CN105045891A - Method and system for improving performance of sequence list, architecture, optimization method and storage apparatus - Google Patents
Method and system for improving performance of sequence list, architecture, optimization method and storage apparatus Download PDFInfo
- Publication number
- CN105045891A CN105045891A CN201510462899.3A CN201510462899A CN105045891A CN 105045891 A CN105045891 A CN 105045891A CN 201510462899 A CN201510462899 A CN 201510462899A CN 105045891 A CN105045891 A CN 105045891A
- Authority
- CN
- China
- Prior art keywords
- performance
- sequence list
- tree construction
- value
- trie tree
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/22—Indexing; Data structures therefor; Storage structures
- G06F16/2228—Indexing structures
- G06F16/2246—Trees, e.g. B+trees
Abstract
The invention provides a method and a system for improving the performance of a sequence list, an improved HBase architecture, a data storage apparatus and an optimization method for a data storage structure of a memory. The method for improving the performance of the distributed sequence list improves the performance of the distributed sequence list by improving an HB<+>-Trie tree structure, and comprises: step 1, selecting an optimal segmentation value for the HB<+>-Trie tree structure, dividing a key value into different segments according to the optimal segmentation value, and performing hierarchical storage on the different segments to form nodes of the HB<+>-Trie tree structure; and step 2, setting a threshold value for each node of the HB<+>-Trie tree structure, and not performing downward extension when the number of data pieces in the node is smaller than the corresponding threshold value. Therefore, the insertion performance, deletion performance and query performance of an existing distributed sequence list system can be improved and the utilization rate of the memory can be reduced.
Description
Technical field
The present invention relates to distributed information processing field, particularly a kind of method, system, the HBase framework of improvement, data storage device and internal storage data storage organization optimization method improving distributed sequence list performance.
Background technology
Existing distributed sequence list commonly uses memory data organization structure B
+the structures such as tree, LSM tree.Adopt B
+the system of tree construction becomes a B by being stored in internal memory with the data construct in disk
+tree, according to user's request by data in magnetic disk swapping in and out internal memory.Adopting the system of LSM tree construction data to be divided into internal memory and disk two parts, is a B in internal memory
+tree construction, data write operation all acts on the B of internal memory
+in tree, whole internal memory write disk is then formed a small documents after exceeding threshold value by memory size.In addition, the internal memory of existing distributed sequence list and disk often adopt the structure storing data of key-value, and each row all can store a key, causes the redundant storage of key.
There is following shortcoming in existing distributed sequence list:
(1) existing distributed sequence list is based upon based on disk storage, memory is on auxiliary basis.Day by day increase at internal memory, data major part is when being even all stored in internal memory, insertion performance, the query performance of existing system are comparatively low.
(2) there is bulk redundancy in data storage, and memory usage is lower.First, existing distributed sequence list system does not carry out necessary compression to the key of data.Secondly, existing system commonly uses key-value data structure storing data, causes the redundant storage of key.
In the research carrying out distributed sequence list memory data structure optimisation technique, inventor find the data organizational structure of existing distributed sequence list system to be directly applied in large internal memory even complete in leave effect unsatisfactory.Especially system inserts performance, deletion performance, query performance and memory usage aspect is all lower.Found by research, cause the main cause of this situation to have 2 points: 1, data organizational structure does not fully compress the common prefix of key, cause system low to key index efficiency; 2, carried out repeatedly storing to key in key-value storage organization, reduced memory usage.HB
+-Trie tree construction can solve the high problem of degraded performance and memory usage in internal memory to a certain extent, and this structure carries out being divided into different sections according to fragmentation value to key, carries out level storage to different segmentations.But there are two prominent questions in this structure: 1, the select permeability of fragmentation value; 2, when application exists hot spot data, the data structure tilt problem that this structure occurs.
