CN103428288A - Method for synchronizing copies on basis of partition state tables and coordinator nodes - Google Patents
Method for synchronizing copies on basis of partition state tables and coordinator nodes Download PDFInfo
- Publication number
- CN103428288A CN103428288A CN2013103501587A CN201310350158A CN103428288A CN 103428288 A CN103428288 A CN 103428288A CN 2013103501587 A CN2013103501587 A CN 2013103501587A CN 201310350158 A CN201310350158 A CN 201310350158A CN 103428288 A CN103428288 A CN 103428288A
- Authority
- CN
- China
- Prior art keywords
- node
- copies
- subregion
- partition holding
- state
- 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
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a method for synchronizing copies on the basis of partition state tables and coordinator nodes. The method is characterized in that the consistency of the copies is maintained by the aid of an authorized copy synchronization mechanism, storage nodes with copies required to be updated transmit synchronization requests to the coordinator nodes, the coordinator nodes judge whether the storage nodes meet copy synchronization operation executing conditions or not according to general partition state tables, and the storage nodes which meet the conditions are authorized to execute copy synchronization operation. The method for synchronizing the copies on the basis of the partition state tables and the coordinator nodes has the advantages that a duplicate detection rate for the consistency of the copies, a synchronization bandwidth occupancy rate and the sizes of inconsistent windows are greatly reduced, the problem of version conflict of multiple copies is solved, and the copy synchronization timeliness is improved; a distributed file storage system can quickly respond to external requests and can quickly detect the consistency of the copies, and accordingly the service capacity of the system and the final consistency efficiency are improved.
Description
Technical field
The present invention relates to technical field of memory, be specifically related to a kind of copies synchronized method based on subregion state table and coordinator node.
Background technology
Traditional data storage technology is stored in a piece of data on a node, needs the user of these data all must arrive on same node and read it.In recent years, along with the user in the Internet is increasing to the demand of resource, network easily occurs congested, and the limited node of disposal ability also can be because access number be too large the machine of delaying.And traditional data storage technology is in extensibility, and the aspects such as high availability can't meet current data storage requirement far away, the eighties in 20th century, the concept of distributed document storage is suggested.
The distributed document storage refers to and utilizes distributed computing technology, by computer network, with node, be connected, memory device on different nodes in network is got up by the Distributed Application software assembly, by the storage system unified management, collaborative work, data storage and business externally are provided jointly, to the user, provide access interface unified, objectification, direct control and the resource management of shielding to physical equipment.Distributed file storage system is by distributed framework, create multiple copies, and by their reasonable layouts on a plurality of server nodes, for user's access, utilize many storage servers share the storage load and process access request, can effectively reduce the node failure rate, reduce subscriber response time.
Store by distributed document, create multiple copies, utilize location server location storage information, not only improved reliability, availability and the access efficiency of system, all kinds of memory devices can be combined with, good extendible stores service is provided.Compare traditional memory device, it is in autgmentability, fail safe, and the aspects such as reliability, all be greatly improved.
Yet, owing to there being a plurality of copies, there will be the nonsynchronous problem of copy, a plurality of copy inconsistencies, cause the result of different user access inconsistent, causes distributed file storage system availability and reliability to reduce.In order to keep the consistency of copy in distributed file storage system, the duplicate of the document strategy has been proposed.Be the consistency of safeguarding copy on the one hand, improve availability and the reliability of distributed file storage system.The overall performance of elevator system on the other hand.For example, by increase copy in storage server, allow more storage server that service is provided, strengthen load balancing and the extended capability of system.
File stored copies Managed Solution under a lot of distributed structure/architecture has been proposed in recent years.According to the framework difference of distributed file system, mainly can be divided into the replication strategy of centralization framework storage system and the replication strategy of symmetrical architecture storage system.
