CN104391931A - Method for efficiently indexing mass data in cloud computing - Google Patents
Method for efficiently indexing mass data in cloud computing Download PDFInfo
- Publication number
- CN104391931A CN104391931A CN201410673448.XA CN201410673448A CN104391931A CN 104391931 A CN104391931 A CN 104391931A CN 201410673448 A CN201410673448 A CN 201410673448A CN 104391931 A CN104391931 A CN 104391931A
- Authority
- CN
- China
- Prior art keywords
- node
- access
- version number
- server
- distributed
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 78
- 230000008569 process Effects 0.000 claims description 8
- 230000003139 buffering effect Effects 0.000 claims description 6
- 238000003780 insertion Methods 0.000 claims description 5
- 230000037431 insertion Effects 0.000 claims description 5
- 230000008520 organization Effects 0.000 claims description 3
- 230000001360 synchronised effect Effects 0.000 claims description 3
- 230000008859 change Effects 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000004992 fission Effects 0.000 description 2
- 238000010923 batch production Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000007620 mathematical function Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
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
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2455—Query execution
- G06F16/24553—Query execution of query operations
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Multi Processors (AREA)
Abstract
The invention discloses a method for efficiently indexing mass data in cloud computing, which adopts logs to record splitting histories of nodes on the basis of a distributed B tree and efficiently accesses the distributed B tree concurrently on the basis of the splitting histories of the nodes, so that the access and indexing efficiency of mass and distributed data in a cloud computing environment is effectively improved. The method of the invention is superior to the transaction method because in each access, the transaction method adds distributed locks to all nodes on the path from the root to the leaf node to be accessed, and the method of the invention only needs to lock the leaf node to be accessed; the method of the present invention is superior to the link method because the number of nodes traversed by the method of the present invention is less than that of the link method.
Description
Technical field
The present invention relates to field of cloud calculation, be specifically related to a kind of mass data efficient index method in cloud computing.
Background technology
To occur in current cloud computing environment a large amount of based on mass data, the internet, applications that provides various information to serve.
The data of these application have the features such as magnanimity, fast growth.System by the method for the key word of each data being carried out to Hash by all Data distribution8 in multiple memory node, store with the easily extensible realizing the mass data increased fast.Hash method has greater efficiency to employing keyword query, but can not improve the efficiency to adopting non-keyword inquiry, does not also support range query.And concerning user, except adopting keyword query data, also like carrying out data query with other attributes or carrying out range query.Such as, in an Online Video system (as Youtube), each video packets is containing information such as video ID, programm name, uplink time, broadcasting times.ID is the key word of each video, can obtain video fast by video ID, but user also inquires about video with programm name or uplink time scope usually.Build the important method that secondary index is the search efficiency improving non-key attribute.In current cloud computing environment, common secondary index is inverted index (invertedindex), and it is scanned all memory nodes by multiple MapReduce process, produces inverted file.Inverted index process is the batch process of an off-line, and the data of up-to-date insertion can not be inquired about in time.Such as, in GoogleBase, insert a record, the next times such as needs again after index (time may be needed) could be inquired by user.In sum, cloud computing environment needs the indexing means supporting online index building and range query.
B-tree indexed had both supported online index building, supported range query again, but due to the magnanimity of application data in cloud computing environment, causing centralized B to set (the B tree of setting up in stand-alone environment) can not the storage demand of satisfying magnanimity index data.Distributed B tree by all B tree nodes are distributed to the storage demand each storage server solving magnanimity index data, and improves Concurrency Access ability by the mode setting internal node at all client buffer B.But it sets the same with centralized B, still there is Concurrency Access problem, namely when inserting in (or deletion) data to distributed B tree, multiple ancestor node divisions (or merging) of a leaf node and this leaf node may be caused, thus cause the access concurrent with it not to be successfully completed.
Current solution B sets Concurrency Access topic and mainly contains two kinds of methods, i.e. transaction method and link method.