Patent documentation 1 (publication number is CN104268709A) discloses a kind of rfid system method for designing adopting distributed LSM to set, the trend that RFID REID to present data volume in the recent period and obviously increases, form mass data gradually, possess the feature of large data.Realize for the rfid system set based on distributed LSM, this fundamental idea of the invention is: on distributed data platform Hadoop, carry out magnanimity RFID data redundant storage, and the B replacing tradition conventional
+tree index structure, uses a kind of novel data directory structure-LSM set, makes full use of its process data edge based on internal memory, carries out the real-time verification storage of magnanimity RFID data.This invention on Hadoop, uses LSM to set index replace traditional B
+tree index is to improve the process advantage of internal memory.But this invention still also exists the problem that fully cannot promote the handling property of LSM tree in internal memory.
Summary of the invention
The object of the invention is to, solve inquiry that above-mentioned existing scheme exists under large memory conditions, deletion, query performance is low and memory usage is low problem.Inventor is on the basis studying available data structure and distributed sequence list system, propose a kind of method, system, the HBase framework of improvement, data storage device and the internal storage data storage organization optimization method that improve distributed sequence list performance, the method for the distributed sequence list performance of described raising is used for HB
+-Trie tree construction improves, and inventor is by the HB after improvement
+-Trie tree construction is defined as BHB
+-Trie tree construction, and adopt this BHB
+the memory data organization structure of-Trie tree construction to HBase system is optimized, thus reaches the object that raising system inserts performance, deletion performance, query performance and memory usage.
The method of the distributed sequence list performance of raising of the present invention, to HB
+-Trie tree construction improves, and comprising: step 1, is HB
+-Trie tree construction selects best fragmentation value, according to described best fragmentation value, key assignments key is divided into different sections, and carries out level storage to different segmentations thus form described HB
+the node of-Trie tree construction; Step 2 is described HB
+each node sets threshold value of-Trie tree construction, in node, data number is less than corresponding threshold value and does not then expand downwards.
The method of the distributed sequence list performance of raising of the present invention, wherein, in described step 1, the computing method of described best fragmentation value c are as follows:
Wherein, c
putfor HB
+the optimum of-Trie tree construction inserts the fragmentation value of performance, c
getfor HB
+the optimum query performance fragmentation value of-Trie tree construction, c
memfor HB
+the fragmentation value that the memory cost of-Trie tree construction is minimum.
The method of the distributed sequence list performance of raising of the present invention, wherein, in described step 2, the computing method of described threshold value t are as follows:
Wherein, c is best fragmentation value, and l is the average length of key assignments key, and n is total data number.
The method of the distributed sequence list performance of raising of the present invention, wherein, the method also comprises: the internal storage data storage mode key-value revising distributed sequence list, is the public key assignments key of all row storages one in same a line.
In addition, the present invention also provides a kind of system improving distributed sequence list performance, and described system is passed through HB
+-Trie tree construction carries out improvement to improve distributed sequence list performance, comprising: fragmentation value selects module, for being HB
+-Trie tree construction selects best fragmentation value, according to described best fragmentation value, key assignments key is divided into different sections, and carries out level storage to different segmentations thus form described HB
+the node of-Trie tree construction; Node B threshold selects module, for being described HB
+each node sets threshold value of-Trie tree construction, in node, data number is less than corresponding threshold value and does not then expand downwards.
The system of the distributed sequence list performance of raising of the present invention, wherein, described fragmentation value is selected in module, and the computing method of described best fragmentation value c are as follows:
Wherein, c
putfor HB
+the optimum of-Trie tree construction inserts the fragmentation value of performance, c
getfor the optimum query performance fragmentation value of HB+-Trie tree construction, c
memfor HB
+the fragmentation value that the memory cost of-Trie tree construction is minimum.
The system of the distributed sequence list performance of raising of the present invention, wherein, described Node B threshold is selected in module, and the computing method of described threshold value t are as follows:
Wherein, c is best fragmentation value, and l is the average length of key assignments key, and n is total data number.