(1) replication strategy of centralization framework storage system
The main consistency detection that adopts analog storage state method, major-minor this law to realize copy in the storage system of centralization framework.Analog storage state method is the list of maintenance documentation piece in each memory node, by heartbeat and namenode, carries out the block message exchange, by namenode, is controlled the synchronous of copy and is copied.Major-minor this law is picked out a reference copies in a plurality of copies of same file, and all write operations all carry out primary copy, and is controlled the synchronous of other copies and copied by primary copy.
(2) replication strategy of symmetrical architecture storage system
Because there is obvious namenode bottleneck problem in the centralization framework, so the better symmetrical architecture of autgmentability is suggested and is tending towards main flow.The main duplicate detection mechanism that adopts " point-to-point " between memory node in the symmetrical architecture storage system.As its name suggests, in the duplicate detection mechanism of " point-to-point ", memory node is by copy check code and version number on its reciprocity distant-end node of request, with local replica, compare, if local replica and far-end copy are inconsistent and the far-end copy version is older, local replica can, by the local replica data-pushing to peer node, be realized the copies synchronized of " point-to-point ".
The replication strategy of traditional symmetrical storage framework has the following disadvantages:
1) the duplicate detection amount is large.The symmetrical architecture storage system that has 3 copies for each subregion, guarantee the consistency of 3 copies, needs between 3 memory nodes to send 6 message and detected.Although this mode is easy to realize, takies massive band width, has reduced the overall performance of system.
2) inconsistent window amplifies.For single memory node, each takes turns the conforming detection of copy all can consume the regular hour, can increase this interval time along with the increasing of number of partitions on memory node, thereby causing a large amount of inconsistent copies to be delayed for a long time could detect, increased the inconsistent window of system, make the efficiency of final consistency lower, the user very easily has access to inconsistent data.
Summary of the invention
For the deficiencies in the prior art, the invention provides the copies synchronized method based on subregion state table and coordinator node, avoided duplicate detection, and inconsistent window is little.
A kind of copies synchronized method based on subregion state table and coordinator node, described copies synchronized method is applied to the distributed file storage system of symmetrical architecture, described distributed file storage system comprises a plurality of memory nodes and a coordinator node, and described copies synchronized method comprises:
(1) initialization
1-1) on each memory node, set up the subregion state table, partition number, subregion state and the version number of the partition holding of described subregion state table real time record respective stored node, after initialization, subregion state and version number are empty;
Described subregion state comprises: " Complete Synchronization ", " up-to-date ", " being synchronized to far-end " and " being updated to this locality " four kinds of states;
1-2) described coordinator node gathers all subregion state tables and sets up the subregion state summary table;
(2) send synchronization request
If in certain memory node, the data of a certain partition holding are modified, and this partition holding stateless conflict, the subregion state that upgrades this partition holding is " up-to-date ", take this memory node simultaneously to send the synchronization request of current partition holding as source node to described coordinator node;
(3) request return instruction
Coordinator node receives described synchronization request, and, according to the subregion state summary table, judges whether current partition holding meets the executive condition of copies synchronized operation, and send corresponding return instruction, is specially:
If 3-1) do not meet described executive condition, to maintain the subregion state of current partition holding constant for coordinator node, and return to " applying for unsuccessfully " to source node;
If 3-2) meet described executive condition, before the coordinator node renewal, the subregion state of partition holding is " being synchronized to far-end ", and returns to " applying for successfully " to source node;
(4) carry out simultaneous operation
Source node receives described return instruction, and makes following operation:
If 4-1) return instruction is " applying for unsuccessfully ", source node keeps the subregion state of current partition holding constant, and waits for copies synchronized;
If 4-2) return instruction is " applying for successfully ", source node is carried out the copies synchronized operation to the memory node at all copies place of current partition holding, makes all copies of current partition holding consistent.
