CN106850856A - A kind of distributed memory system and its synchronization caching method - Google Patents
A kind of distributed memory system and its synchronization caching method Download PDFInfo
- Publication number
- CN106850856A CN106850856A CN201710190374.8A CN201710190374A CN106850856A CN 106850856 A CN106850856 A CN 106850856A CN 201710190374 A CN201710190374 A CN 201710190374A CN 106850856 A CN106850856 A CN 106850856A
- Authority
- CN
- China
- Prior art keywords
- client
- data
- caching
- cache
- buffer area
- 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
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1097—Protocols in which an application is distributed across nodes in the network for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/56—Provisioning of proxy services
- H04L67/568—Storing data temporarily at an intermediate stage, e.g. caching
Abstract
The invention discloses a kind of distributed memory system, including several clients, metadata node, monitor node and several memory nodes for being connected with the network switch, cache synchronization module is provided with described each client, caching record module is provided with described each memory node;A kind of synchronization caching method of distributed memory system, comprises the following steps that:Cache synchronization module is set in each client, on several memory nodes of distributed storage service end, each memory node sets cache synchronization module, is intercepted when data are written and read operation, and in centre insertion caching process, data are read out in client finally.The present invention can improve the hit rate of local client caching, and ensure the uniformity of data, finally improve the reading performance in distributed memory system.
Description
Technical field
The invention belongs to technical field of data storage, more particularly to a kind of distributed memory system and its synchronization caching side
Method.
Background technology
As shown in Figure 1 in existing distributed storage cluster, service end includes:Metadata node, monitor node, storage
Node.But largest is memory node, the data of user are preserved, and directly provide clients with the branch of read-write requests
Hold.Monitor node (some systems are also management node) is mainly the state for recording all service ends and client, this node
Can be communicated with other nodes within often several seconds excessively, determine whether each service end and client are normal.
In order to improve the readwrite performance of system, client and server memory node may all provide data buffer storage, to add
The speed of fast read-write.But client realizes that read-write cache will bring more complicated problem in a distributed system.Such as visitor
Family end is when data are write, if be cached to locally, now the machine of delaying or service error occurs in client, it will produce data to lose
Lose.Client is when data are read, if directly reading data from local cache, but possible other clients have updated phase before
Data are closed, and local read buffer can not be obtained, and be at this moment legacy data by what is obtained.
So in a distributed system, client-side program does not provide data buffer storage typically, using synchronous processing method, and
The scheme of data buffer storage is only provided in service end memory node.But if a reliable data can be provided in client
Caching, can significantly improve storage performance.There is the possibility of loss of data due to the write buffer of client, the present invention is
For accelerate client reading performance to design.
The content of the invention
Goal of the invention:For problems of the prior art, the present invention provides one kind by service end memory node
On record the status information of each client-cache, when service end carries out data and updates, notify that relative clients end is discarded local
Cache information, the final reading performance for realizing accelerating client, and the Multipoint synchronous mode by caching, it is to avoid client reads data
When obtain old data cached, finally realize a kind of reliable distributed memory system and its synchronization caching method.
Technical scheme:In order to solve the above technical problems, the present invention provide a kind of distributed memory system, including with
Several clients, metadata node, monitor node and several memory nodes that the network switch is connected, described each client
Cache synchronization module is provided with end, caching record module is provided with described each memory node;
Memory node:Data for preserving user, and directly provide clients with the support of read-write requests;
Monitor node:State for recording all service ends and client.
A kind of synchronization caching method of distributed memory system as described above, comprises the following steps that:In each client
Upper setting cache synchronization module, on several memory nodes of distributed storage service end, each memory node sets caching
Synchronization module, is intercepted when data are written and read operation, and in centre insertion caching process, finally in client to data
It is read out.