The present invention also provides a kind of HBase framework of improvement, its HB using above-mentioned arbitrary described method to improve
+-Trie tree construction replaces the B of HBase in internal memory
+tree construction.
In addition, the present invention also provides a kind of data storage device, adopts to comprise above-mentioned any one and improve the system of distributed sequence list performance.
In addition, the present invention also provides a kind of internal storage data storage organization optimization method, comprises above-mentioned any one and improves the method for distributed sequence list performance.
The present invention and existing HB
+-Trie tree is compared, and automatically can select fragmentation value; And for preventing data skew, be each node sets threshold value, in node, data number is less than threshold value and does not expand downwards.
The total technique effect of the present invention is:
(1) based on to HB
+the improvement of-Trie tree construction, proposes BHB
+-Trie tree construction, it can select the optimal segmentation value of tree construction automatically, makes system insert performance, query performance and Installed System Memory expense and totally reaches best.
(2) BHB
+-Trie tree construction can select the threshold value of node data number automatically, decreases the impact of hot spot data on system performance.
(3) improve system to insert performance, delete performance and query performance.
(4) memory usage is improved.
Accompanying drawing explanation
Fig. 1 is the process flow diagram of the method for the distributed sequence list performance of raising of the present invention.
(A), (B) of Fig. 2, (C) are in distributed sequence list respectively, fragmentation value and insert performance, fragmentation value and query performance, graph of a relation between fragmentation value and memory cost.
Fig. 3 adopts the method for the distributed sequence list performance of raising of the present invention to HB
+figure after-Trie tree construction improves, i.e. BHB
+-Trie tree construction.
Fig. 4 is the schematic diagram of modifying to the key-value internal storage data storage mode of distributed sequence list of embodiments of the invention.
Fig. 5 (A) is existing HBase Organization Chart, and Fig. 5 (B) is the HBase Organization Chart after improvement of the present invention.
Fig. 6 is the pie graph of the system of the distributed sequence list performance of raising of the present invention.
Fig. 7 is the schematic diagram of data storage device of the present invention.
Description of reference numerals
The system of the distributed sequence list performance of 1 raising
11 fragmentation values select module
12 Node B threshold select module
2 data storage devices
Embodiment
In order to make object of the present invention, technical scheme and advantage clearly understand, below in conjunction with accompanying drawing, the HBase framework of the method for the distributed sequence list performance of raising of the present invention, system, improvement, data storage device and internal storage data storage organization optimization method are further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.
The method of the distributed sequence list performance of raising of the present invention, needs to use the experimental data that can embody application data characteristic to test distributed sequence list system, system according to test result based on HB
+-Trie tree construction selects best fragmentation value and node data number threshold value.As shown in Figure 1, the method for the distributed sequence list performance of raising of the present invention, by HB for the process flow diagram of the method for the distributed sequence list performance of raising of the present invention
+-Trie tree construction carries out improvement to improve distributed sequence list performance, the HB after improvement
+-Trie tree construction is defined as BHB
+-Trie tree construction, described method comprises: step 1, is existing HB
+-Trie tree construction selects best fragmentation value, according to described best fragmentation value, key assignments key is divided into different sections, and carries out level storage to different segmentations thus form described HB
+the node of-Trie tree construction; Step 2 is described each node sets threshold value, and in node, data number is less than corresponding threshold value and does not then expand downwards.
First, step 1 is specifically described.Inventor uses TPC-H data research fragmentation value and inserts performance, relation between query performance and memory cost, and result of study is respectively as shown in Fig. 2 (A), (B), (C).Result of study shows, exist different fragmentation values make respectively system insert performance the highest, query performance is the highest and memory cost is minimum.Suppose that the fragmentation value making system insert best performance is c
put, make the fragmentation value of query performance optimum be c
get, the fragmentation value making Installed System Memory expense minimum is c
mem, make the fragmentation value of entire system optimum be then have
Due to c
put, c
getand c
membetween three, numerical value is close, and Installed System Memory expense is at c
memleft side amplitude of variation is comparatively large, right side amplitude of variation comparatively relaxes.Therefore, inventor adopts linear normalization model, to c
put, c
getand c
memdistribute the proportion of 25%, 25% and 50% respectively, and result is rounded up process.The normalization computing method of best fragmentation value are as shown in formula (1).
