a block-based storage device is directly connected to an input / output bus of a host computer through a SCSI or an IDE. The DAS provides high performance and strong data protection, but there are limitations in storage device scalability. . NAS is a structure that manages all the metadata on one server, which describes how files are stored on storage devices, to provide file services that can share data among hosts with different platforms. There is a problem that the file server is a bottleneck because the input and output is through a single server. A SAN is a structure that supports fast and highly scalable interconnects to address storage device connectivity constraints and storage device sharing issues, and to support more client hosts and storage devices, and storage applications such as file systems There is a limit to storage scalability because it requires understanding of the file system's metadata on different platforms for sharing and the need for separate expensive devices.

In order to overcome the limitations of the aforementioned DAS, NAS and SAN, an object-based storage system has been proposed. Here, the object-based storage system divides the data processing path used in the storage environment into the metadata area and the real data area, thereby integrating the concepts of the NAS and the SAN through the object interface, as well as high performance and high scalability. Meet the requirements for storage environments such as data sharing across platforms.

However, in the conventional object-based storage system, the data read / write of the client is fixed by fixedly storing the size of the chunk, which is a data storage unit of the data server, and the number of partitions regardless of the client's request frequency. write) Could not respond appropriately to the request frequency. As such, in the conventional object-based storage system, as the frequency of data read / write requests increases, a fixed input chunk (IOPS) can be distributed through I / O distribution by replicating a chunk having a fixed size and disposing it in another data server. Could provide This could improve the preservation and availability of data through replication in the conventional object-based storage system, but there is a limit to inefficient use of the storage space of the data server.

Therefore, in the object-based storage system of the related art, a technology for minimizing the waste of storage space while ensuring data preservation and availability, and proactively providing IOPS constantly while reflecting the frequency of data read / write requests from clients is proposed. Need to be more urgent.

Therefore, the prior art as described above was able to guarantee the preservation and availability of the data in the object-based storage system, but there is a problem that wastes the storage space of the data server, it is an object of the present invention to solve this problem.

Therefore, the present invention divides and distributes chunks when the client's request frequency increases, and when the client's request frequency decreases, the client requests in the object-based storage system to move the partitioned chunks to some servers and keep them together with the remaining partitioned chunks. The purpose is to provide a data management method according to the frequency.

In addition, the present invention provides a data management method according to the client request frequency in the object-based storage system, the object-based storage system that duplicates and distributes the chunk when the request frequency of the client increases, and deletes the divided chunk when the client request frequency is reduced. The purpose is.

The objects of the present invention are not limited to the above-mentioned objects, and other objects and advantages of the present invention which are not mentioned can be understood by the following description, and will be more clearly understood by the embodiments of the present invention. Also, it will be readily appreciated that the objects and advantages of the present invention may be realized by the means and combinations thereof indicated in the claims.

In order to achieve the above object, the present invention provides a data management method, comprising: determining a corresponding management range according to a management type for a data storage unit; A measuring step of measuring a client request frequency for the data storage unit; And a management step of managing the data storage unit according to a management type corresponding to the management range to which the measured client request frequency value belongs, wherein the management range includes a response IOPS calculated by the data server for user satisfaction. Input Output Per Second) value is a range formed between the first response value and the second response value, which is a response IOPS value calculated by multiplying the first response value by a predetermined fraction.

delete

In the present invention as described above, if the request frequency of the client increases, the chunk is divided and distributed, and if the request frequency of the client decreases, the partitioned chunk is moved to some servers and stored together with the remaining partitioned chunks to store the data server. There is an effect that space can be used efficiently.

In addition, the present invention has an effect that the storage space of the data server can be efficiently used by replicating the chunks when the request frequency of the client increases and then distributing them, and deleting the divided chunks when the request frequency of the client decreases.

BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, features and advantages of the present invention will become more apparent from the following detailed description of the present invention when taken in conjunction with the accompanying drawings, It can be easily carried out. In the following description, well-known functions or constructions are not described in detail since they would obscure the invention in unnecessary detail. Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of an embodiment of an object-based storage system to which the present invention is applied.