In copies synchronized method of the present invention, each memory node is safeguarded respectively a subregion state summary table, has recorded partition number, subregion state and the version number of all partition holdings in the respective stored node; Coordinator node is safeguarded the subregion state summary table, has recorded all partition informations of all nodes in this system, comprises partition number, subregion state and version number.Coordinator node receives the synchronization request of memory node, according to the subregion state summary table, judges whether this memory node meets the condition of carrying out the copies synchronized operation, licenses to qualified memory node, allows this node carry out the copies synchronized operation; Ineligible memory node, show to have other replica node to send the copies synchronized application prior to source node, and the source node hold mode is constant, and wait for copies synchronized.The copies synchronized mechanism of this mandate can effectively reduce the useless generation copied, reduce synchronization bandwidth and take, can well manage copy version, avoid the version conflict problem of a plurality of copies simultaneously, and increase the ageing of copies synchronized, reduce inconsistent window.And the copies synchronized operation is carried out between memory node, and without coordinator node, coordinator node only is responsible for authorizing according to the simultaneous operation of request authoritative copy, can effectively alleviate the pressure of coordinator node.
Four kinds of subregion states in described step (1) are changed according to the finite state machine principle.
Realize four kinds of subregion state conversions by the finite state machine principle, can limit the order that four kinds of subregion states are changed mutually.
Synchronization request in described step (2) carries partition number, subregion state and the version number of current partition holding.
The information of carrying in synchronization request is convenient to coordinator node and is processed.
Executive condition in described step (3) is the conflict of current partition holding stateless and version conflict.
The transfer principle of described state conflict by state determine, if current partition holding does not meet the switch condition of finite state machine, thinks current partition holding state conflict.
Described version conflict is judged according to following methods: if up-to-date in the not all copy of the copy version of current partition holding, think current partition holding version conflict.
The judgement of state conflict is prior to version conflict, when existence conflict not just can further judge whether to exist version conflict.State conflict both can judge according to the Condition Areas table on memory node also can be according to the total list deciding of subregion state on coordinator node.Version conflict judged on coordinator node, and by the version number of the copy of depositing on more all memory nodes, whether the version that upgrades is attempted in judgement is up-to-date in all copy version, if not up-to-date, is version conflict.State conflict generally due to other memory nodes prior to current memory node application simultaneous operation and this simultaneous operation do not carry out current memory node and cause.The time difference of these two requests is inconsistent window.By the Synchronization Design framework of coordinator node, and the state of state summary table and Version Control, inconsistent window can effectively be reduced.
As preferably, described copies synchronized operating basis NWR strategy carries out.
The consistency of distributed system (Consistency), availability (Availability), subregion tolerable (Tolerance of network Partition) can't meet simultaneously, adopt classical NWR strategy, in above three kinds of performances, make balance.N means to need the backup of N data, and W means that data write N node and just are counted as merit, and it is up-to-date that the data guarantee that R means to read R node wherein has portion.For example needing to back up number is 3 parts, writes 2 parts and means that data successfully write, and must reading 2 piece of data guarantees, portion is wherein arranged is up-to-date.
Described step 3-2) in:
Before the copies synchronized operation, source node is updated to " being synchronized to far-end " by the state of current partition holding in the subregion state table;
In the copies synchronized operating process, the memory node at each copy place is updated to " being updated to this locality " by the subregion state of respective stored subregion;
After copies synchronized has operated, source memory node and copy place memory node and coordinator node all are updated to " Complete Synchronization " by the subregion state of corresponding partition holding.
