WO2021107211A1

WO2021107211A1 - In-memory database-based time-series data management system

Info

Publication number: WO2021107211A1
Application number: PCT/KR2019/016673
Authority: WO
Inventors: 김학철; 한혁; 진성일
Original assignee: 주식회사 리얼타임테크
Priority date: 2019-11-27
Filing date: 2019-11-29
Publication date: 2021-06-03
Also published as: KR102102313B1

Abstract

The present invention relates to a technology in which, instead of storing time-series data in units of records, pieces of time-series data having the same attribute are divided into units of data blocks having a certain size and stored to be connected in a temporal manner, and indexes are generated and managed in units of data blocks, thereby improving the management and analysis performance of time-series data analyzed time-sequentially in an in-memory database.

Description

In-memory database-based time series data management system

The present invention does not store time series data in record units, but divides time series data of the same attribute into data blocks of a certain size and stores them in a time-connected manner, and creates and manages indexes in units of data blocks, thereby providing an in-memory database It relates to a technology that can improve the management and analysis performance of time series data analyzed chronologically in a system.

In the current rapidly changing information society, building and operating reliable information is becoming an essential condition for the survival of countries, companies, or users. In order to build such reliable information, most of the recent companies are operating more tasks by connecting them to a Relational Database Management System (RDBMS). When operating such a database system for business, high performance, high efficiency and reliability are naturally required basic conditions.

In the relational database management system, information is generated in large quantities due to cloud, mobile, and globalization of services, and has become one of the most important resources that change rapidly. In order to process such a large amount of data, an in-memory database technology has been recently proposed and introduced.

An in-memory database is a database management system that is installed and operated in the main memory of data storage, and has an advantage in that the processing speed is faster than that installed on a disk.

In particular, as the amount of data increases with the advent of big data and the requirements for fast analysis processing increase, research on in-memory databases is being conducted more actively.

On the other hand, recently, due to the development of IoT technology, etc., a large amount of time series data is being collected in real time, and such time series data is composed of a specific time and one or more data values collected at the corresponding time.

Currently, the in-memory database stores and manages time series data as multiple records composed of multiple fields in a similar way to general data, and creates and manages indexes for each record.

However, the conventional in-memory database that manages time series data in record units and creates indexes in record units causes performance degradation in consideration of the characteristics of time series data analysis.

That is, most analyzes of time series data tend to analyze data by sorting time series data in chronological order. , the processing time becomes very long.

In addition, as the time series data continuously increases by managing time series data at a specific time in record units and creating record unit indexes, the size of the index also increases continuously, resulting in a problem of index data management as well as a decrease in index performance. there is

Accordingly, the present invention was created in consideration of the above circumstances. Time series data is divided into data blocks of a certain size and sequentially stored in chronological order, and an index for time series data is created in units of data blocks, so that time series data rapidly increases Its technical purpose is to provide an in-memory database-based time series data management system that not only solves the problem of increasing the index size, but also efficiently searches time series data analyzed in chronological order using data blocks. There is this.

According to one aspect of the present invention for achieving the above object, time series data is stored in units of data blocks of a certain size, but as for time series data of the same property, a plurality of data blocks are connected in a chain form so that time series data is sequentially sequential. A time series data storage unit for storing data in the form of, and a record storage unit for storing a record table including time series data information including start time and end time for all of the time series data currently stored for each attribute information of the time series data and time series data information consisting of first data block information; One index table is stored for data blocks in which time series data of the same attribute is stored, but the index table is a time series consisting of index information for each data block including start time and end time for each data block and data block location information. Time series data storage processing of searching and storing a data index unit and a data block to store the time series data for a time series data storage query, and updating the record storage unit and the time series data index unit in response to the data block in which the time series data is stored; For a time series data search query including a time condition, the first data block information is obtained by searching the record table including the query time condition, and data block location information that satisfies the query time condition is obtained from the index table including the first data block. Including a time series data processing unit that performs time series data retrieval processing to obtain, retrieve a data block corresponding to a data block location from the time series data storage unit, and call and return time series data corresponding to a search query from the retrieved data block An in-memory database-based time series data management system is provided.