1) transaction method: it is when accessing at every turn, need distributed transaction pin from root to the path of a leaf node all nodes, cause all access all will pin root node, to the queuing of root node lock B set amendment access all become
Serial access, causes access efficiency lower.
2) Link method: it solves Concurrency Access problem by the mode setting up link between node, solves the lower problem of concurrent capability that transaction method exists.But it is not suitable for the distributed B tree in cloud computing environment, because existing distributed B sets to adopt improve Concurrency Access ability in the mode of all its internal nodes of client buffer, and the method postponing to upgrade is adopted to share buffer joint renewal cost.Under delay generation patterns, a client is in the secondary access time of same node, in server end, corresponding node may to divide repeatedly, according to Distribution Strategy, these nodes newly split off can be distributed in other memory nodes, therefore subsequent access all needs to travel through a lot of node in multiple server, affects access efficiency.
Hash, the result of the fixed size obtained by the data that unidirectional mathematical function (being sometimes referred to as " hash algorithm ") is applied to any amount.If changed in input data, then Hash also can change.Hash can be used for many operations, comprises authentication and digital signature.Also referred to as " eap-message digest ".
The binary value of random length is mapped as the less binary value of regular length by hash algorithm, and this little binary value is called cryptographic hash.Cryptographic hash be one piece of data uniquely and extremely compact numeric representation form.If hash one section of plaintext and even only change a letter of this paragraph, Hash subsequently all will produce different values.Finding hash to be the input that two of same value are different, is computationally impossible, so the cryptographic hash of data can the integrality of check data.
Summary of the invention
The technical problem to be solved in the present invention is: for magnanimity and the distributivity feature of data in cloud computing environment, and there is the lower problem of rate of people logging in existing distributed b-tree indexed method, the present invention proposes a kind of mass data efficient index method in cloud computing environment, on the basis that distributed B sets, adopt daily record to record the division history of node, and based on the distributed B tree of the efficient Concurrency Access of node split history, effectively improve magnanimity in cloud computing environment, distributed data access and index efficiency.
The technical solution adopted in the present invention is:
A kind of mass data efficient index method in cloud computing, described method adopts daily record to record the division history of node on the basis that distributed B sets, and based on the distributed B tree of the efficient Concurrency Access of node split history, effectively improve magnanimity in cloud computing environment, distributed data access and index efficiency.
Described method particular content is as follows:
Every station server is provided with a B tree node division daily record, for recording the division history of all B tree nodes be distributed in this server;
Division history be one in chronological sequence order sequence log file, each division of node is all charged in daily record as a record;
Interrecord structure in daily record is: < LowValue, UpValue, Pointer, Version, preRecord >, wherein LowValue and UpValue is minimum value in index node and maximal value respectively, and they have recorded the scope of the value that node stores; Pointer is a pointer, and point to the memory location of split vertexes, it comprises two codomains, i.e. index node storage server address IP and the position of index node in this server; Version is the version number of index node, is mainly used in judging when accessing to want access division daily record, and index node often divides once, and the version of two split vertexes adds 1 on the version basis of its father node; PreRecord is a pointer, and it once divides record before pointing to; Divide the log recording division situation of all nodes, by preRecord, the division history of each node is connected into a chained list;
After a node is divided into two nodes, one of them node can be distributed in an other station server, and along with data constantly increase, the node split off still constantly divides, and is distributed in different servers;
Each node has division daily record, by these division daily record cascades, the node organization split off is become a tree.Its accessing operation needs the quantity of access services device only relevant with the height of tree.Therefore, SINC drastically increases access efficiency.