In the execution copies synchronized operates forward and backward and simultaneous operation process, the associated storage node is according to the subregion state of the corresponding partition holding of operation real-time update.Because the realization of data read-write operation and copies synchronized is quite independent, employing is carried out the data manipulation strategy based on subregion state, set different read-write strategies for setting different subregion states, efficiently solve the subregion read/write conflict problem occurred in state conversion process, prevent the problem of many versions confusion, concrete grammar is as follows:
1) subregion state of copy place partition holding is " being updated to this locality ":
If read operation, read the data of respective partition the distant-end node of corresponding copies synchronized operational correspondence (memory node at each copy place) from this memory node,
If retouching operation, this memory node no longer accept to upload, the retouching operation such as deletion, and described retouching operation is transferred to the distant-end node of the copies synchronized operational correspondence that this memory node is corresponding, to the operation of modifying of respective stored subregion in distant-end node;
2) subregion state of copy place partition holding is " being synchronized to far-end ":
If read operation, directly from current memory node reading out data,
If retouching operation, wait for after copies synchronized has operated, then will carry out this retouching operation;
3) subregion state of copy place partition holding is " up-to-date " or " Complete Synchronization " state:
If read-write operation, directly from current memory node reading out data,
If retouching operation, carry out this read-write operation immediately, and, after this read-write operation completes, corresponding partition holding state is updated to " up-to-date ", and sends the synchronization request of this partition holding to coordinator node.
Described distributed file storage system also comprises a hot standby node, the synchronization request continuous several times that current memory node sends is overtime, think the coordinator node machine of delaying, hot standby node is automatically taken over coordinator node and is received synchronization request and send return instruction, after coordinator node recovers, then synchronization request is transferred to coordinator node and process.
Hot standby node, as the secondary node of coordinator node, when coordinator node is delayed machine, is enabled automatically, has improved the reliability of system.
Copies synchronized method based on subregion state table and coordinator node provided by the invention, safeguard the copy consistency in distributed file storage system by the copies synchronized operation mechanism of authorizing, the duplicate detection rate, the synchronization bandwidth that greatly reduce copy consistency take and inconsistent window, thereby reduce the pressure of copies synchronized operation to distributed file storage system, make distributed file storage system rapider to the detection of the response of external request, copy consistency, the service ability of elevator system and the efficiency of final consistency.
The accompanying drawing explanation
The Organization Chart of the distributed file storage system of the symmetrical architecture with coordinator node that Fig. 1 is the present embodiment;
Fig. 2 is the flow chart that the present embodiment is based on the copies synchronized method of subregion state table and coordinator node;
The finite state machine schematic diagram that Fig. 3 is the subregion state conversion.
Embodiment
Below in conjunction with the drawings and specific embodiments, the present invention will be further described.
The distributed file storage system that the applied environment of the copies synchronized method based on subregion state table and coordinator node in the present embodiment is symmetrical architecture.As shown in Figure 1, the distributed file storage system of this symmetrical architecture comprises the memory node of a plurality of equities, a coordinator node and a hot standby node.Originally execute communicating by letter of memory node and coordinator node in example, and the communication between memory node all adopts the synchronous communications protocol based on http protocol.
The copies synchronized method based on subregion state table and coordinator node of the present embodiment, the specific works flow process as shown in Figure 2, comprising:
(1) initialization
1-1) set up subregion state table (being Hash table in the present embodiment) on each memory node, partition number, subregion state and the version number of the partition holding of real time record respective stored node, after initialization, subregion state and version number are empty;
1-2) described coordinator node gathers all subregion state tables and sets up the subregion state summary table;
Subregion state has following four kinds:
Complete Synchronization: show that all copies keep the state of Complete Synchronization;
Up-to-date: as to show that copy is up-to-date, and with other copies, do not produce and synchronize;
Be synchronized to far-end: show up-to-date copy, and this copy is being synchronized to the far-end copy;
Be updated to this locality: show old copy, and from far-end, receive up-to-date copy.
Four kinds of subregion states are changed according to the finite state machine principle, and specifically transfer process is as shown in Figure 3:
(a) subregion state of current partition holding is " up-to-date ":
If this partition holding is modified, it is still " up-to-date " state;
If this subregion sends synchronization request to coordinator node, and the return information of receiving is transformed to " being synchronized to far-end " state during for " applying for successfully ";
Copies synchronized operation source node sent if receive, be transformed to " being updated to this locality " state;
(b) subregion state of current partition holding is " being synchronized to far-end ":
The copies synchronized operation is transformed to " Complete Synchronization " state after carrying out;
(c) when copy in " being updated to this locality " state:
The copies synchronized operation is finished, and is transformed to " Complete Synchronization " state;
(d) subregion state of current partition holding is " Complete Synchronization ":
If the data of current partition holding are modified, be transformed to " up-to-date " state,
If current subregion receives the copies synchronized operation that source node is sent, be transformed to " being updated to this locality ".
(2) send synchronization request
If in certain memory node, the data of a certain partition holding are modified, and this partition holding stateless conflict, what this partition holding can meet the state machine principle is converted to " up-to-date " from current subregion state, the subregion state that upgrades the respective partition in the subregion state table of this node is " up-to-date ", take this memory node simultaneously to send the synchronization request of current subregion as source node to coordinator node; This synchronization request carries partition number, subregion state and the version number of current partition holding.
(3) request is returned
Coordinator node receives synchronization request, and according to the subregion state summary table, and whether what judge current partition holding meets the copy executive condition, to source node, sends return instruction:
3-1) first judged whether state conflict, if there is state conflict to source node, to return to " applying for unsuccessfully ";
If 3-2) stateless conflict, further determined whether version conflict, if without version conflict, to source node, return to " applying for successfully ".
Obtain the subregion state of current partition holding from the subregion state table of memory node or the subregion state summary table coordinator node, if the subregion state of current partition holding can't be converted to " being synchronized to far-end " from current state according to the finite state machine principle, think current partition holding state conflict;
Obtain the version number of all copies of current partition holding in subregion state summary table from coordinator node, if in the not all copy of the copy version of current partition holding " up-to-date ", think current partition holding version conflict;
(4) carry out simultaneous operation
Source node receives described return instruction, and makes following operation:
If 4-1) return instruction is " applying for unsuccessfully ", source node keeps the subregion state of current partition holding constant, and waits for copies synchronized;
If 4-2) return instruction is for to apply for successfully, source node is updated to " being synchronized to far-end " by the state of current partition holding in the subregion state table; And carry out the copies synchronized operation according to the NWR strategy to the memory node at all copies place of current partition holding, make all copies of current partition holding consistent; The memory node at each copy place changes to " being updated to this locality " by the subregion state of respective stored subregion after receiving the copies synchronized operation of source node transmission; After copies synchronized has operated, source memory node and copy place memory node and coordinator node are updated to " Complete Synchronization " by the subregion state of respective partition.
For preventing the chaotic problem of the version caused because of the user writable operation, the user carries out data manipulation according to the subregion state of accessed copy place partition holding:
1) subregion state of copy place partition holding is " being updated to this locality ":
If read operation, from this memory node, the far-end of corresponding copies synchronized operational correspondence joint (sending the source node of copies synchronized operation) reads the data of respective partition putting,
If retouching operation, this node no longer accept to upload, the retouching operation such as deletion, and described retouching operation is transferred to the distant-end node of the copies synchronized operational correspondence that this node is corresponding, to the operation of modifying of respective partition in distant-end node;
2) subregion state of copy place subregion is " being synchronized to far-end ":
If read operation, directly from the present node reading out data,
If retouching operation, wait for after copies synchronized has operated, then will carry out this retouching operation;
3) subregion state of copy place subregion is " up-to-date " or " Complete Synchronization " state:
If read-write operation, directly from the present node reading out data;
If retouching operation, carry out this read-write operation immediately, and, after this read-write operation completes, corresponding subregion state is updated to " up-to-date ", and send the synchronization request of this subregion to coordinator node.
Described distributed file storage system also comprises a hot standby node, the synchronization request continuous several times that current memory node sends is overtime, it in the present embodiment, is 3 times, think the coordinator node machine of delaying, hot standby node is automatically taken over coordinator node and is received synchronization request and send return instruction, after coordinator node recovers, then synchronization request is transferred to coordinator node and process.The machine if coordinator node and hot standby node are delayed simultaneously, memory node can carry out the retry operation of 3 times, and congested for fear of request, the stand-by period of 3 retry operations increases progressively gradually, is respectively 10s, 20s, 30s.If 3 all failures of retry, memory node reactivates the copies synchronized strategy of " point-to-point ", and by final " point-to-point " copies synchronized strategy that starts whole storage cluster of the mode of Gossip message propagation.
The above; be only the specific embodiment of the present invention, but protection scope of the present invention is not limited to this, anyly is familiar with those skilled in the art in the technical scope that the present invention discloses; the variation that can expect easily or replacement, within all should being encompassed in protection scope of the present invention.
Claims (9)
1. the copies synchronized method based on subregion state table and coordinator node, it is characterized in that, described copies synchronized method is applied to the distributed file storage system of symmetrical architecture, described distributed file storage system comprises a plurality of memory nodes and a coordinator node, and described copies synchronized method comprises:
(1) initialization
1-1) on each memory node, set up the subregion state table, partition number, subregion state and the version number of the partition holding of described subregion state table real time record respective stored node, after initialization, subregion state and version number are empty;
Described subregion state comprises: " Complete Synchronization ", " up-to-date ", " being synchronized to far-end " and " being updated to this locality " four kinds of states;
1-2) described coordinator node gathers all subregion state tables and sets up the subregion state summary table;
(2) send synchronization request
If in certain memory node, the data of a certain partition holding are modified, and this partition holding stateless conflict, the subregion state that upgrades this partition holding is " up-to-date ", take this memory node simultaneously to send the synchronization request of current partition holding as source node to described coordinator node;
(3) request return instruction
Coordinator node receives described synchronization request, and, according to the subregion state summary table, judges whether current partition holding meets the executive condition of copies synchronized operation, and send corresponding return instruction, is specially:
If 3-1) do not meet described executive condition, to maintain the subregion state of current partition holding constant for coordinator node, and return to " applying for unsuccessfully " to source node;
If 3-2) meet described executive condition, before the coordinator node renewal, the subregion state of partition holding is " being synchronized to far-end ", and returns to " applying for successfully " to source node;
(4) carry out simultaneous operation
Source node receives described return instruction, and makes following operation:
If 4-1) return instruction is " applying for unsuccessfully ", source node keeps the subregion state of current partition holding constant, and waits for copies synchronized;
If 4-2) return instruction is " applying for successfully ", source node is carried out the copies synchronized operation to the memory node at all copies place of current partition holding, makes all copies of current partition holding consistent.
2. the copies synchronized method based on subregion state table and coordinator node as claimed in claim 1, is characterized in that, four kinds of subregion states in described step (1) are changed according to the finite state machine principle.
3. the copies synchronized method based on subregion state table and coordinator node as claimed in claim 2, is characterized in that, the synchronization request in described step (2) carries partition number, subregion state and the version number of current partition holding.
4. the copies synchronized method based on subregion state table and coordinator node as claimed in claim 3, is characterized in that, the executive condition in described step (3) is the conflict of current partition holding stateless and version conflict.
5. the copies synchronized method based on subregion state table and coordinator node as claimed in claim 4, it is characterized in that, the transfer principle of described state conflict by state determine, if current partition holding does not meet the switch condition of finite state machine, thinks current partition holding state conflict.
6. the copies synchronized method based on subregion state table and coordinator node as claimed in claim 5, it is characterized in that, described version conflict is judged according to following methods: if up-to-date in the not all copy of the copy version of current partition holding, think current partition holding version conflict.
7. the copies synchronized method based on subregion state table and coordinator node as claimed in claim 6, is characterized in that, described copies synchronized operating basis NWR strategy carries out.