As shown in FIG. 1, an object-based storage system to which the present invention is applied includes at least one client 110 requesting a connection to data, a client server 120 relaying a request of the client 110, and data. Metadata server 130 for locating data based on information received from client server 120 by managing metadata describing how the data is stored, and at least two or more data servers 140 storing actual data. ), The compressed data server 150 compresses and stores separate data for space saving of the data server 140.

The object-based storage system of the present invention may be applied as a low-cost, high availability mass storage system such as a server based computing service, an IPTV service, an email service, a web storage service, and the like.

Hereinafter, a process of a user performing access to data stored in the object-based storage system of the present invention through the client 110 will be briefly described with reference to FIG. 1.

First, the client 110 requests the client server 102 to access arbitrary data (1).

Next, the client server 120 analyzes the request of the client 110 and requests the metadata server 130 for information about the location of the corresponding data (②). In this case, the client server 120 communicates with the client 110 through an interface environment such as 'web service' or 'poxix'.

Next, the metadata server 130 determines the location of the data in the data server 140, and then allows the connection between the client 110 and the data server 140, according to the policy of the operator 110 and Direct connection between the data server 140 is made, or a connection between the client 110 and the data server 140 is made through the client server 120 (③).

Next, as the client 110 is directly connected to the data server 140 or connected to the data server 140 via the client server 120, the client 110 accesses corresponding data stored in the data server 140 and manages data. Perform a function (eg, save, delete, modify, view, etc.) (④). At this time, when the storage function and the delete function by the client 110 is performed, the data server 140 notifies the metadata server 130 of the result of the execution. This can be done immediately on a policy basis or at the latest after a successful implementation.

Meanwhile, the compressed data server 150 compresses and stores inaccessible data in a predetermined period of time among the data of the data server 140 based on a least recently used (LRU) algorithm, thereby providing availability and preservation of the data. While increasing the storage space of the data server 140 can be saved. Here, the compressed data server 150 is directly connected to the data server 140 or indirectly through the metadata server 130.

In particular, the storage system of the present invention manages data stored in the data server 140 for each type according to a request frequency of the client 110. To this end, in the storage system of the present invention, if the request frequency increases according to the request frequency per unit time of the client 110 (that is, the request frequency of read / write), each partitioned chunk in which the chunk is divided into a predetermined size is divided into data servers. (140) to apply distributed management type, or if the request frequency decreases, move each partitioned chunk to some server and apply the management type to keep the remaining partitioned chunks together (see FIG. 2 to be described later) 'or' A method of applying a management type for distributing each of the replication chunks in which the chunks are replicated in a predetermined multiple when the request frequency increases to the data server 140, or applying a management type that deletes some of the replication chunks when the request frequency decreases. 3) to provide a constant input output per second (IOPS) from the data server 140 to the client 110, and to improve the storage and viewing speed for the data. Can.

To this end, in the storage system of the present invention, a criterion for determining whether to divide or replicate the chunk of the data server 140 (that is, to determine how much to divide or replicate the chunk when the request frequency of the data server 140 increases to some extent). For the user's satisfaction in the data server 140 [hereinafter referred to as "first response IOPS (Ap)"), and whether to divide or delete the divided chunks of the data server 140 as a reference. As a criterion (ie, a criterion for determining whether to merge or delete a split chunk when the request frequency of the data server 140 is reduced to a certain degree), a predetermined decimal number (for example, 0.1 in the first response IOPS (Ap)) is used. A decimal number in the range of ˜0.5, which is described for convenience in the present invention as 0.5, that is, IOPS '(hereinafter referred to as “second response IOPS (Tp)”) calculated by multiplying by multiplying by Tp = Ap / 2 is set. Preferably, the predetermined decimal number for calculating the second response IOPS (Tp) is set to 0.5 or less to prevent excessively performing the merge (or delete the duplicate chunk) after moving the split chunk. Do it.

In addition, in the storage system of the present invention, 'request IOPS' for the data server 140 (hereinafter referred to as "request IOPS (Dp)") as a value corresponding to a request frequency per unit time for a specific chunk in the client 110. Measure

Accordingly, in the storage system of the present invention, by determining in which range the 'request IOPS (Dp)' falls within the management range determined between the 'first response IOPS (Ap)' and the 'second response IOPS (Tp)', Data stored in the data server 140 is variably managed according to a management type.

First, according to the request frequency of the client 110 of the present invention, when the request frequency increases, each partitioned chunk that divides the chunk into a predetermined size is distributed to the data server 140, and when the request frequency becomes low, each partitioned chunk Will be described in detail with reference to FIG.

2 is a flowchart illustrating a variable data management method according to a client request frequency in an object-based storage system according to the present invention.

First, in the object-based storage system according to the present invention, the first response IOPS (Ap) and the second response IOPS (Tp) are set in advance (S201). Here, the derivation process of the first response IOPS (Ap) will be described later in detail with reference to FIGS. 4 and 5.

Thereafter, the storage system measures a request IOPS Dp for a specific chunk of the data server 140 (S202). Here, the request IOPS (Dp) may be measured at the request of the system administrator or periodically measured. In addition, the request IOPS (Dp) may be measured as a value of a measurement moment or an average value calculated for at least one day or more.

Thereafter, the storage system determines which range the request IOPS Dp falls within a range formed between the first response IOPS (AP) and the second response IOPS (Tp) (S203). At this time, in the management range determined between the first response IOPS (Ap) and the second response IOPS (Tp), the request IOPS (Dp)> the first response IOPS (Ap), the second response IOPS (Tp) ≤ request IOPS (Dp) ≤ ≤ first response IOPS (Ap), request IOPS (Dp) <second response IOPS (Tp).

First, if the request IOPS (Dp) is greater than the first response IOPS (Ap) (Dp> Ap), the storage system increases the request frequency for a particular chunk of the data server 140, and thus, the specific chunk is 1 / n ( n is divided by at least two natural numbers) and distributed in the data server 140 (S204). In this case, the storage system checks the state of the storage capacity of the data server 140 and only sequentially distributes the divided chunks from the server with the low storage capacity, or accepts the divided chunks of the data server 140. It is possible to arrange and distribute the divided chunks arbitrarily. For example, if n is 3 and the data server 140 is equal to A, B, C, and D in the order of low storage capacity, the divided chunks are distributed to each of the data servers 140 A, B, and C. .

In addition, as a result of re-measuring the request IOPS (Dp), the storage system is a process of redistributing and splitting each partition chunk in the data server 140 when the request IOPS (Dp) is still larger than the first response IOPS (Ap). Is repeated (Dp &gt; Ap) and terminates if the request IOPS (Dp) is not less than the second response IOPS (Tp) and not greater than the first response IOPS (Ap) (Tp &lt; Dp &lt; Ap).

Next, if the request IOPS (Dp) is less than the second response IOPS (Tp) (Dp &lt; Tp), the storage system is in a state in which the request frequency for the specific partition chunk of the data server 140 is reduced, so that The set is moved to the data server 140 having the set of the remaining divided chunks and merged with the set of the remaining divided chunks (S205). At this time, since the process of merging the split chunks is performed by the data server 140 satisfying the case where the request IOPS (Dp) is smaller than the second response IOPS (Tp), the total number of split chunks is changed (odd to even number). In some data servers 140, the respective divided chunks may be combined and biased.

In addition, the storage system re-measures the request IOPS (Dp) of the data server 140, if the request IOPS (Dp) is less than the second response IOPS (Tp) (Dp <Tp), the data server ( The specific partition chunk set of 140 is moved to the data server 140 having the remaining partition chunk sets and merged. On the other hand, if the storage system re-measures the request IOPS (Dp) of the data server 140 and the request IOPS (Dp) is not smaller than the second response IOPS (Tp) and not greater than the first response IOPS (Ap) ( Tp ≤ Dp ≤ Ap).

In general, in the storage system, if the request frequency for a particular chunk of the data server 140 increases, the 'split and batch process' (S204) is repeatedly performed in a specific chunk while Dp> Ap is satisfied, and Tp≤Dp≤Ap If it is satisfied, it ends. After that, since the request frequency for the chunk decreases after a predetermined time, the storage system repeatedly performs the 'move and sum chip process' (S205) in the divided chunk while satisfying Dp <Tp, and Tp≤Dp≤Ap. Exit when satisfied.

On the other hand, according to the request frequency of the client 110 of the present invention, 'if the request frequency increases, each replication chunk that replicates the chunks in a predetermined multiple is distributed to the data server 140, and if the request frequency is lowered, some of the replication chunks The method of deleting 'will be described in detail with reference to FIG. 3.

3 is a flowchart illustrating another embodiment of a variable data management method according to a client request frequency in an object-based storage system according to the present invention.

Since each of steps S301 to S303 in FIG. 3 corresponds to steps S201 to S203 in FIG. 2, those skilled in the art may easily understand the detailed description.

As in FIG. 2, the storage system determines which range the request IOPS Dp falls within a range formed between the first response IOPS AP and the second response IOPS Tp (S303). Further, in the range formed between the first response IOPS (Ap) and the second response IOPS (Tp), the request IOPS (Dp)> the first response IOPS (Ap), the second response IOPS (Tp) ≤ the request IOPS (Dp) ≤ first response IOPS (Ap), request IOPS (Dp) <second response IOPS (Tp).

First, if the request IOPS (Dp) is greater than the first response IOPS (Ap) (Dp> Ap), the storage system is in a state where the request frequency for a particular chunk of the data server 140 is increased, so that n (n At least two natural numbers) are duplicated and distributed to the data server 140 (S304). In this case, the storage system checks the state of the storage capacity of the data server 140 and only sequentially distributes the replication chunks from the server having the low storage capacity, or only accommodates the replication chunks of the data server 140. You can check and randomly distribute the duplicate chunks. For example, if n is 3 and the data server 140 is equal to A, B, C, and D in the order of low storage capacity, the replication chunks are distributed to each of the data servers 140 A, B, and C. .

In addition, if the storage system re-measures the request IOPS (Dp) of the data server 140 and the request IOPS (Dp) is larger than the first response IOPS (Ap) (Dp> Ap), the storage system re-replicates the chunk to copy the chunk. Repeat the process of distributedly distributing to the data server 140 is not stored. At this time, the storage system may re-replicate the original chunks or re-replicate each copy chunk according to the administrator's policy. On the other hand, if the storage system re-measures the request IOPS (Dp) of the data server 140 and the request IOPS (Dp) is not smaller than the second response IOPS (Tp) and not greater than the first response IOPS (Ap) ( Tp ≤ Dp ≤ Ap).

Next, if the request IOPS (Dp) is less than the second response IOPS (Tp) (Dp <Tp), the storage system deletes the replication chunk since the request frequency for the specific replication chunk of the data server 140 is reduced ( S305). In this case, the storage system preferably checks the state of the storage capacity and deletes the replication chunk first from the data server 140 having no free space.

In addition, when the request IOPS (Dp) is smaller than the second response IOPS (Tp) as a result of re-measuring the request IOPS (Dp) of the data server 140 (Dp <Tp), the storage server satisfies the data server ( Delete only the specific replication chunk of 140). On the other hand, the storage system ends if the request IOPS Dp of the data server 140 is not smaller than the second response IOPS Tp and not greater than the first response IOPS Ap (Tp ≦ Dp ≦ Ap).

In general, in the storage system, if the request frequency for a particular chunk of the data server 140 increases, the 'clone and batch process' (S304) is repeatedly performed in a specific chunk while satisfying Dp> Ap and Tp≤Dp≤Ap. If it is satisfied, it ends. After that, since a predetermined time elapses, the request frequency for the chunk decreases. Therefore, when the partition chunk satisfies Dp &lt; Tp, the 'deletion process' (S305) is repeated and Tp &lt; Dp &lt; Quit. Hereinafter, a process of calculating the first response IOPS (Ap) will be described in detail.

4 is a flowchart illustrating a derivation process of the first response IOPS Ap.

The storage system calculates the connection response time (Rs) and the response satisfaction (S) of the data server 140 (S401, S402), through which the first response IOPS (Ap) is derived (S403).

Specifically, first, the connection response time Rs (sec) of the data server 140 is calculated using Equation 1 below (S401).

Here, π represents the maximum IOPS of the data server 140, ρ represents the double available IOPS ratio. In general, ρ is set to 50% to 80% of π.

Next, the response satisfaction degree S of the data server 140 is calculated using Equation 2 below (S402).

Here, R represents the guaranteed response time of the minimum data server 140 that is a customer service service agreement (SLA) matter. R can be specified separately according to the user's data connection request frequency and required IOPS. In general, the total IO response time per user request on the web is required within 2 seconds.

Finally, the first response IOPS (Ap) of the data server 140 is derived using Equation 3 below (S403).

Where Κ is the number of subscriber requests per second, Φ is IOPS per chunk, τ (bytes) is the initial chunk size, and N is the optimal block size. In particular, the maximum IOPS (π) and the optimal block size (N) of the aforementioned data server 140 are derived through FIG. 5 to be described later.

FIG. 5A is a flowchart illustrating the derivation process of the maximum IOPS (π) and the optimal block size (N).

In order to derive the maximum IOPS (π) and the optimal block size (N), the storage system transmits the measurement data corresponding to a multiple of the unit block size (for example, when it increases by 2 times) to the data server 140. Afterwards, the inner product (product) of the 'maximum write IOPS' and the 'block size' of the measured data are used. At this time, the primary storage system (M ₂ -M _3), (M ₂ -M ₁₎ between the three inner product (multiplication) consecutive M _1, M _2, and when M _3, two adjacent inner product (product) Comparisons are made with each other to derive maximum IOPS (π) and optimal block size (N). In the present invention, in order to derive the maximum IOPS (π), a better inner product can be calculated in the case of using a sequential distribution of data and read IOPS, but a poor situation can be solved by using the maximum write IOPS. Assume In general, write IOPS and read IOPS are proportional.

Hereinafter, the derivation process of the maximum IOPS (π) and the maximum block size (N) will be described in detail.

First, the storage system selects the data server 140 for measuring the maximum IOPS (π), and sets the unit block size B ₁ as a storage unit of the data server 140 (S501). At this time, the storage system selects any data server when the characteristics of the entire data server 140 are the same, or selects a specific data server when it is not. In addition, the storage system may be variously set as a unit block size such as 1K, 2K, 4K, etc., which may be determined by the storage system administrator in a policy manner.

Next, the storage system transmits the measurement data corresponding to one unit block size (N ₁ = B ₁ ) to the data server 140 selected in step S501 and then measures the IOPS (I ₁ ) of the measurement data. Calculate and record M ₁ (= B ₁ * I ₁ ), which is the inner product (product) of the block size B ₁ of the measured data and the measured IOPS I ₁ (S502). At this time, the storage system records the corresponding measurement data at an arbitrary position of the data server 140.

Next, the storage system transmits the measurement data corresponding to twice the previous measurement data (ie, N ₂ = 2 * B ₂ ) to the data server 140 selected in step S501 and then IOPS (I ₂ ) of the measurement data. ) And calculate and record M ₂ (= N ₂ * I ₂ ), which is the inner product (product) of N ₂ , the block size of the measured data, and I ₂ , the measured IOPS (S503).

Next, the storage system transmits the measurement data corresponding to twice the previous measurement data (N ₃ = 2 * N ₂ ) to the data server 140 selected in step S501 and then outputs IOPS (I ₃ ) of the measurement data. Measure and calculate and record M ₃ (= N ₃ * I ₃ ), which is the inner product (product) of N ₃ , the block size of the measured data, and I ₃ , the measured IOPS (S504).

On the other hand, the storage system is a product of three consecutive products (that is, M ₁ , M ₂₎ And M ₃ ) (S502 to S504), the difference between successive inner products 'the difference between M ₁ and M ₂ (ie, M ₂ -M ₁ ) (S ₁ )' and the difference between M ₂ and M ₃ ( That is, M ₃ -M ₂ ) (S ₂ ) 'are compared (S505). In this case, when S ₁ > S ₂ , the storage system derives the most recently measured IOPS 'I ₃ ' as the maximum IOPS (π), and the maximum block size (N) as the block size of the most recently measured data. derives a ₃ N / 2 in N _{2 (i.e., N 3 = 2 * N 2} ) (S506). On the other hand, if S ₁ <S ₂ , the storage system calculates and records M ₄ , that is, the data selected in step S501 for measurement data corresponding to twice the previous measurement data (N ₄ = 2 * N ₃ ). the M ₄ server then transferred to 140 measure the IOPS (I ₄₎ of the measurement data, and the block of the sample data size of N ₄ and the IOPS of the inner product (multiplication) of the I ₄ measurements (= N ₄ * I ₄ ) Is calculated and recorded (S507).