The inventive method adopts the mode of version number to judge whether the node stored in buffer joint and server there occurs change, in distributed B tree, each node of server end is provided with a version number, and the internal node cushioned in each client also preserves the version number of corresponding node; Postpone update strategy owing to adopting, the version number of buffering is less than the version number of server node sometimes, and this explanation is not also synchronized to the corresponding node in this buffer zone to the amendment of server interior joint;
Division history in conjunction with version number and node realizes distributed B height of tree Concurrency Access, method is: if the version number of access node is identical with the version number of the client buffer sending access, just this node locked and access this node, otherwise just access the division daily record of this node, access node and the version number thereof of down hop is found from corresponding division record, and the version number of request of access and node is passed to next-hop node, go down with this, until find data or there is not next-hop node.
Based on said method, set insertion < key at B, the flow process of pointer > becomes:
A) in client, the B tree that the internal node searching buffering is formed, obtains < key, the position of the leaf node f that pointer > should insert and version number v thereof;
B) server of request of access to leaf node f place is sent;
C) in the server, obtain the version number v ' of node f, if v '=v, proceed to d), otherwise proceed to e;
D) in home server, pin node f, inserted by < key, pointer > in node f, operation terminates;
E) search daily record, find < key, the position of the leaf node f that pointer > should insert and version number v thereof, forward to b).
Beneficial effect of the present invention: the magnanimity and the distributivity feature that the present invention is directed to data in cloud computing environment, and there is the lower problem of access efficiency in existing distributed b-tree indexed method, a kind of mass data efficient index method in cloud computing environment is proposed, it adopts daily record to record the division history of node on the basis that distributed B sets, and based on the distributed B tree of the efficient Concurrency Access of node split history.Thus effectively improve magnanimity, distributed data access and index efficiency in cloud computing environment.
Accompanying drawing explanation
Fig. 1 is the structure of node split daily record;
Fig. 2 is the tree that split vertexes is formed;
Fig. 3 is the fission process of node a;
Fig. 4 is link organizational form;
Fig. 5 is the organizational form of the inventive method;
Fig. 6 is update example.
Embodiment
Below according to Figure of description, in conjunction with specific embodiments, the present invention is further described:
A kind of mass data efficient index method in cloud computing, described method adopts daily record to record the division history of node on the basis that distributed B sets, and based on the distributed B tree of the efficient Concurrency Access of node split history, effectively improve magnanimity in cloud computing environment, distributed data access and index efficiency.
Described method particular content is as follows:
Every station server is provided with a B tree node division daily record, for recording the division history of all B tree nodes be distributed in this server;
Division history be one in chronological sequence order sequence log file, its structure as shown in Figure 1, all charge in daily record as a record by each division of node;
Interrecord structure in daily record is: < LowValue, UpValue, Pointer, Version, preRecord >, wherein LowValue and UpValue is minimum value in index node and maximal value respectively, and they have recorded the scope of the value that node stores; Pointer is a pointer, and point to the memory location of split vertexes, it comprises two codomains, i.e. index node storage server address IP and the position of index node in this server; Version is the version number of index node, is mainly used in judging when accessing to want access division daily record, and index node often divides once, and the version of two split vertexes adds 1 on the version basis of its father node; PreRecord is a pointer, and it once divides record before pointing to; Divide the log recording division situation of all nodes, by preRecord, the division history of each node is connected into a chained list;
After a node is divided into two nodes, one of them node can be distributed in an other station server, and along with data constantly increase, the node split off still constantly divides, and is distributed in different servers;
Each node has division daily record, by these division daily record cascades, the node organization split off is become a tree, as shown in Figure 2.Its accessing operation needs the quantity of access services device only relevant with the height of tree.Therefore, the inventive method drastically increases access efficiency.
Such as, suppose the fission process of a node a as shown in Figure 3, so as shown in Figure 4, organizational form of the present invention as shown in Figure 5 for the link organizational form of the rear each node of division.If the pointed a111 of certain client buffer, and data are in a222, so in link method, and travel through 8 nodes from a111 and just can obtain data, need the inventive method only to need traversal four times.