8. the copies synchronized method based on subregion state table and coordinator node as claimed in claim 7, is characterized in that described step 3-2) in:
Before the copies synchronized operation, source node is updated to " being synchronized to far-end " by the state of current partition holding in the subregion state table;
In the copies synchronized operating process, the memory node at each copy place is updated to " being updated to this locality " by the subregion state of respective stored subregion;
After copies synchronized has operated, source memory node and copy place memory node and coordinator node all are updated to " Complete Synchronization " by the subregion state of corresponding partition holding.
9. the copies synchronized method based on subregion state table and coordinator node as claimed in claim 8, it is characterized in that, described distributed file storage system also comprises a hot standby node, the synchronization request continuous several times that current memory node sends is overtime, think the coordinator node machine of delaying, hot standby node is automatically taken over coordinator node and is received synchronization request and send return instruction, after coordinator node recovers, then synchronization request is transferred to coordinator node and process.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310350158.7A CN103428288B (en) | 2013-08-13 | 2013-08-13 | Based on the copies synchronized method of subregion state table and coordinator node |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310350158.7A CN103428288B (en) | 2013-08-13 | 2013-08-13 | Based on the copies synchronized method of subregion state table and coordinator node |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103428288A true CN103428288A (en) | 2013-12-04 |
| CN103428288B CN103428288B (en) | 2016-03-09 |
Family
ID=49652452
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310350158.7A Active CN103428288B (en) | 2013-08-13 | 2013-08-13 | Based on the copies synchronized method of subregion state table and coordinator node |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103428288B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105516230A (en) * | 2014-10-15 | 2016-04-20 | 深圳市腾讯计算机系统有限公司 | Data processing method and device |
| CN105550306A (en) * | 2015-12-14 | 2016-05-04 | 北京奇虎科技有限公司 | Multi-copy data reading/writing method and system |
| CN105610879A (en) * | 2014-10-31 | 2016-05-25 | 深圳市华为技术软件有限公司 | Data processing method and data processing device |
| CN107219997A (en) * | 2016-03-21 | 2017-09-29 | 阿里巴巴集团控股有限公司 | A kind of method and device for being used to verify data consistency |
| CN110968641A (en) * | 2018-09-30 | 2020-04-07 | 阿里巴巴集团控股有限公司 | Data writing control method and device and electronic equipment |
| CN115328880A (en) * | 2022-10-13 | 2022-11-11 | 浙江智臾科技有限公司 | Distributed file online recovery method, system, computer equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101187931A (en) * | 2007-12-12 | 2008-05-28 | 浙江大学 | Distribution type file system multi-file copy management method |
| CN101188544A (en) * | 2007-12-04 | 2008-05-28 | 浙江大学 | File transfer method for distributed file server based on buffer |
| CN102368267A (en) * | 2011-10-25 | 2012-03-07 | 曙光信息产业(北京)有限公司 | Method for keeping consistency of copies in distributed system |
| CN103139302A (en) * | 2013-02-07 | 2013-06-05 | 浙江大学 | Real-time copy scheduling method considering load balancing |
| CN103150347A (en) * | 2013-02-07 | 2013-06-12 | 浙江大学 | Dynamic replica management method based on file heat |
-
2013
- 2013-08-13 CN CN201310350158.7A patent/CN103428288B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101188544A (en) * | 2007-12-04 | 2008-05-28 | 浙江大学 | File transfer method for distributed file server based on buffer |
| CN101187931A (en) * | 2007-12-12 | 2008-05-28 | 浙江大学 | Distribution type file system multi-file copy management method |
| CN102368267A (en) * | 2011-10-25 | 2012-03-07 | 曙光信息产业(北京)有限公司 | Method for keeping consistency of copies in distributed system |
| CN103139302A (en) * | 2013-02-07 | 2013-06-05 | 浙江大学 | Real-time copy scheduling method considering load balancing |