In addition, the data block stored in the time series data storage unit includes a header area and a body area in which actual time series data is stored, and the header area includes NEXT data block information and corresponding data connected to the corresponding data block in a chain form. It includes the start time of the block and the end time of the corresponding data block, and the body area is characterized in that the time series data collection time and data items in the form in which one or more data values are arranged in a preset order are sequentially stored in the time series data collection time order An in-memory database-based time series data management system is provided.

In addition, one or more time stamp items are allocated to the header area of the data block, and the time stamp items are set as collection times of data items stored in the body area, and the data item collection times set in the time stamp items are sequentially assigned to the body area. There is provided an in-memory database-based time-series data management system, characterized in that it is set as a data item collection time at a location having a predetermined interval from the stored data items.

In addition, the time series data processing unit identifies a timestamp item close to the query time condition in the header area of the data block when performing time series data search, and data that satisfies the query time condition from the data item at the timestamp position checked in the body area There is provided an in-memory database-based time series data management system, characterized in that by starting an item search and searching for time series data corresponding to a query search result.

In addition, the time series data is configured to further include a metadata storage unit for storing a meta table including information on the type of data corresponding to the data value and the arrangement order thereof, the time series data processing unit having a plurality of data values In addition to setting the storage environment of a data item consisting of a time series data collection time and data values for time series data, when a time series data search query is performed, an arrangement order corresponding to the type of data to be searched is obtained from the metadata storage unit, , The in-memory database-based time series data management system is provided, characterized in that the data value of the arrangement order is called as the time series data requested to be searched in the data item satisfying the query time condition.

In addition, the time series data processing unit is an in-memory database-based time series data management system, characterized in that it supports to set a data storage environment of time series data in a UDT (User Defined Type) method for time series data having a plurality of data values. is provided

In addition, the time series data processing unit performs a time series data search for a time series data deletion query to search for a data item in which the time series data to be deleted is stored, and then sets the collection time to “NULL” in the searched data item. There is provided an in-memory database-based time series data management system, characterized in that it performs.

In addition, the time series data processing unit performs a data packing operation for data items set to "NULL" in a predetermined period unit, but after clearing the data items set to "NULL" in the corresponding data block, copies the data items after There is provided an in-memory database-based time series data management system, characterized in that the data packing operation is performed in such a way that the data is moved to the previous data item location.

According to the present invention, analysis and management performance of time series data can be improved by dividing and storing time series data in chronological order and storing them in units of data blocks of a certain size, and creating and managing indexes in units of data blocks.

In addition, since the header area of the data block includes time series data time information for a predetermined position stored in the body area of the data block, improved time series data retrieval performance can be guaranteed.

1 is a diagram showing a schematic configuration of an in-memory database-based time series data management system according to a first embodiment of the present invention .

2 is a diagram illustrating a query for a UDT function supported by the time series data processing unit 100 shown in FIG. 1 .

3 is a diagram illustrating a meta table structure of the metadata storage unit 200 shown in FIG. 1 .

4 is a diagram illustrating a record table structure of the record storage unit 300 shown in FIG. 1;

FIG. 5 is a diagram illustrating a data block structure stored in the time series data storage unit 400 shown in FIG. 1;

FIG. 6 is a diagram illustrating an index table structure of the time series data indexing unit 500 shown in FIG. 1. FIG.

7 is a flowchart for explaining a time series data storage operation in the time series data management system based on the in-memory database shown in FIG. 1;

FIG. 8 is a flowchart for explaining a time series data search operation in the time series data management system based on the in-memory database shown in FIG. 1. FIG.