The inventive method adopts the mode of version number to judge whether the node stored in buffer joint and server there occurs change, in distributed B tree, each node of server end is provided with a version number, and the internal node cushioned in each client also preserves the version number of corresponding node; Postpone update strategy owing to adopting, the version number of buffering is less than the version number of server node sometimes, and this explanation is not also synchronized to the corresponding node in this buffer zone to the amendment of server interior joint;
Division history in conjunction with version number and node realizes distributed B height of tree Concurrency Access, method is: if the version number of access node is identical with the version number of the client buffer sending access, just this node locked and access this node, otherwise just access the division daily record of this node, access node and the version number thereof of down hop is found from corresponding division record, and the version number of request of access and node is passed to next-hop node, go down with this, until find data or there is not next-hop node.
Based on said method, set insertion < key at B, the flow process of pointer > becomes:
A) in client, the B tree that the internal node searching buffering is formed, obtains < key, the position of the leaf node f that pointer > should insert and version number v thereof;
B) server of request of access to leaf node f place is sent;
C) in the server, obtain the version number v ' of node f, if v '=v, proceed to d), otherwise proceed to e;
D) in home server, pin node f, inserted by < key, pointer > in node f, operation terminates;
E) search daily record, find < key, the position of the leaf node f that pointer > should insert and version number v thereof, forward to b).
Such as, in figure 6, client1 inserts < key1, and during pointer1 >, the buffered version inserting node is identical with the version in server, just completes update in this locality.And client2 inserts < key2, during pointer2 >, because the buffered version inserting node is different with the version in server, therefore access log, the leaf node obtaining being inserted into is in server2, request of access is handed to server2, in server2, completes update.
As can be seen from above-mentioned insertion flow process, only need pin when revising B tree node this leaf node that will access, avoiding each access and will lock root node to this paths of access node; Equally, if the update of leaf node causes internal node to revise, also only this node need be pinned when revising certain internal node just passable.Therefore the method substantially increases concurrency.
It is that transaction method will add distributed lock to root to all nodes on this paths of the leaf node that will access, and the inventive method only need lock the leaf node that will access because in each access that the inventive method is better than transaction method; And to be better than link method be that the nodes that will travel through because of the inventive method is less than link method.
Experiment embodiment:
Complete in 126 computing machines of experiment in four racks, these machines connect into network by gigabit ethernet switch, are about 14Gb ps in rack to half-band width, are about 6.5Gbps between rack to half-band width.Every platform computing machine You Liangge 2.8GHz Intel is to strong CPU, 4GB internal memory, the SCSI hard disk of two 10000 turns per minute.All computer run is the RedHat Enterprise Linux AS4.0 of 2.6.9 at kernel version.
Each node of B tree preserves 175 couples of < key, pointer > value, a < key, pointer > is to accounting for 22Byte, wherein key accounts for 10 Byte, and pointer accounts for 22 Byte.Pointer is made up of two parts: IP address (4Byte) and side-play amount (8Byte).Before the experiment, in B tree, existing 400 nodes, 6.4 ten thousand < key, pointer > couple, be distributed in four station servers.The node of B tree is placed on server end, and load is produced by client, and server end and client lay respectively in different computing machines; Every station server is for B tree provides 32MB buffer zone in internal memory, and each client computer runs four threads simultaneously, their access same B tree.
In order to assess the performance of the inventive method, in same experimental situation, also tested transaction method and link method, to be used for performance comparison.In an experiment, load is produced by client computer, independent of the server storing B tree node; Each client computer runs four threads simultaneously.The load that thread produces has three classes:
A) insert type load.Operation in load is all update, and the key in each update produces at random from 0 ~ 105 integer space.
B) type load is searched.Operation in load is all query manipulation, and the key in each query manipulation selects randomly from the key set of B tree.
C) mixed type load.Operation in load is made up of update and search operation, and wherein update accounts for 30%, and search operation accounts for 70%, and the producing method of the key in two kinds of operations is with identical above.