Further, the specific method of caching process when carrying out write operation to data is:
Step one:File writing module receives write request first, is then smaller object file division, and to object
Carry out exclusive number:ID;
Step 2:Cache synchronization module starts and sets up a certain size buffer area;Then the right of each write request is preserved
The numbering of image data content and the object;Whether capacity has completely judged finally is accounted for buffer area, when appearance shared by buffer area
Amount removes oldest target cache when being judged as having expired;When capacity shared by buffer area be judged as less than when carry out step 3 operation protect
Deposit object includes that numbering and content juxtaposition object are " transitory state " to local cache
Step 3:After client local cache preserves data, to server end sending object content, object ID and client
The ID of itself, and this data cached is designated " transitory state ";
Step 4:Service end receive data after, in conservation object content to permanent storage media, then upgating object ID and
During client id is recorded to services cache;
Step 5:Judge for a plurality of record in caching record module with the presence or absence of the object ID, if do not deposited
Then notifying client write operation to complete and remove object " transitory state " and final write operation terminates, if there is then being walked
Rapid six;
Step 6:The object ID and client id are traveled through, the cache invalidation of other clients object ID is then notified, and
Asynchronous wait returning result;If it is asynchronous have result if return to asynchronous result and terminate, if asynchronous coming to nothing enters step
Seven;
Step 7:Monitor node records the client entirety cache invalidation without asynchronous returning result first, then client
End timed communication task triggering cache synchronization module removes all cachings of the client.
Further, the specific method of caching process when carrying out read operation to data is:
Step 1:File read through model receives file read request, and then whether local cache is sentenced including object data
It is disconnected, back read data is returned if comprising object data to requestor, read operation process terminates;If object data is not included
Into step 2:
Step 2:Server receives read request and object data is obtained from service end, then upgating object ID and client
During ID is recorded to services cache, and the data object result for reading is returned to client.Whether client judges local cache
It has been expired that, remove oldest target cache and enter step 3 if caching expire, if it is determined that caching less than being then directly entered
Step 3:
Step 3:Conservation object data are to local cache, including numbering and content and finally return that reading data, read operation
Journey terminates.
Further, whether buffer area accounts for capacity in the step of caching process when carrying out write operation to data two
Full criterion is to judge whether the space that buffer area is used is more than certain ratio (such as 90 percent), if greatly
The capacity that accounts for for then judging buffer area in the ratio has been expired, if be not more than the ratio if judge buffer area account for capacity less than.
Further, cached in the step of caching process when carrying out read operation to data 2 whether full judgement mark
Standard is to judge whether the space that buffer area is used is more than certain ratio (such as 90 percent), if greater than 9 percent
Ten judge caching expire, if no more than 90 percent judge cache less than.
Compared with prior art, the advantage of the invention is that:
In whole flow process, the client machine of delaying is the cache invalidation equivalent to this node to the present invention, without influence data
Uniformity.If there is machine of delaying, the situation for increasing or reducing in the memory node of server end, then be notified that all clients
All local cache datas are discarded at end, it is ensured that the uniformity of data.Therefore, for reading the application services for writing few Web etc more,
During whole operation, the burden of writing of server is increased to a certain degree, but can be substantially to accelerate read request performance and reaction
Speed.
The present invention can improve the hit rate of local client caching, and ensure the uniformity of data, finally improve point
Reading performance in cloth storage system.
Brief description of the drawings
Fig. 1 is the structural representation of distributed memory system in the prior art;
Fig. 2 is structural representation of the invention;
Fig. 3 is the overall flow figure that the present invention carries out caching process during write operation to data;
Fig. 4 is the overall flow figure that the present invention carries out caching process during read operation to data.
Specific embodiment
With reference to the accompanying drawings and detailed description, the present invention is furture elucidated.
The invention provides a kind of caching by multiple clients and the synchronous method of service end, by service end memory node
The status information of each client-cache is recorded, when service end carries out data renewal, notifies that relative clients end is discarded local slow
Information is deposited, the final reading performance for realizing accelerating client.The granularity cached in the present invention is the object in units of an object
Can be a complete file, or a part for big file.Visitor finally avoid by the method for synchronization of this caching
Read the unreliable problem of legacy data cached during data in family end.