The configuration shown in the embodiments and drawings described in the present invention is only a preferred embodiment of the present invention, and does not express all the technical spirit of the present invention, so the scope of the present invention is the embodiment and drawings described in the text should not be construed as being limited by That is, since the embodiment may have various changes and may have various forms, it should be understood that the scope of the present invention includes equivalents capable of realizing the technical idea. In addition, since the object or effect presented in the present invention does not mean that a specific embodiment should include all of them or only such effects, it should not be understood that the scope of the present invention is limited thereby.

All terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs, unless otherwise defined. Terms defined in a commonly used dictionary should be interpreted as being consistent with the meaning of the context of the related art, and cannot be interpreted as having an ideal or excessively formal meaning not explicitly defined in the present invention.

1 is a diagram illustrating a schematic configuration of an in-memory database-based time series data management system according to a first embodiment of the present invention.

Referring to FIG. 1 , an in-memory database-based time series data management system according to the present invention includes a time series data processing unit 100 , a metadata storage unit 200 , a record storage unit 300 , and a time series data storage unit 400 . and a time series data indexing unit 500 .

The time series data processing unit 100 processes a query for time series data requested by the user. The time-series data processing unit 100 processes time-series data-related queries such as storage, deletion, and correction of time-series data, a search for analysis, and calculation of sum or difference.

In addition, the time series data processing unit 100 provides a UDT (User Defined Type) function for arbitrarily defining a storage structure for time series data in response to a user's request. This function can be effectively used especially when there are multiple types of data values (columns) simultaneously acquired for one time series data.

2 shows a field name for storing time series data using a User Defined Type (UDT) function defined by a user, a query statement (a) for setting a data storage environment of time series data, and a record for time series data corresponding thereto The query statement (b) for creating a table is exemplified. According to Fig. 2 (a), the time series data field type is set to "dustsensor", and for each column name of "t, v1, v2, v3", each column type of "time, int, double, double" has You can set the time series data storage environment. Then, as shown in FIG. 2(b), a table having a field name of “DustSensingValue” can be created. That is, the field name of a field in which time series data information is stored in the record table in FIG. 4 is set to "DustSensingValue".

The metadata storage unit 200 is a device for effectively storing and managing time series data, and manages metadata for time series data based on a user-defined time series data field name.

3 illustrates a meta table structure of the metadata storage unit 200 . Referring to FIG. 3 , the metadata storage unit 200 stores the number of time series data columns in units of data items and time series data by time units stored in the time series data storage unit 400 by user-defined time series data field names (UDT names). An upper table (a) in which the item size (byte) is stored and a lower table (b) in which a column name and column type for each time series data column sequence of the user-defined time series data field name (UDT name) are stored. The lower table (b) is created as many as the number of columns of time series data in the upper table (a). For example, when FIG. 2 is applied, the number of time series data columns (t, v1, v2, v3) is set to "4" for "DustSensingValue" in the upper table (a), and four lower tables (b) are created. For example, if the time series data column sequence number is "3" in the lower table, "Temperature" corresponding to "v2", which is the third column among columns (t, v1, v2, v3), is set as the column name, and the third column The third column type corresponding to "double" is set as the column type.

The record storage unit 300 stores general column information on general properties such as object ID, location, and date for each record ID (hereinafter oid) allocated to time series data and time series data information on the corresponding object. In this case, oid may be an id for a collection device that collects time series data, and general columns are stored in the form of a general relational table.

4 illustrates the record table structure of the record storage unit 300 . Referring to FIG. 4, the record table includes a general column area for each oid and a time series data area as shown in (a). And, as shown in (b), the time series data area includes the total start time, the total end time, the number of timestamps, and first data block information for the entire time series data. Here, the first data block information may be a storage location of a corresponding data block stored in the time series data storage unit 400 . In addition, the number of timestamps is the number of time information in the corresponding data block included in the header area in the data block structure of the time series data storage unit 400 in which the time series data is stored.