By selecting best fragmentation value automatically and reasonably, system can be made to insert performance, query performance and memory cost and comprehensively to reach optimum efficiency, before avoiding, being carried out the drawback arranged by programmer based on experience value.
Consider that the existence of hot spot data can cause existing HB
+in-Trie tree, some node data number is many, some node data number is few, thus causes the unbalanced problem of data structure.Step 2 of the present invention provides node data number Threshold selection scheme.Node data number crosses that I haven't seen you for ages has a strong impact on system performance, and reduces Installed System Memory utilization factor.For solving this problem, can be each Node configuration threshold value, and lower one deck when only having the data number of node to be greater than this number of threshold values, could be expanded.This method just avoids the problem of Sparse in data structure, and can with the operation automatic expansion of system.The system of selection of threshold value:
Inventor is according to existing HB
+the scheme that the feature of-Trie tree and this programme are selected fragmentation value, has extrapolated the formula of calculated threshold, and has proved this effect by experiment.If optimal threshold is t, data total number is n, the average length of key assignments key is l, the value of optimization segment value ChunkSize is c, so then there is following formula
Making system performance reach optimum fragmentation value can by formula (1) reckoning herein out.Now system is divided into l/c section, BHB
+-Trie tree construction also just divides in order to l/c layer, and so the average data number of per node on average is the result of calculation in formula (2).
By arranging optimal threshold, make system that hot spot data can be avoided the impact of system performance and memory usage.
Fig. 3 adopts the method for the distributed sequence list performance of raising of the present invention to HB
+figure after-Trie tree construction improves, i.e. BHB
+the figure of-Trie tree construction.The BHB that the present invention proposes
+-Trie tree construction, at HB
+on the basis of-Trie tree construction, add fragmentation value selection function for automatically selecting best fragmentation value, and Node B threshold selection function is used for for each node adds data number threshold value.
In addition, the method for the distributed sequence list performance of raising of the present invention, also comprises: the internal storage data storage mode key-value revising distributed sequence list, is the public key assignments key of all row storages one in same a line.Namely the data store organisation key-value that distributed sequence list is conventional is improved, to decrease the redundant storage to key, scheme after improvement as shown in Figure 4, on the left of Fig. 4, form is the data store organisation key-value that the distributed sequence list used before improving is commonly used, this chart illustrate only the situation that there are two different key assignments in data store organisation key-value, certainly in practical situations both, data store organisation key-value usually comprises more key assignments key, key1 and key2 in figure represents key assignments 1 and key assignments 2 respectively, c1, c2, c3, v1, v2, v3, v4, v5, v6 represents row 1 respectively, row 2, row 3, value 1, value 2, value 3, value 4, value 5, value 6, as can be seen from the figure, , in database, data line often has a lot of row, the corresponding key of each row in original scheme, this just causes the repeatedly storage of key, thus cause data redundancy too high, Installed System Memory utilization factor reduces.And form is the data store organisation of the distributed sequence list after the present invention improves on the right side of Fig. 4, this programme be all row storages public key in same a line with reduce key storage redundancy, improve Installed System Memory utilization factor.By this improvement, the redundant storage of system to key can be reduced, significantly improve memory usage.。
In addition, the present invention also provides a kind of HBase framework of improvement, wherein HBase (HadoopDatabase) be a high reliability, high-performance, towards row, telescopic distributed memory system, utilize HBase technology can erect large-scale structure storage cluster on cheap PCServer.