System architecture of the invention such as accompanying drawing 2.Service end has multiple memory nodes in this distributed storage, and each is deposited
Storage node increases by one " the caching record module " of present invention design;Client also has multiple, and the present invention is increased in client
" the cache synchronization module " of design.The position that service end preserves data is determined by distributed memory system itself, may be distributed to
On any memory node.Flow of the invention is the read-write flow of data intercept, in the treatment of centre insertion caching.For
The method for preserving each target cache data is:
1. when " cache synchronization module " starts, a certain size buffer area is set up.
2. " cache synchronization module " preserves the object data content of each write request and the numbering of the object.When buffer area is fast
When full (such as about 90% uses), then it is dirty data to use conventional algorithm to put oldest content.
3. after client local cache preserves data, to server end sending object content, object ID and client itself
ID, but temporarily put this it is data cached be " transitory state " mark.
4. after service end receives data, in conservation object content to permanent storage media, and object ID and client certainly
The ID of body is preserved as a record of " caching record module ".Here same object ID, can have a plurality of client id to remember
Record, because multiple client may all cache the object.
5. for be not present in record this client initiate other clients, server to these clients send out
The notice for sending object ID caching record to fail, but asynchronous wait returning result.
6. after all cache invalidations notify to distribute, result is not to wait for, directly notifies that initiation write request client is removed this and delayed
" transitory state " mark deposited, terminates this write request.
7. for the module of the asynchronous returning result of wait, the asynchronous end if returning result is found within a certain period of time.
Still come to nothing after otherwise repeatedly retrying, then notify monitor node (or management node) the client global failure.
8. for receiving the client of object ID cache invalidation notice, then this data cached is set to dirty data.
9. client timed task constantly communicates with monitor node (or management node), it is found that client global failure leads to
After knowing, local client can put all local caches for dirty data.
For the method for client reading object data is:
1. client searches local cache with the presence or absence of the object data and object number, and is not dirty data.
2. if it does, and do not have " transitory state " indicate, illustrate data effectively, then be directly returned to requestor, reading
Object Process terminates.
3. if it does not, rear end server initiates read request.
4. server end obtains object data from itself storage, and addition or renewal should in " caching record module "
Object ID and client id corresponding relation are recorded.
5. then data are returned to client by server end.
6. client can finally be provided to requestor equally current data buffer storage to local, reads Object Process and terminates.When
When buffer area is expired soon (such as about 90% uses), then it is dirty data to use conventional algorithm to put oldest content.
Wherein, dirty data is set in top-operation, shows that the segment data is invalid, can covered.It is actual quite with remove number
According to, but it is very quick.In whole flow process, the client machine of delaying is the cache invalidation equivalent to this node, without influence number
According to uniformity.If there is machine of delaying, the situation for increasing or reducing in the memory node of server end, then be notified that all visitors
All local cache datas are discarded at family end, it is ensured that the uniformity of data.As for service end memory node delay machine or data are lost
Mistake is ensured by the method for many copies of service end, is not belonging to scope.
From the foregoing, for reading the application services for writing few Web etc, during whole operation, increasing to a certain degree more
The burden of writing of server is added, but can be substantially to accelerate read request performance and reaction speed.
In the present invention in distributed memory system, client deployment cache synchronization module, in the memory node of service end
Deployment caching record module, the flow of whole read-write is shown in accompanying drawing 3 and 4, and the method write request flow includes:
1. the memory node of service end stores user data in units of object, and an object is less than 4M (configurable) greatly
Small user data, can be a complete file, or a part for big file, or one the one of long data block
Part.That is for the file less than 4M, complete file is an object;For the file more than 4M, in units of 4M
Cutting;For not being the data block in units of file, such as one simulation block device, is also to be cut in units of 4M.