| CN103150347A (en) * | 2013-02-07 | 2013-06-12 | 浙江大学 | Dynamic replica management method based on file heat |
Non-Patent Citations (1)
| Title |
|---|
| 王禹 等: "P2P 存储系统副本一致性维护策略", 《计算机工程》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105516230A (en) * | 2014-10-15 | 2016-04-20 | 深圳市腾讯计算机系统有限公司 | Data processing method and device |
| CN105516230B (en) * | 2014-10-15 | 2019-06-25 | 深圳市腾讯计算机系统有限公司 | A kind of data processing method and device |
| CN105610879A (en) * | 2014-10-31 | 2016-05-25 | 深圳市华为技术软件有限公司 | Data processing method and data processing device |
| CN105610879B (en) * | 2014-10-31 | 2019-05-17 | 华为技术有限公司 | Data processing method and device |
| CN105550306A (en) * | 2015-12-14 | 2016-05-04 | 北京奇虎科技有限公司 | Multi-copy data reading/writing method and system |
| CN107219997A (en) * | 2016-03-21 | 2017-09-29 | 阿里巴巴集团控股有限公司 | A kind of method and device for being used to verify data consistency |
| CN107219997B (en) * | 2016-03-21 | 2020-08-18 | 阿里巴巴集团控股有限公司 | Method and device for verifying data consistency |
| CN110968641A (en) * | 2018-09-30 | 2020-04-07 | 阿里巴巴集团控股有限公司 | Data writing control method and device and electronic equipment |
| CN110968641B (en) * | 2018-09-30 | 2023-05-02 | 阿里巴巴集团控股有限公司 | Data writing control method and device and electronic equipment |
| CN115328880A (en) * | 2022-10-13 | 2022-11-11 | 浙江智臾科技有限公司 | Distributed file online recovery method, system, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103428288B (en) | 2016-03-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9747301B2 (en) | Distributed file system using consensus nodes | |
| CN103428288B (en) | Based on the copies synchronized method of subregion state table and coordinator node | |
| US9984140B1 (en) | Lease based leader election system | |
| CN113641511B (en) | Message communication method and device | |
| CN102521038B (en) | Virtual machine migration method and device based on distributed file system | |
| US8676760B2 (en) | Maintaining data integrity in data servers across data centers | |
| US20140244578A1 (en) | Highly available main memory database system, operating method and uses thereof | |
| CN103488546A (en) | Backup and recovery method capable of supporting online concurrency of multi-level data and database | |
| CN113987064A (en) | Data processing method, system and equipment | |
| EP4213038A1 (en) | Data processing method and apparatus based on distributed storage, device, and medium | |
| CN105493474A (en) | System and method for supporting partition level journaling for synchronizing data in a distributed data grid | |
| CN105426427A (en) | MPP database cluster replica realization method based on RAID 0 storage | |
| WO2016177231A1 (en) | Dual-control-based active-backup switching method and device | |
| WO2018010501A1 (en) | Global transaction identifier (gtid) synchronization method, apparatus and system, and storage medium | |
| WO2014177085A1 (en) | Distributed multicopy data storage method and device | |
| CN105069152A (en) | Data processing method and apparatus | |
| US10097630B2 (en) | Transferring data between sites | |
| CN109639773A (en) | A kind of the distributed data cluster control system and its method of dynamic construction | |
| CN110807039A (en) | Data consistency maintenance system and method in cloud computing environment | |
| CN103365740B (en) | A kind of data cold standby method and device | |
| CN105760391A (en) | Data dynamic redistribution method and system, data node and name node | |
| CN102185717A (en) | Service processing equipment, method and system | |
| JP2012008934A (en) | Distributed file system and redundancy method in distributed file system | |
| CN102982033A (en) | Small documents storage method and system thereof | |
| WO2015196692A1 (en) | Cloud computing system and processing method and apparatus for cloud computing system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20131204 Assignee: Hundsun Technologies Inc. Assignor: Zhejiang University Contract record no.: 2018330000035 Denomination of invention: Method for synchronizing copies on basis of partition state tables and coordinator nodes Granted publication date: 20160309 License type: Common License Record date: 20180417 |
|
| EE01 | Entry into force of recordation of patent licensing contract |