That is, the record storage unit 300 allocates one oid to time series data having the same property, and stores only the first data block information of the time series data storage unit 400 in which the actual time series data is stored for the oid. Therefore, even when time series data continuously increases, only pointer information for the first data block is recorded in the record table, so record management is easy.

The time series data storage unit 400 stores the actual time series data in units of data blocks of a certain size, but the time series data of the same attribute is connected in a chain form by a plurality of data blocks, and the time series data is stored in a continuous form in chronological order. Here, the time series data block means a continuous space of a predetermined size in the memory, and the time series data is arranged in chronological order in the corresponding space. At this time, when all time series data is stored in one data block, the data blocks of the same size are reallocated and information about the reallocated data blocks is registered in the previous data block table so that the data blocks are sequentially connected. is saved

5 illustrates a data block structure stored in the time series data storage unit 400 .

The data block consists of a header area and a body area where actual time series data is stored as shown in FIG. 5(a), and the header area is a NEXT data block connected to the corresponding data block in a chain form. It includes information and the corresponding data block start time and the corresponding data block end time, and the body area is such that the time series data collection time and data items in the form of one or more data values arranged in a preset order are sequentially stored in the time series data collection time order. is composed For example, if the query of FIG. 2 is applied, the column arrangement order stored in the data item is “t,v1,v2,v3”, and the time series data item in the body area is “start time (t), first sensed value (CO _{2 )”} ), the second sensed value (temperature), and the third sensed value (humidity)”.

Here, the data block start time may be the start time of the first data item stored in the body area, and the data block end time may be the start time of the last data item stored in the body area. That is, time series data stored in the body region are basically stored in ascending order with respect to time.

Also, one or more timestamp items may be allocated to the header area of the data block. The time stamp item is set as the collection time of a certain number of data items stored in the body area. The data item collection time set in the time stamp item is the data item collection time at a location with a certain interval from the data items sequentially stored in the body area. can be set to For example, the data item collection time set for the time stamp item may be set as the data item collection time stored at a multiple of the sequence number corresponding to a quotient obtained by dividing the total number of data items stored in the body region by the number of time stamp items.

In FIG. 5(b), three timestamps including a start time t1 of the first unit block, a start time t4 of the fourth unit block, and a start time t7 of the seventh unit block are exemplified. In this case, the number of timestamps may be set to correspond to the number of data items stored in the data block. For example, the number of timestamps may be increased as the size of the data block increases, that is, as the number of data items increases, in order to further improve the time-series data search speed.

That is, the time series data processing unit 100 roughly predicts the location of the time series data corresponding to the time to be searched in the data block through the timestamp information in the header area when searching the time series data, so that the time series data can be searched more quickly. Do.

The time series data indexing unit 500 is to provide a method for effectively searching the time series data stored in the time series data storage unit 400, that is, index tables for each data block in which time series data having the same property is stored, that is, oid. configurable. For example, one time series data collection device may be configured as a unit, or one time series data table may be created for all time series data collection devices included in a specific table record.

6 illustrates an index table structure of the time series data indexing unit 500 .

Referring to FIG. 6 , the index table structure stored in the time series data indexing unit 500 has a structure in which index information for each data block including start time and end time for each data block and the corresponding data block information is repeated. In this case, the data block information may be a corresponding data block location stored in the time series data storage unit 400 .

In addition, in the index table, maximum and minimum values of time series data of a corresponding data block may be further added to further improve data analysis performance. This may be appropriately referred to in response to an operator when requesting time series data analysis thereafter.

Next, an operation of the in-memory database-based time series data management system according to the present invention will be described with reference to the flowcharts shown in FIGS. 7 and 8 .

The in-memory database-based time series data management system according to the present invention basically performs a time series data storage process and a time series data search process.

7 is a flowchart illustrating a time series data storage process in an in-memory database-based time series data management system according to the present invention.

Referring to FIG. 7 , the time series data processing unit 100 sets a field name for storing time series data in the UDT method, a time series data value and an arrangement order thereof, and a storage environment of a data type for each data value (ST100). When the query of FIG. 2 is applied, the time series data processing unit 100 sets the field name in which the time series data of the record storage unit 300 is stored to “DustSensingValue” and “DustSensingValue” in the metadata storage unit 200 . Creates a meta table named UDT. And, set the column type (time, int, double, double) for each column item (t, v1, v2, v3) in the lower table among meta tables. For example, "t" in the column item is the start time at which the corresponding time series data is collected, and "v1, v2, v3" is the time series data, for example, "CO ₂ , temperature, humidity" may be matched.

Hereinafter, a state in which the columns (t, v1, v2, v3) are _{matched with "start time, CO 2} , temperature, humidity" will be described as an example.

In the above state, when a time series data storage query is received (ST120), the time series data processing unit 100 searches whether a record table for the corresponding time series data exists in the record storage unit 300 based on the attribute information. At this time, if the record table does not exist, the record table is created by allocating an oid for attribute information of the corresponding time series data (ST130).

Next, the time series data processing unit 100 searches for a data block to store the time series data using the time series data index unit 500 based on the time information included in the time series data storage query, and in the time series data storage unit 400 The corresponding time series data is stored in the searched data block position in a preset column order (t, v1, v2, v3) (ST140).

For the time series data newly registered in step ST140, the time series data processing unit 100 allocates a new data block using the time series data index unit 500, and stores the corresponding time series data in the newly allocated data block.

In addition, with respect to time series data to which data blocks have already been allocated in step ST140, data block information in which the corresponding time series data is to be stored is searched using the time series data indexing unit 500 based on time information. For example, in FIG. 6 , it is determined whether there is a storage area remaining in the last data block corresponding to the time series data, and if there is, the time series data is stored in the body area of the data block of the time series data storage unit 400 . And, if there is no storage area remaining in the last data block, after allocating a new data block, the corresponding time series data is stored in the newly allocated data block. At this time, the next data block information is set as the corresponding data block information in the header area of the previous data block.

That is, the time series data processing unit 100 stores the time series data in the allocated data block, but stores the time series data to be sorted in ascending order with respect to time.

Thereafter, the time series data processing unit 100 updates management information related to the corresponding data block based on the time series data storage information stored in the time series data storage unit 400 (ST150). That is, the time series data processing unit 100 updates the metadata storage unit 200 , the record storage unit 300 , and the time series data index unit 500 .

The time series data processing unit 100 updates the number of time series data items and the data item size in the upper table corresponding to the corresponding UDT name (“DustSensingValue”) in the metadata storage unit 200 .

In addition, at least one of the object start time, the object end time, the number of timestamps, and the first data block information is updated by reflecting the time series data information currently stored for the DustSensingValue name for the record ID in the record storage unit 300 . For example, the object start time and object end time, the number of timestamps, and first data block information are newly registered for the time series data to be newly stored. In addition, the object end time may be updated for additionally stored time series data, and if the number of time series data stored in the corresponding data block is greater than the reference number, the number of timestamps may be increased. In this case, when the number of timestamps is increased, the time series data processing unit 100 resets the timestamp position set in the header area of the corresponding data block according to the number of timestamps. For example, if the number of data stored in one data block is 10 and the number of timestamps (refer to FIG. 4(b)) is “3”, the timestamp position is “ When the number of data increases to 11 in the state set to t1, t4, t7", the number of timestamps (refer to (b) of FIG. 4) is changed to "4" and the time series data table (Fig. 5 (b) )), you can set the timestamp position to "t1, t3, t6, t8".

In addition, the time series data indexing unit 500 updates at least one of start time, end time, and lower node information for the corresponding record ID. For example, a start time, an end time, and a lower node are newly registered for the time series data to be newly stored. And, for the additionally stored time series data, the start time or the end time for the corresponding sub-node is updated, and at least one of the start time, the end time, the maximum value, the minimum value, and the data block information is updated in the corresponding sub-node information. .

Thereafter, the time series data processing unit 100 returns a time series data storage completion query (ST160).

8 is a flowchart illustrating a time series data search process in an in-memory database-based time series data management system.

Referring to FIG. 8 , in a state in which time series data is stored through the process as shown in FIG. 7 , when a time series data search query including a time condition is received ( ST210 ), the time series data processing unit 100 analyzes the search query to obtain metadata The storage unit 200 extracts the storage sequence number of the column name corresponding to the data type (column) to be searched (ST220). That is, the time series data processing unit 100 extracts the meta table based on the UDT name included in the search query, and checks the batch sequence number of the column name in the extracted meta table. For example, it can be confirmed that "Temperature", which is the type of time series data requested to be queried, is the second storage sequence (v2) in the table whose UDT name is "DustSensingValue".

Next, the time series data processing unit 100 searches the data block index information including the query time condition in the total start time and the total end time from the record storage unit 300, and retrieves the first data block information from the searched data block index information. Acquire (ST230).

Then, the time series data processing unit 100 searches the index table including the first data block in the time series data index unit 500 based on the first data block information obtained in step ST230, but starting from the highest node table to the lower node. A data block that satisfies the query time condition is searched for while sequentially moving to the table (ST240). That is, from each table shown in Fig. 6, the sub-node information in which the query time condition is located between the start time and the end time is extracted, and the query time condition is located again between the start time and the end time from the sub-node information. Acquire data block location information.

Next, the time series data processing unit 100 searches for a corresponding data block in the time series data storage unit 400 based on the data block location information obtained from the time series data index unit 500 , and approaches the query time condition in the searched data block. The location of the data item to be used is searched (ST250).

In this case, the time series data processing unit 100 may search for a data location corresponding to the query time condition more quickly by using the timestamp stored in the header area of the data block as shown in FIG. 5 .

For example, when the query time condition in FIG. 5 is between “t4” and “t7”, the time series data processing unit 100 may search for the query time condition from the position t4 in the body area of the corresponding data block table.

In the present invention, it is also possible to set the number of timestamps in the data block table arbitrarily by the user. This is to set the start time of the time series data search in the data block table to a position close to the query time condition, and can be set to increase the number of timestamps.

Then, the time series data processing unit 100 calls the time series data value corresponding to the batch sequence extracted in step ST220 from the corresponding data item of the body region corresponding to the searched query time condition as the target query time series data (ST260). For example, in the data item satisfying the time condition of the data block, the second time series data value is called as the temperature value of the query.

In this case, when the time condition is within a certain range, all query target time series data satisfying the time condition corresponding to the corresponding range are extracted in the same manner as described above.

Thereafter, the time series data processing unit 100 returns the called time series data as a search result (ST270).

Also, when the number of data blocks searched for in step ST240 is two or more, the time series data processing unit 100 may obtain time series data for each data block in a parallel processing method according to the type of query. For example, when calculating the sum of time series data between times t1 and t2, time series data may be obtained in a parallel manner for each data block, and then the query result may be calculated by summing the data.

Meanwhile, in the present invention, according to a request from the outside, the time series data processing unit 100 may perform a deletion query on the time series data stored in the time series data storage unit 400 .

That is, the time series data processing unit 100 searches for the data item in which the time series data to be deleted is stored by performing the time series data search described above for the time series data deletion query, and then sets the time information (t) in the searched data item to "NULL" You can perform a delete query in the way you set it. Thereafter, the time series data processing unit 100 recognizes that the data item for which time information is found as “NULL” is deleted.

In this case, the time series data processing unit 100 may check a timestamp for a data block included in a data item set to "NULL", and may change timestamp information as necessary. For example, the timestamp information may be changed when a data item set to “NULL” is set as a timestamp or when a data item set to “NULL” is a predetermined number or more in a corresponding data block.

In addition, the time series data processing unit 100 collectively performs a data packing operation for data items set to "NULL" in units of a predetermined period. That is, the time series data processing unit 100 performs a data packing operation in such a way that after clearing a data item set to “NULL” in a data block, the data item after the data item is copied and moved to the position of the previous data item. Since this is an operation that moves data in a block unit in memory, performance issues do not occur significantly.

Thereafter, the time series data processing unit 100 updates all information related to the data block, including the header of the data block on which the data packing operation is performed, if necessary.

Claims

A time series data storage unit for storing time series data in units of data blocks of a certain size, but for time series data of the same attribute, a plurality of data blocks are connected in a chain form to store the time series data in a chronological sequence;

A record storage unit for storing a record table including time series data information consisting of start time and end time for all of the time series data currently stored for each attribute information of the time series data and first data block information;

One index table is stored for data blocks in which time series data of the same attribute is stored, but the index table is a time series consisting of index information for each data block including start time and end time for each data block and data block location information. a data index;

Time series data storage processing of searching for and storing a data block to store the time series data for a time series data storage query, and updating the record storage unit and the time series data index unit in response to the data block in which the time series data is stored, including a time condition For a time series data search query, the first data block information is obtained by searching the record table including the query time condition, the data block location information satisfying the query time condition is obtained from the index table including the first data block, and time series data and a time series data processing unit that searches for a data block corresponding to a data block position in the storage unit, and performs time series data search processing to call and return time series data corresponding to a search query from the searched data block. In-memory database-based time series data management system.
According to claim 1,

The data block stored in the time series data storage unit consists of a header area and a body area in which actual time series data is stored,

The header area contains information on the next data block that is connected to the data block in a chain form, the start time of the data block and the end time of the data block, and the body area contains the time series data collection time and one or more data values. An in-memory database-based time series data management system, characterized in that data items arranged in a preset order are sequentially stored in the time series data collection time order.
3. The method of claim 2,

One or more timestamp items are allocated to the header area of the data block, and the time stamp items are set to a collection time of data items stored in the body area,

In-memory database-based time series data management system, characterized in that the data item collection time set in the time stamp item is set as the data item collection time at a location having a predetermined interval from the data items sequentially stored in the body area.
4. The method of claim 3,

The time series data processing unit identifies a timestamp item close to the query time condition in the header area of the data block when performing time series data search, and searches for data items satisfying the query time condition from the data item at the timestamp position checked in the body area In-memory database-based time series data management system, characterized in that by starting the search for time series data corresponding to the query search results.
3. The method of claim 2,

It is configured to further include a metadata storage unit for storing a meta table including information on a data type corresponding to a data value for time series data and an arrangement order thereof,

The time series data processing unit sets a storage environment for a data item including a time series data collection time and data values for time series data having a plurality of data values, and searches the metadata storage unit when performing a time series data search query An in-memory database-based time series data management system, characterized in that a batch order corresponding to a target data type is acquired, and a data value of the batch order is called as the time series data requested to be retrieved from a data item that satisfies a query time condition.
6. The method of claim 5,

The time series data processing unit is an in-memory database-based time series data management system, characterized in that it supports to set the data storage environment of the time series data in a UDT (User Defined Type) method for time series data having a plurality of data values.
According to claim 1,

The time series data processing unit performs a time series data search for a time series data deletion query to search for a data item in which the time series data to be deleted is stored, and then performs a deletion query in a manner that sets the collection time to “NULL” in the searched data item. In-memory database-based time series data management system, characterized in that.
8. The method of claim 7,

The time series data processing unit performs a data packing operation for data items set to "NULL" in a predetermined period unit,

In-memory database-based time series data management system, characterized in that after clearing the data item set to "NULL" in the corresponding data block, the data packing operation is performed by copying the data item after it and moving it to the position of the previous data item. .