2. when client " cache synchronization module " starts, a certain size buffer area is set up, buffer area can be stored in interior
Deposit or the high-speed Medium such as SSD in, without influenceing its function.
3. " cache synchronization module " first determines whether whether buffer area is full (such as about 90% uses) soon.If it is adopt
It is dirty data to put oldest cache object with conventional algorithm.Afterwards, this conservation object content and correlation number are saved in this
Ground buffer area, this is data cached for " transitory state " indicates for juxtaposition, represents that data not can use finally also.
4. client sending object content, object ID and the client ID of itself.
5. after service end receives data, in conservation object content to permanent storage media, and in " caching record module "
Upgating object ID and client id application relation record.This record is stored in internal memory, it may not be necessary to permanently stored.
6. " caching record module " searches all records with object ID as key assignments, may have a plurality of here, such as:
Object ID, the ID1 of client 1
Object ID, the ID2 of client 2
This shows that this object ID has caching in multiple client.The content of these numberings can be set with various ways
Put, as long as ensureing uniqueness.
7. " caching record module " traverse object ID records, for other clients for not being the initiation of this client, take
Business device to these clients initiate object ID cache invalidation notice, but asynchronous wait returning result.
8. after all cache invalidations notify to distribute, return to this write request and terminate to client.After client has notice
" transitory state " mark of the caching is removed, terminates this write request.
9., for waiting asynchronous returning result module, returning result then asynchronous end is had within a certain period of time.Else if
Repeatedly (generally 3 times) retry send object ID cache invalidation notice, still come to nothing, then notify monitor node (or pipe
Reason node) the client global failure.
10. client receives object ID cache invalidation notice, then this data cached is set to dirty data.
11. client timed tasks constantly communicate with monitor node (or management node), find client global failure
After notice, local client can put all local caches for dirty data.
Include for read request flow:
1. when client " cache synchronization module " receives data read request, first determine whether local cache with the presence or absence of the number
According to and reference numeral.
2. if it does, and there is no " transitory state " to indicate that local data of directly taking away returns to requestor, read request
Terminate.This point can greatly accelerate the process for reading data.
3. if it does not, moving towards back-end server initiates read request.
4. server end, after receiving request, obtains object data, and add in " caching record module " in itself storage
Plus or upgating object ID and client id record.
5. then data are returned to client by server end.
6. client " cache synchronization module " is general judges whether buffer area is full (such as about 90% uses) soon.If
Be then removed using conventional algorithm it is oldest data cached.Then this read request data is cached, finally the contents of object is returned
Back to requestor, read request terminates.
Therefore, the invention provides a kind of multi-client read buffer and the synchronous method of service end.The method can be improved
The hit rate of local client caching, and ensure the uniformity of data, finally improve the reading performance in distributed memory system.
Embodiments of the invention is the foregoing is only, is not intended to limit the invention.It is all in principle of the invention
Within, the equivalent made should be included within the scope of the present invention.The content category that the present invention is not elaborated
In prior art known to this professional domain technical staff.
Claims (6)
1. a kind of distributed memory system, it is characterised in that including several clients, the metadata that are connected with the network switch
Node, monitor node and several memory nodes, are provided with cache synchronization module in described each client, it is described each deposit
Caching record module is provided with storage node;
Memory node:Data for preserving user, and directly provide clients with the support of read-write requests;
Monitor node:State for recording all service ends and client.
2. a kind of synchronization caching method of distributed memory system as claimed in claim 1, it is characterised in that specific steps are such as
Under:Cache synchronization module is set in each client, and on several memory nodes of distributed storage service end, each is deposited
Storage node sets cache synchronization module, is intercepted when data are written and read operation, and inserts caching process in centre, finally
Data are written and read in client.
3. a kind of synchronization caching method of distributed memory system according to claim 2, it is characterised in that the logarithm
Specific method according to caching process when carrying out write operation is:
Step one:File writing module receives write request first, is then smaller object file division, and carried out to object
Exclusive number:ID;
Step 2:Cache synchronization module starts and sets up a certain size buffer area;Then the number of objects of each write request is preserved
According to content and the numbering of the object;Whether capacity has completely judged finally is accounted for buffer area, when capacity is sentenced shared by buffer area
Oldest target cache is removed when breaking to have expired;When capacity shared by buffer area be judged as less than when carry out step 3 operation conservation object
Include that numbering and content juxtaposition object are " transitory state " to local cache
Step 3:After client local cache preserves data, to server end sending object content, object ID and client itself
ID, and this data cached is designated " transitory state ";
Step 4:After service end receives data, in conservation object content to permanent storage media, then upgating object ID and client
During end ID is recorded to services cache;
Step 5:Judge for a plurality of record in caching record module with the presence or absence of the object ID, if there is no then
Notify client write operation to complete and remove object " transitory state " and final write operation terminates, if there is then carrying out step
Six;
Step 6:The object ID and client id are traveled through, the cache invalidation of other clients object ID is then notified, and it is asynchronous
Wait returning result;If it is asynchronous have result if return to asynchronous result and terminate, if asynchronous coming to nothing enters step 7;
Step 7:Monitor node records the client entirety cache invalidation without asynchronous returning result first, and then client is determined
When communication task triggering cache synchronization module remove all cachings of the client.
4. a kind of synchronization caching method of distributed memory system according to claim 2, it is characterised in that the logarithm
Specific method according to caching process when carrying out read operation is:
Step 1:File read through model receives file read request, then whether local cache is judged including object data, such as
Fruit then returns back read data to requestor comprising object data, and read operation process terminates;Enter if object data is not included
Step 2:
Step 2:Server receives read request and object data is obtained from service end, and then upgating object ID and client id are arrived
In services cache record, and the data object result for reading is returned to client;Client judges whether local cache has expired,
Remove oldest target cache and enter step 3 if caching expire, if it is determined that caching less than being then directly entered step
Three:
Step 3:Conservation object data are to local cache, including numbering and content and finally return that reading data, read operation process knot
Beam.
5. a kind of synchronization caching method of distributed memory system according to claim 3, it is characterised in that the logarithm
According to when carrying out write operation two the step of caching process in the capacity that accounts for of buffer area whether full criterion is to judge buffer area
Whether the space for being used is more than 90 percent, judges that the capacity that accounts for of buffer area has been expired if greater than 90 percent, such as
Fruit be not more than 90 percent judge buffer area account for capacity less than.
6. a kind of synchronization caching method of distributed memory system according to claim 4, it is characterised in that the logarithm
According to when carrying out read operation 2 the step of caching process in cache whether full criterion is to judge the sky that buffer area is used
Between whether be more than 90 percent, if greater than 90 percent judge caching expired, if no more than 90 percent
Judge caching less than.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710190374.8A CN106850856A (en) | 2017-03-28 | 2017-03-28 | A kind of distributed memory system and its synchronization caching method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710190374.8A CN106850856A (en) | 2017-03-28 | 2017-03-28 | A kind of distributed memory system and its synchronization caching method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106850856A true CN106850856A (en) | 2017-06-13 |
Family
ID=59130719
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710190374.8A Pending CN106850856A (en) | 2017-03-28 | 2017-03-28 | A kind of distributed memory system and its synchronization caching method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106850856A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107608627A (en) * | 2017-08-21 | 2018-01-19 | 云宏信息科技股份有限公司 | A kind of remote data classification storage method, electronic equipment and storage medium |
CN107643988A (en) * | 2017-09-15 | 2018-01-30 | 郑州云海信息技术有限公司 | A kind of storage system and storage method with failover mechanism |
CN108762673A (en) * | 2018-05-24 | 2018-11-06 | 浪潮电子信息产业股份有限公司 | A kind of remote data access processing system |
WO2019048969A1 (en) * | 2017-09-05 | 2019-03-14 | International Business Machines Corporation | Asynchronous update of metadata tracks in response to a cache hit generated via an i/o operation over a bus interface |
CN109691065A (en) * | 2018-08-23 | 2019-04-26 | 袁振南 | Distributed memory system and its data read-write method, storage terminal and storage medium |
CN110187825A (en) * | 2018-06-26 | 2019-08-30 | 西安奥卡云数据科技有限公司 | The super more copies of fusion of one kind accelerate storage system |
CN112764690A (en) * | 2021-02-03 | 2021-05-07 | 北京同有飞骥科技股份有限公司 | Distributed storage system |
WO2021102673A1 (en) * | 2019-11-26 | 2021-06-03 | Citrix Systems, Inc. | Document storage and management |
CN114051056A (en) * | 2022-01-13 | 2022-02-15 | 阿里云计算有限公司 | Data caching and reading method and data access system |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102136003A (en) * | 2011-03-25 | 2011-07-27 | 上海交通大学 | Large-scale distributed storage system |
CN102541983A (en) * | 2011-10-25 | 2012-07-04 | 无锡城市云计算中心有限公司 | Method for synchronously caching by multiple clients in distributed file system |
CN104156327A (en) * | 2014-08-25 | 2014-11-19 | 曙光信息产业股份有限公司 | Method for recognizing object power failure in write back mode in distributed file system |
US20150026417A1 (en) * | 2012-12-28 | 2015-01-22 | Huawei Technologies Co., Ltd. | Caching Method for Distributed Storage System, a Lock Server Node, and a Lock Client Node |
CN104580437A (en) * | 2014-12-30 | 2015-04-29 | 创新科存储技术(深圳)有限公司 | Cloud storage client and high-efficiency data access method thereof |
CN105549905A (en) * | 2015-12-09 | 2016-05-04 | 上海理工大学 | Method for multiple virtual machines to access distributed object storage system |
-
2017
- 2017-03-28 CN CN201710190374.8A patent/CN106850856A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102136003A (en) * | 2011-03-25 | 2011-07-27 | 上海交通大学 | Large-scale distributed storage system |
CN102541983A (en) * | 2011-10-25 | 2012-07-04 | 无锡城市云计算中心有限公司 | Method for synchronously caching by multiple clients in distributed file system |
US20150026417A1 (en) * | 2012-12-28 | 2015-01-22 | Huawei Technologies Co., Ltd. | Caching Method for Distributed Storage System, a Lock Server Node, and a Lock Client Node |
CN104156327A (en) * | 2014-08-25 | 2014-11-19 | 曙光信息产业股份有限公司 | Method for recognizing object power failure in write back mode in distributed file system |
CN104580437A (en) * | 2014-12-30 | 2015-04-29 | 创新科存储技术(深圳)有限公司 | Cloud storage client and high-efficiency data access method thereof |
CN105549905A (en) * | 2015-12-09 | 2016-05-04 | 上海理工大学 | Method for multiple virtual machines to access distributed object storage system |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107608627A (en) * | 2017-08-21 | 2018-01-19 | 云宏信息科技股份有限公司 | A kind of remote data classification storage method, electronic equipment and storage medium |
CN107608627B (en) * | 2017-08-21 | 2020-10-02 | 云宏信息科技股份有限公司 | Remote data hierarchical storage method, electronic equipment and storage medium |
US11010295B2 (en) | 2017-09-05 | 2021-05-18 | International Business Machines Corporation | Asynchronous update of metadata tracks in response to a cache hit generated via an i/o operation over a bus interface |
WO2019048969A1 (en) * | 2017-09-05 | 2019-03-14 | International Business Machines Corporation | Asynchronous update of metadata tracks in response to a cache hit generated via an i/o operation over a bus interface |
US10565109B2 (en) | 2017-09-05 | 2020-02-18 | International Business Machines Corporation | Asynchronous update of metadata tracks in response to a cache hit generated via an I/O operation over a bus interface |
GB2579754A (en) * | 2017-09-05 | 2020-07-01 | Ibm | Asynchronous update of metadata tracks in response to a cache hit generated via an I/O operation over a bus interface |
GB2579754B (en) * | 2017-09-05 | 2020-12-02 | Ibm | Asynchronous update of metadata tracks in response to a cache hit generated via an I/O operation over a bus interface |
CN107643988A (en) * | 2017-09-15 | 2018-01-30 | 郑州云海信息技术有限公司 | A kind of storage system and storage method with failover mechanism |
CN108762673A (en) * | 2018-05-24 | 2018-11-06 | 浪潮电子信息产业股份有限公司 | A kind of remote data access processing system |
CN110187825A (en) * | 2018-06-26 | 2019-08-30 | 西安奥卡云数据科技有限公司 | The super more copies of fusion of one kind accelerate storage system |
CN109691065A (en) * | 2018-08-23 | 2019-04-26 | 袁振南 | Distributed memory system and its data read-write method, storage terminal and storage medium |
CN109691065B (en) * | 2018-08-23 | 2021-11-09 | 袁振南 | Distributed storage system and data read-write method thereof, storage terminal and storage medium |
WO2021102673A1 (en) * | 2019-11-26 | 2021-06-03 | Citrix Systems, Inc. | Document storage and management |
US11580148B2 (en) | 2019-11-26 | 2023-02-14 | Citrix Systems, Inc. | Document storage and management |
CN112764690A (en) * | 2021-02-03 | 2021-05-07 | 北京同有飞骥科技股份有限公司 | Distributed storage system |
CN114051056A (en) * | 2022-01-13 | 2022-02-15 | 阿里云计算有限公司 | Data caching and reading method and data access system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106850856A (en) | A kind of distributed memory system and its synchronization caching method | |
US10469577B2 (en) | Caching method and system based on cache cluster | |
CN103885895B (en) | Write performance in fault-tolerant cluster storage system | |
US10268719B2 (en) | Granular buffering of metadata changes for journaling file systems | |
AU2002335503B2 (en) | Disk writes in a distributed shared disk system | |
CN103942252B (en) | A kind of method and system for recovering data | |
US20060136472A1 (en) | Achieving cache consistency while allowing concurrent changes to metadata | |
CN103186554B (en) | Distributed data mirror method and storage back end | |
CN105549905A (en) | Method for multiple virtual machines to access distributed object storage system | |
US20020032671A1 (en) | File system and file caching method in the same | |
CN105635196B (en) | A kind of method, system and application server obtaining file data | |
WO2018137327A1 (en) | Data transmission method for host and standby devices, control node, and database system | |
CN105224255B (en) | A kind of storage file management method and device | |
CN106527974B (en) | A kind of method that writing data, equipment and system | |
CN105701219B (en) | A kind of implementation method of distributed caching | |
JPH0827755B2 (en) | How to access data units at high speed | |
CN103530388A (en) | Performance improving data processing method in cloud storage system | |
CN101286127A (en) | Multi-fork diary memory continuous data protecting and restoration method | |
CN110399348A (en) | File deletes method, apparatus, system and computer readable storage medium again | |
CN111782612A (en) | File data edge caching method in cross-domain virtual data space | |
US10387308B2 (en) | Method and apparatus for online reducing caching devices | |
CN107046575A (en) | A kind of cloud storage system and its high density storage method | |
CN107329859A (en) | A kind of data guard method and storage device | |
CN100394404C (en) | System and method for management of metadata | |
AU2002248570B2 (en) | Managing checkpoint queues in a multiple node system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170613 |