Fig. 5 (A) for the original Organization Chart of HBase, Fig. 5 (B) be the framework that the present invention applies in HBase.In the original framework of HBase, adopting the system of LSM tree construction data to be divided into internal memory and disk two parts, is a B in internal memory
+tree construction, data write operation all acts in the B+ tree of internal memory, and memory size is then by whole internal memory write disk after exceeding threshold value, and thus comprise multiple B+ tree construction in disk, in accompanying drawing, in disk, only citing depicts 3 B
+tree construction, actual conditions then may comprise more B+ tree constructions.HB after the improvement that the present invention adopts above-mentioned arbitrary described method to build
+-Trie tree construction and BHB
+-Trie tree construction replaces the B of HBase in internal memory
+tree construction.When writing with a brush dipped in Chinese ink disk, the present invention carries out disk according to original mechanism and writes with a brush dipped in Chinese ink, all the other modules and the original mechanism of HBase completely compatible.Before the present invention needs the online implementing after improvement, first with reflecting that the relation under this machine environment and application data between fragmentation value and system performance measured by the data test collection of this application, by testing to obtain c
put, c
getand c
mem, then calculate by formula (1) fragmentation value making system performance optimum, then calculate optimal threshold according to formula (2).Be finally the best fragmentation value of Operation system setting and optimal threshold, then online implementing, thus reach the object that raising system inserts performance, query performance and memory usage.
In addition, the present invention also provides a kind of system 1 improving distributed sequence list performance, and as shown in Figure 6, described system is passed through HB
+-Trie tree construction carries out improvement to improve distributed sequence list performance, comprising: fragmentation value selects module 11, for being HB
+-Trie tree construction selects best fragmentation value, according to described best fragmentation value, key assignments key is divided into different sections, and carries out level storage to different segmentations thus form described HB
+the node of-Trie tree construction; Node B threshold selects module 12, for being described HB
+each node sets threshold value of-Trie tree construction, in node, data number is less than corresponding threshold value and does not then expand downwards.
The system of the distributed sequence list performance of raising of the present invention, wherein, described fragmentation value is selected in module, and the computing method of described best fragmentation value c are as follows:
Wherein, c
putfor HB
+the optimum of-Trie tree construction inserts the fragmentation value of performance, c
getfor HB
+the optimum query performance fragmentation value of-Trie tree construction, c
memfor HB
+the fragmentation value that the memory cost of-Trie tree construction is minimum.
The system of the distributed sequence list performance of raising of the present invention, wherein, described Node B threshold is selected in module, and the computing method of described threshold value t are as follows:
Wherein, c is best fragmentation value, and l is the average length of key assignments key, and n is total data number.
In addition, the present invention also provides a kind of data storage device 2, as shown in Figure 7, adopts and comprises above-mentioned any one and improve the system of distributed sequence list performance.
In addition, the present invention also provides a kind of internal storage data storage organization optimization method, comprises above-mentioned any one and improves the method for distributed sequence list performance.
Claims (10)
1. improve a method for distributed sequence list performance, it is characterized in that, to HB
+-Trie tree construction improves, and comprising:
Step 1 is HB
+-Trie tree construction selects best fragmentation value, according to described best fragmentation value fragmentation value, key assignments key is divided into different sections, and carries out level storage to different segmentations thus form described HB
+the node of-Trie tree construction;
Step 2 is described HB
+each node sets threshold value of-Trie tree construction, in node, data number is less than corresponding threshold value and does not then expand downwards.
2. the method for the distributed sequence list performance of raising according to claim 1, is characterized in that, in described step 1, the computing method of described best fragmentation value c are as follows:
Wherein, c
putfor HB
+the optimum of-Trie tree construction inserts the fragmentation value of performance, c
getfor HB
+the optimum query performance fragmentation value of-Trie tree construction, c
memfor HB
+the fragmentation value that the memory cost of-Trie tree construction is minimum.
3. the method for the distributed sequence list performance of raising according to claim 1, is characterized in that, in described step 2, the computing method of described threshold value t are as follows:
Wherein, c is best fragmentation value, and l is the average length of key assignments key, and n is total data number.