Fix at nodes, server and client computer be when increasing, the test result of above-mentioned three class loads draws: no matter in which kind of load, the inventive method is all better than link method and transaction method.It is that transaction method will add distributed lock to root to all nodes on this paths of the leaf node that will access, and the inventive method only need lock the leaf node that will access because in each access that the inventive method is better than transaction method; And the inventive method is better than link method is that the nodes that will travel through because of the inventive method is less than link method.
Online index and range query that cloud computing needs to adopt distributed B to set supports the mass data increased fast.Experimental result shows, context of methods significantly improves the access efficiency that distributed B sets.
Claims (3)
1. a mass data efficient index method in cloud computing, is characterized in that: described method adopts daily record to record the division history of node on the basis that distributed B sets, and based on the distributed B tree of the efficient Concurrency Access of node split history.
2. mass data efficient index method in a kind of cloud computing according to claim 1, is characterized in that:
Every station server is provided with a B tree node division daily record, for recording the division history of all B tree nodes be distributed in this server;
Interrecord structure in daily record is: < LowValue, UpValue, Pointer, Version, preRecord >, wherein LowValue and UpValue is minimum value in index node and maximal value respectively, and they have recorded the scope of the value that node stores; Pointer is a pointer, and point to the memory location of split vertexes, it comprises two codomains, i.e. index node storage server address IP and the position of index node in this server; Version is the version number of index node, is mainly used in judging when accessing to want access division daily record, and index node often divides once, and the version of two split vertexes adds 1 on the version basis of its father node; PreRecord is a pointer, and it once divides record before pointing to; Divide the log recording division situation of all nodes, by preRecord, the division history of each node is connected into a chained list;
After a node is divided into two nodes, one of them node can be distributed in an other station server, and along with data constantly increase, the node split off still constantly divides, and is distributed in different servers;
Each node has division daily record, by these division daily record cascades, the node organization split off is become a tree;
In distributed B tree, each node of server end is provided with a version number, and the internal node cushioned in each client also preserves the version number of corresponding node; Postpone update strategy owing to adopting, the version number of buffering is less than the version number of server node sometimes, and this explanation is not also synchronized to the corresponding node in this buffer zone to the amendment of server interior joint;
If the version number of access node is identical with the version number of the client buffer sending access, just this node locked and access this node, otherwise just access the division daily record of this node, access node and the version number thereof of down hop is found from corresponding division record, and the version number of request of access and node is passed to next-hop node, go down with this, until find data or there is not next-hop node.
3. mass data efficient index method in a kind of cloud computing according to claim 2, is characterized in that: set insertion < key at B, the flow process of pointer > becomes:
A) in client, the B tree that the internal node searching buffering is formed, obtains < key, the position of the leaf node f that pointer > should insert and version number v thereof;
B) server of request of access to leaf node f place is sent;
C) in the server, obtain the version number v ' of node f, if v '=v, proceed to d), otherwise proceed to e;
D) in home server, pin node f, inserted by < key, pointer > in node f, operation terminates;
E) search daily record, find < key, the position of the leaf node f that pointer > should insert and version number v thereof, forward to b).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410673448.XA CN104391931A (en) | 2014-11-21 | 2014-11-21 | Method for efficiently indexing mass data in cloud computing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410673448.XA CN104391931A (en) | 2014-11-21 | 2014-11-21 | Method for efficiently indexing mass data in cloud computing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104391931A true CN104391931A (en) | 2015-03-04 |
Family
ID=52609835
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410673448.XA Pending CN104391931A (en) | 2014-11-21 | 2014-11-21 | Method for efficiently indexing mass data in cloud computing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104391931A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105868398A (en) * | 2016-04-20 | 2016-08-17 | 国网福建省电力有限公司 | Method for low-overhead data transmission of distributed file system based on Fat-B tree |
| CN106156126A (en) * | 2015-04-08 | 2016-11-23 | 阿里巴巴集团控股有限公司 | Process the data collision detection method in data task and server |
| CN113204435A (en) * | 2021-07-01 | 2021-08-03 | 阿里云计算有限公司 | Data processing method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102063466A (en) * | 2010-12-03 | 2011-05-18 | 中国科学院软件研究所 | Database forced access control method and system based on policy |
| US8037102B2 (en) * | 2004-02-09 | 2011-10-11 | Robert T. and Virginia T. Jenkins | Manipulating sets of hierarchical data |
-
2014
- 2014-11-21 CN CN201410673448.XA patent/CN104391931A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8037102B2 (en) * | 2004-02-09 | 2011-10-11 | Robert T. and Virginia T. Jenkins | Manipulating sets of hierarchical data |
| CN102063466A (en) * | 2010-12-03 | 2011-05-18 | 中国科学院软件研究所 | Database forced access control method and system based on policy |
Non-Patent Citations (1)
| Title |
|---|
| 黄斌，彭宇行，彭小宁: "云计算中海量数据高效索引方法", 《计算机应用研究》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106156126A (en) * | 2015-04-08 | 2016-11-23 | 阿里巴巴集团控股有限公司 | Process the data collision detection method in data task and server |
| CN106156126B (en) * | 2015-04-08 | 2019-10-11 | 阿里巴巴集团控股有限公司 | Handle the data collision detection method and server in data task |
| CN105868398A (en) * | 2016-04-20 | 2016-08-17 | 国网福建省电力有限公司 | Method for low-overhead data transmission of distributed file system based on Fat-B tree |
| CN105868398B (en) * | 2016-04-20 | 2019-04-26 | 国网福建省电力有限公司 | A kind of distributed file system low overhead data transmission method based on Fat-B tree |
| CN113204435A (en) * | 2021-07-01 | 2021-08-03 | 阿里云计算有限公司 | Data processing method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8880502B2 (en) | Searching a range in a set of values in a network with distributed storage entities | |
| CN102591970B (en) | Distributed key-value query method and query engine system | |
| US20130191523A1 (en) | Real-time analytics for large data sets | |
| US20080086464A1 (en) | Efficient method of location-based content management and delivery | |
| US20070112795A1 (en) | Scalable retrieval of data entries using an array index or a secondary key | |
| CN104408111A (en) | Method and device for deleting duplicate data | |
| CN104077423A (en) | Consistent hash based structural data storage, inquiry and migration method | |
| CN106570113B (en) | Mass vector slice data cloud storage method and system | |
| CN103353901B (en) | The orderly management method of table data based on Hadoop distributed file system and system | |
| CN105843933B (en) | The index establishing method of distributed memory columnar database | |
| CN112015820A (en) | Method, system, electronic device and storage medium for implementing distributed graph database | |
| CN103617276A (en) | Method for storing distributed hierarchical RDF data | |
| CN103823846A (en) | Method for storing and querying big data on basis of graph theories | |
| US20190278757A1 (en) | Distributed Database Management System with Dynamically Split B-Tree Indexes | |
| Liang et al. | Mid-model design used in model transition and data migration between relational databases and nosql databases | |
| Gao et al. | An efficient ring-based metadata management policy for large-scale distributed file systems | |
| CN105357247A (en) | Multi-dimensional cloud resource interval finding method based on hierarchical cloud peer-to-peer network | |
| CN104391931A (en) | Method for efficiently indexing mass data in cloud computing | |
| US9239852B1 (en) | Item collections | |
| CN107493309B (en) | File writing method and device in distributed system | |
| CN112685417B (en) | Database operation method, system, device, server and storage medium | |
| CN105205158A (en) | Big data retrieval method based on cloud computing | |
| Batko et al. | Scalable Similarity Search in Metric Spaces. | |
| Chihoub et al. | A scalability comparison study of data management approaches for smart metering systems | |
| CN103220355B (en) | Multi-user configuration method in content distributing network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20150304 |
|
| RJ01 | Rejection of invention patent application after publication |