Afterwards, the storage system is called 'difference between M ₂ and M ₃ (ie M ₃ -M ₂ ) (S ₂ )' and 'difference between M ₃ and M ₄ (ie M ₄ -M ₃ ) (S ₃ )' (S508), if S ₂ > S ₃ , the storage system derives the most recently measured IOPS as 'I ₄ ' as the maximum IOPS (π), and most recently measured data as the maximum block size (N). is derived by the block size of N ₄ / N 2 of ₃ (that _{is, N 4 = 2 * N 3} ) (S509). On the other hand, if S ₂ <S ₃ , the storage system performs a process of calculating M ₅ again and then compares the difference between inner products again (S510). The following process is a process that can be easily inferred by those skilled in the art by referring to the same process as steps S508 to S511, and a detailed description thereof will be omitted.

In particular, the storage system is a storage system {S ₁ ,... The maximum IOPS (π) and the optimal block size (N) are derived when S _{k -1} is the maximum among the set of, S _k } (a natural number with k≥2). Assuming that the storage system derives the maximum IOPS (π) and the optimal block size (N) in the aforementioned S509 phase, wherein the set of drive made between subsequent inner product is _{{S 1, S 2, S} 3} because S _1, S S ₂ is the maximum at ₂ and S ₃ . That is, in order to derive the maximum IOPS (π) and the optimal block size (N) in step S509, S ₁ <S ₂ and S ₂ > S ₃ must be satisfied, so that S ₁ <S ₂ > S ₃ and Similarly, S ₂ should be maximum.

Therefore S510 after the step, the storage system is repeated to perform a process, such as steps S507 to S510 until the step S _{k -1} from the difference between the set of successive dot S _k up then the maximum IOPS (π) and the optimal block size (N) is derived.

FIG. 5B is a detailed flowchart of FIG. 5A.

The storage system can specifically implement the derivation process of the maximum IOPS (π) and the maximum block size (N) as shown in FIG. 5B.

As mentioned above, the storage system selects the data server 140 from which the maximum IOPS (π) and the maximum block size (N) will be derived, and the unit block size (B ₁ ), which is the basic storage unit of the data server 140. Select (S551).

Next, the storage system is measured record IOPS (I ₁₎ by sending the measured data (B ₁₎ corresponding to one unit block size, the inner product (multiplication) of the block size (B ₁₎ and IOPS (I ₁₎ M ₁ (= B ₁ * I ₁ ) is recorded (S552).

Next, the storage system measures and records IOPS (I ₂ ) by transmitting measurement data (N ₂ = 2 * B ₁ ) corresponding to twice the unit block size, and measures block size (N ₂ ) and IOPS (I _2). M ₂ (= N ₂ * I ₂ ), which is the inner product of (), is recorded (S553). At this time, the storage system inserts the difference between M ₂ and M ₁ into S _max (= M ₂ -M ₁ ) and substitutes M ₂ into M _N-1 (S554).

Next, the inner product of the storage system is the measurement data (N = ₂ * N 2) transmits the IOPS measured record (I _N), and the block size (N) and IOPS (I _N) for corresponding to the unit block size of four times M _N (= N * I _N ), which is a product, is recorded (S555). At this time, the storage system substitutes M _N and M _N-1 to S _N (= M _N -M _N-1 ) (S556).

On the other hand, the storage system S _N and S _max of the S _N is a large (S557), assigns the S _N to S _max than S _max, and (S558), re-S555 step by substituting the M _N to M _N-1 compared Repeat the process of comparing S _N with the previous S _max by doubling the block size by moving. Here, the storage system compares S _N and S _max and if S _N is smaller than S _max (S557), the maximum block size is 1/2 of the previous N and the maximum IOPS is I _N (S559).

6A is an exemplary diagram for chunk partitioning and distributed storage according to the present invention.

As shown in FIG. 6A, the storage system divides chunk 1 into half and then distributes data to data server 1 (that is, chunk 1_1) and data server 3 (that is, chunk 1_2) that belong to data server 140, respectively. After dividing the chunk 2 into 1/2, the chunk 2 is divided and stored in the data server 2 (that is, the chunk 2_1) and the data server 4 (that is, the chunk 2_2) belonging to the data server 140, respectively.

Figure 6b is an illustration of chunking and concentrated storage according to the present invention.

As shown in FIG. 6B, the storage system combines the chunk 1_1 of the data server 1 and the chunk 1_2 of the data server 3 and concentrates the data in the chunk 1 on the data server 1 due to a change in user's access frequency and response time. The chunks 2_1 of the server 2 and the chunks 2_2 of the data server 4 are combined, and the data are stored in the chunk 2 of the data server 2. In this case, the storage system may store the chunk 1 and the chunk 2 after combining the compressed data server 150 as described above.

Figure 7a is an exemplary view for chunk replication and distributed storage according to the present invention.

As shown in FIG. 7A, the storage system replicates Chunk 1 twice and distributes the data to Data Server 1 (that is, Chunk 1) and Data Server 3 (that is, Chunk 1) belonging to the data server 140, respectively. After replicating the chunk 2 twice, the data is distributed and stored in the data server 2 (that is, the chunk 2) and the data server 4 (that is, the chunk 2) belonging to the data server 140, respectively.

Figure 7b is an illustration of chunk deletion in accordance with the present invention.

As shown in FIG. 7B, the storage system deletes the chunk 1 of the data server 3 and the chunk 2 of the data server 4 due to a change in the user's access frequency and response time.

On the other hand, the method of the present invention as described above can be written in a computer program. And the code and code segments constituting the program can be easily inferred by a computer programmer in the art. In addition, the written program is stored in a computer-readable recording medium (information storage medium), and read and executed by a computer to implement the method of the present invention. The recording medium may include any type of computer readable recording medium.

The present invention described above is capable of various substitutions, modifications, and changes without departing from the technical spirit of the present invention for those skilled in the art to which the present invention pertains. It is not limited by the drawings.

1 is a configuration diagram of an embodiment of an object-based storage system to which the present invention is applied;

2 is a flowchart illustrating a variable data management method according to a client request frequency in an object-based storage system according to the present invention;

3 is a flowchart illustrating another embodiment of a variable data management method according to a client request frequency in an object-based storage system according to the present invention;

4 is a flowchart illustrating a derivation process of a first response IOPS (Ap);

5a is a flowchart illustrating a process of deriving a maximum IOPS (π) and an optimal block size (N);

5b is a detailed flowchart of FIG. 5a;

6a is an exemplary view of chunk division and distributed storage according to the present invention;

Figure 6b is an exemplary view for chunking and concentrated storage according to the present invention,

Figure 7a is an exemplary view for chunk replication and distributed storage according to the present invention,

Figure 7b is an illustration of chunk deletion in accordance with the present invention.

* Explanation of symbols for the main parts of the drawings

110: client 120: client server

130: metadata server 140: data server

150: compressed data server

Claims (9)

delete In the data management method, A decision step of determining a corresponding management range according to a management type for the data storage unit; A measuring step of measuring a client request frequency for the data storage unit; And And a management step of managing the data storage unit according to a management type corresponding to the management range to which the measured client request frequency value belongs. The management range is, The first response value, which is a response IOPS (Input Output Per Second) value calculated by the data server for user satisfaction, is formed between the second response value, which is a response IOPS value calculated by multiplying the first response value by a predetermined fraction. The data management method characterized by the above-mentioned range. The method of claim 2, The first response value is, The number of subscriber requests per second and initial time for the response satisfaction time calculated using the maximum IOPS of the data server and the guaranteed response time of the data server, which is a service level agreement (SLA), Data management method calculated by applying 'size of data storage unit' and 'optimal block size'. The method of claim 3, wherein The maximum IOPS and the optimal block size is, And a maximum value in a set obtained by obtaining a difference between successive inner product values using the inner product values of the measured data transmitted to the data server in a predetermined multiple and the IOPS for the measured data. The method according to any one of claims 2 to 4, The management step, And distributing the data storage unit into at least two natural numbers when the client request frequency value is greater than the first response value. The method of claim 5, The management step, And when the client request frequency value is smaller than the second response value, move the divided data storage unit after the division to a predetermined data server and combine the data. The method according to any one of claims 2 to 4, The management step, And distributing the data storage unit by a multiple of at least two natural numbers when the client request frequency value is greater than the first response value. The method of claim 7, wherein The management step, And deleting the duplicated data storage unit when the client request frequency value is smaller than the second response value. delete

Images