4. the method improving distributed sequence list performance according to claim 1, it is characterized in that, the method also comprises:
Revising the internal storage data storage mode key-value of distributed sequence list, is the public key assignments key of all row storages one in same a line.
5. improve a system for distributed sequence list performance, it is characterized in that, described system is passed through HB
+-Trie tree construction carries out improvement to improve distributed sequence list performance, comprising:
Fragmentation value selects module, for being HB
+-Trie tree construction selects best fragmentation value, according to described best fragmentation value, key assignments key is divided into different sections, and carries out level storage to different segmentations thus form described HB
+the node of-Trie tree construction;
Node B threshold selects module, for being described HB
+each node sets threshold value of-Trie tree construction, in node, data number is less than corresponding threshold value and does not then expand downwards.
6. the system of the distributed sequence list performance of raising according to claim 5, is characterized in that, described fragmentation value is selected in module, and the computing method of described best fragmentation value c are as follows:
Wherein, c
putfor HB
+the optimum of-Trie tree construction inserts the fragmentation value of performance, c
getfor HB
+the optimum query performance fragmentation value of-Trie tree construction, c
memfor HB
+the fragmentation value that the memory cost of-Trie tree construction is minimum.
7. the system of the distributed sequence list performance of raising according to claim 6, is characterized in that, described Node B threshold is selected in module, and the computing method of described threshold value t are as follows:
Wherein, c is best fragmentation value, and l is the average length of key assignments key, and n is total data number.
8. the HBase framework improved, is characterized in that, the HB improved by the method for the distributed sequence list performance of described raising arbitrary in Claims 1-4
+-Trie tree construction replaces the B of HBase in internal memory
+tree construction.
9. a data storage device, is characterized in that, adopts and comprises in claim 5 to 7 that any one improves the system of distributed sequence list performance.
10. an internal storage data storage organization optimization method, its spy is, comprises in Claims 1-4 that any one improves the method for distributed sequence list performance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510462899.3A CN105045891B (en) | 2015-07-31 | 2015-07-31 | Improve sequence list performance methodology, system, framework, optimization method and storage device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510462899.3A CN105045891B (en) | 2015-07-31 | 2015-07-31 | Improve sequence list performance methodology, system, framework, optimization method and storage device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105045891A true CN105045891A (en) | 2015-11-11 |
| CN105045891B CN105045891B (en) | 2018-08-31 |
Family
ID=54452438
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510462899.3A Active CN105045891B (en) | 2015-07-31 | 2015-07-31 | Improve sequence list performance methodology, system, framework, optimization method and storage device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105045891B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106682184A (en) * | 2016-12-29 | 2017-05-17 | 华中科技大学 | Light-weight combination method based on log combination tree structure |
| CN108628892A (en) * | 2017-03-21 | 2018-10-09 | 北京京东尚科信息技术有限公司 | Method, apparatus, electronic equipment and the readable storage medium storing program for executing of ordered data storage |
| CN108710661A (en) * | 2018-05-11 | 2018-10-26 | 武汉轻工大学 | A kind of data managing method and device for realizing text file based on memory sequence list |
| CN112667644A (en) * | 2021-01-20 | 2021-04-16 | 浪潮云信息技术股份公司 | Hybrid index memory database storage engine management method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101866358A (en) * | 2010-06-12 | 2010-10-20 | 中国科学院计算技术研究所 | Multidimensional interval querying method and system thereof |
| CN102521405A (en) * | 2011-12-26 | 2012-06-27 | 中国科学院计算技术研究所 | Massive structured data storage and query methods and systems supporting high-speed loading |
| EP2854033A1 (en) * | 2013-07-26 | 2015-04-01 | Huawei Technologies Co., Ltd. | Data recovery method, data recovery device, and distributed storage system |
-
2015
- 2015-07-31 CN CN201510462899.3A patent/CN105045891B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101866358A (en) * | 2010-06-12 | 2010-10-20 | 中国科学院计算技术研究所 | Multidimensional interval querying method and system thereof |
| CN102521405A (en) * | 2011-12-26 | 2012-06-27 | 中国科学院计算技术研究所 | Massive structured data storage and query methods and systems supporting high-speed loading |
| EP2854033A1 (en) * | 2013-07-26 | 2015-04-01 | Huawei Technologies Co., Ltd. | Data recovery method, data recovery device, and distributed storage system |
Non-Patent Citations (1)
| Title |
|---|
| 刘云生: ""利用HB+树的复合索引问题的解决方案"", 《计算机工程》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106682184A (en) * | 2016-12-29 | 2017-05-17 | 华中科技大学 | Light-weight combination method based on log combination tree structure |
| CN106682184B (en) * | 2016-12-29 | 2019-12-20 | 华中科技大学 | Lightweight merging method based on log merging tree structure |
| CN108628892A (en) * | 2017-03-21 | 2018-10-09 | 北京京东尚科信息技术有限公司 | Method, apparatus, electronic equipment and the readable storage medium storing program for executing of ordered data storage |
| CN108628892B (en) * | 2017-03-21 | 2020-11-20 | 北京京东尚科信息技术有限公司 | Method and device for storing ordered data, electronic equipment and readable storage medium |
| CN108710661A (en) * | 2018-05-11 | 2018-10-26 | 武汉轻工大学 | A kind of data managing method and device for realizing text file based on memory sequence list |
| CN108710661B (en) * | 2018-05-11 | 2020-10-27 | 武汉轻工大学 | Method and device for realizing data management of text file based on memory sequence list |
| CN112667644A (en) * | 2021-01-20 | 2021-04-16 | 浪潮云信息技术股份公司 | Hybrid index memory database storage engine management method |
| CN112667644B (en) * | 2021-01-20 | 2023-02-21 | 浪潮云信息技术股份公司 | Hybrid index memory database storage engine management method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105045891B (en) | 2018-08-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9189506B2 (en) | Database index management | |
| CN105320775A (en) | Data access method and apparatus | |
| CN1979469A (en) | Index and its extending and searching method | |
| CN110147204B (en) | Metadata disk-dropping method, device and system and computer-readable storage medium | |
| EP2735978A1 (en) | Storage system and management method used for metadata of cluster file system | |
| CN110347852B (en) | File system embedded with transverse expansion key value storage system and file management method | |
| CN110168532B (en) | Data updating method and storage device | |
| CN103678491A (en) | Method based on Hadoop small file optimization and reverse index establishment | |
| CN102915278A (en) | Data deduplication method | |
| CN103176754A (en) | Reading and storing method for massive amounts of small files | |
| CN105045891A (en) | Method and system for improving performance of sequence list, architecture, optimization method and storage apparatus | |
| CN102402602A (en) | B+ tree indexing method and device of real-time database | |
| CN103838853A (en) | Mixed file system based on different storage media | |
| CN107967122A (en) | A kind of method for writing data of block device, device and medium | |
| CN111459884B (en) | Data processing method and device, computer equipment and storage medium | |
| CN103914483A (en) | File storage method and device and file reading method and device | |
| CN102819601A (en) | Information retrieval method and information retrieval equipment | |
| CN109213445A (en) | A kind of management method, management system and the relevant apparatus of storage system metadata | |
| CN104424219A (en) | Method and equipment of managing data documents | |
| CN109558456A (en) | A kind of file migration method, apparatus, equipment and readable storage medium storing program for executing | |
| KR101438667B1 (en) | Database method for b+ tree based on PRAM | |
| CN104391961A (en) | Tens of millions of small file data read and write solution strategy | |
| CN107766355B (en) | Hierarchical data management method, hierarchical data management system and instant messaging system | |
| CN103902582B (en) | A kind of method and apparatus for reducing data warehouse data redundancy | |
| CN104375782A (en) | Read-write solution for tens of millions of small file data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |