CN105376269A - Virtual machine storage system, virtual machine storage system realizing method and virtual machine storage system realizing device - Google Patents

Virtual machine storage system, virtual machine storage system realizing method and virtual machine storage system realizing device Download PDF

Info

Publication number
CN105376269A
CN105376269A CN201410390308.1A CN201410390308A CN105376269A CN 105376269 A CN105376269 A CN 105376269A CN 201410390308 A CN201410390308 A CN 201410390308A CN 105376269 A CN105376269 A CN 105376269A
Authority
CN
China
Prior art keywords
data
server
virtual machine
write request
storage system
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
Application number
CN201410390308.1A
Other languages
Chinese (zh)
Other versions
CN105376269B (en
Inventor
高原
全小飞
广小明
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tianyi Cloud Technology Co Ltd
Original Assignee
China Telecom Corp Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by China Telecom Corp Ltd filed Critical China Telecom Corp Ltd
Priority to CN201410390308.1A priority Critical patent/CN105376269B/en
Publication of CN105376269A publication Critical patent/CN105376269A/en
Application granted granted Critical
Publication of CN105376269B publication Critical patent/CN105376269B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Landscapes

  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Memory System Of A Hierarchy Structure (AREA)

Abstract

The invention discloses a virtual machine storage system, a virtual machine storage system realizing method and a virtual machine storage system realizing device. The virtual machine storage system realizing method comprises the steps of receiving a writing request of a client operating system, buffering writing request data into the buffer memory of the virtual machine, and uploading the writing request data to a server according to the state of the server. According to the virtual machine storage system, the virtual machine storage system realizing method and the virtual machine storage system realizing device, after the writing request is received, through buffering the writing request data into the buffer memory of the virtual machine and uploading the writing request data to the server according to the state of the server, a congestion problem generated by writing outburst in a virtual machine storage system in cloud calculating can be settled.

Description

Virtual machine storage system and its implementation and device
Technical field
The present invention relates to field of cloud calculation, particularly relate to a kind of virtual machine storage system and its implementation and device.
Background technology
In cloud computing, when virtual machine runs, required software resource all leaves on storage server.In prior art, virtual machine storage system is by the virtual disk interface being positioned at client, monitor of virtual machine (VirtualMachineMonitor, and be positioned at the input and output (inputandoutput of server VMM), IO) service and virtual disk reflection formed, wherein, the process of VMM primary responsibility and server mutual, cache management then depends on the internal mechanism of client operating system.The feature that virtual machine stores is in client operating system lower floor, is shielded the details of client operating system and server interaction by virtual technology, and client operating system and application program thereof are changed without the need to essence just can according to the mode operation of cloud computing.
But in cloud computing environment, because virtual memory to have carried out the shielding of network and server details to upper strata, client operating system then cannot understand the situation of current network and server, therefore cannot opportunity of selecting properly uploading data.Even if client does not have process take virtual disk, server still may be in busy state.Client likely continues when server is busy to submit write request to, like this, server can be made busier on the one hand, even cause loss of data at buffer overflow, waste bandwidth resource; Client operating system can not obtain server response within the time expected on the other hand, causes relative program to run slowly or collapse.
Summary of the invention
In view of this, the technical problem that the disclosure will solve is: how to solve the congestion problems produced by writing burst in virtual machine storage system in cloud computing.
First aspect, the disclosure provides a kind of virtual machine storage system implementation method, comprising:
Receive the write request of client operating system;
By write request data buffer storage in virtual machine buffer memory;
According to the state of server by write request data upload to server.
Wherein, according to the state of server, write request data upload is comprised to server:
Judge the state of server, if the state of server is idle, then by the data upload of write request to server, if the state of server is busy, then by write request data buffer storage in virtual machine buffer memory.
Alternatively, write request data buffer storage is comprised at virtual machine buffer memory: by write request data buffer storage at nonvolatile memory.
And/or
This virtual machine storage system implementation method also comprises:
According to the type of write request data by write request data upload to server.
Further, according to the type of write request data, write request data upload is comprised to server:
If write request data are user data, then by write request data upload to server; And/or
If write request data are internal memory swap data and/or ephemeral data, then write request data are stored in nonvolatile memory;
And/or
This virtual machine storage system implementation method also comprises:
If write request data are ephemeral data, ephemeral data is cached in internal memory, after at least one ephemeral data is merged into total ephemeral data, is stored into nonvolatile memory.
Alternatively, nonvolatile memory comprises journal disk.
Alternatively, if client or server meet accident when restarting, also comprise:
Read the control section of preserving in journal disk;
Journal disk is scanned, the data segment finding dispensing section, submit to section to identify from control section;
The data segment uploaded dispensing section successively, submit to section to identify;
After uploading, normally start client operating system.
Alternatively, this virtual machine storage system realizes also comprising:
Data segment in journal disk is submitted to server;
After receiving the confirmation of receipt that server returns, the control section of change data segment is for submitting section to;
When receiving after submission that server returns confirms, the control section of change data segment is available segments;
Reclaim the data segment of available segments mark.
Alternatively, this virtual machine storage system realizes also comprising:
Receive the read request of client operating system;
In virtual machine buffer memory, data corresponding to read request are searched according to read request;
If find the data that read request is corresponding in virtual machine buffer memory, then read data, if do not find data corresponding to read request in virtual machine buffer memory, then read request is sent to server.
Second aspect, the disclosure provides a kind of virtual machine storage system implement device, comprising:
Receiver module, for receiving the write request of client operating system;
Cache module, communicates with receiver module, for by write request data buffer storage in virtual machine buffer memory;
Control module, communicates with cache module, for according to the state of server by write request data upload to server.
Wherein, control module is also for the state judging server, if the state of server is idle, then by the data upload of write request to server, if the state of server is busy, then by write request data buffer storage in virtual machine buffer memory.
Alternatively, cache module is used for write request data buffer storage at nonvolatile memory;
And/or
This device also comprises:
Control module also for according to the type of write request data by write request data upload to server.
Alternatively, control module also for:
If write request data are user data, then by write request data upload to server; And/or
If write request data are internal memory swap data and/or ephemeral data, then write request data are stored in nonvolatile memory;
And/or
Device also comprises:
If control module is also ephemeral data for write request data, ephemeral data is cached in internal memory, after at least one ephemeral data is merged into total ephemeral data, is stored into nonvolatile memory.
Alternatively, nonvolatile memory comprises journal disk.
Alternatively, if client or server meet accident when restarting, control module also for:
Read the control section of preserving in journal disk;
Journal disk is scanned, the data segment finding dispensing section, submit to section to identify from control section;
The data segment uploaded dispensing section successively, submit to section to identify;
After uploading, normally start client operating system.
Alternatively, control module also for:
Data segment in journal disk is submitted to server;
After receiving the confirmation of receipt that server returns, the control section of change data segment is for submitting section to;
When receiving after submission that server returns confirms, the control section of change data segment is available segments;
Reclaim the data segment of available segments mark.
Alternatively, this device also comprises:
Receiver module is used for: the read request receiving client operating system;
Control module also for:
In virtual machine buffer memory, data corresponding to read request are searched according to read request;
If find the data that read request is corresponding in virtual machine buffer memory, then read data, if do not find data corresponding to read request in virtual machine buffer memory, then read request is sent to server.
The third aspect, the disclosure provides a kind of virtual machine storage system, comprising:
Be configured with the client of virtual machine storage system implement device;
Server, communicates with client, provides service for the user for client.
The virtual machine storage system that the disclosure provides and its implementation and device, after receiving write request, by by write request data buffer storage in virtual machine buffer memory, and according to the state of server by write request data upload to server, the congestion problems produced by writing burst in virtual machine storage system in cloud computing can be solved.
According to below with reference to the accompanying drawings to detailed description of illustrative embodiments, further feature of the present invention and aspect will become clear.
Accompanying drawing explanation
Figure 1A illustrates the flow chart of the virtual machine storage system implementation method of one embodiment of the invention.
Figure 1B illustrates the structural representation of the virtual storage system of one embodiment of the invention.
Fig. 2 illustrates the flow chart of the virtual machine storage system implementation method of another embodiment of the present invention.
Fig. 3 A illustrates the structure chart of the VMCache of the virtual machine storage system implementation method of one embodiment of the invention.
Fig. 3 B illustrates the schematic flow sheet of the virtual machine storage system implementation method of one embodiment of the invention.
Fig. 4 A illustrates the schematic diagram of the page-map management of the virtual machine storage system implementation method of one embodiment of the invention.
Fig. 4 B illustrates the schematic diagram of the memory cache layout of the virtual machine storage system implementation method of one embodiment of the invention.
Fig. 4 C illustrates the journal disk structural representation of the virtual machine storage system implementation method of one embodiment of the invention.
Fig. 4 D illustrates the schematic diagram of the section state transitions of the virtual machine storage system implementation method of one embodiment of the invention.
Fig. 4 E illustrates the schematic diagram of the segment layout of the virtual machine storage system implementation method of one embodiment of the invention four.
Fig. 5 A illustrates the flow chart reading method in the virtual machine storage system implementation method of one embodiment of the invention.
Fig. 5 B illustrates the flow chart of write method in the virtual machine storage system implementation method of one embodiment of the invention.
Fig. 6 illustrates the flow chart of data reconstruction method in the virtual machine storage system implementation method of one embodiment of the invention.
Fig. 7 illustrates the flow chart of delayed submission confirmation method in the virtual machine storage system implementation method of one embodiment of the invention.And
Fig. 8 illustrates the structured flowchart of the virtual machine storage system implement device of one embodiment of the invention.
Embodiment
With reference to the accompanying drawings the present invention is described more fully, exemplary embodiment of the present invention is wherein described.
Figure 1A illustrates the flow chart of the virtual machine storage system implementation method of one embodiment of the invention.As shown in Figure 1A, this virtual machine storage system implementation method mainly comprises:
Step S102, receives the write request of client operating system.
Step S104, by write request data buffer storage in virtual machine buffer memory.
Step S106, according to the state of server by write request data upload to server.
Particularly, Figure 1B illustrates the structural representation of the virtual storage system of one embodiment of the invention, as shown in Figure 1B, client operating system 11 comprises application program 12, file system 13, virutal machine memory management 14, virtual disk interface 15 etc., VMM comprises IO processor 16, virtual machine buffer memory 17 and network storage protocol client 18 etc.VMM19 can be divided into IO processor 16, virtual machine buffer memory 17 and network storage protocol client 18 according to function, wherein, network storage protocol client 18 can storage system (NetworkFileSystem Network Based, or the client 10 of the system such as small computer system interface (InternetSmallComputerSystemInterface, iSCSI) NFS).The read-write requests receiving client operating system 11 is responsible for by IO processor 16, classifies, queues up and dispatch request; VMCache primary responsibility management VMM19 in write buffer memory; The client 10 that network storage protocol client 18 is responsible for realizing network storage protocol communicates with server 20.
Client operating system 11 realizes alternately by VMM19 and server 20, cache management depends on the virtual machine buffer memory 17 (VirtualMachineCache in VMM19, VMCache), this virtual machine buffer memory 17 particularly write request buffer memory independent of client operating system 11, be that granularity performs with block, cross-platform higher, do not need to change operating system simultaneously and both can be transplanted to other system easily, portable ability is higher.In cloud computing, several operation systems can be dynamically loaded into this client 10 and run.
Particularly, when the read-write requests of client operating system 11 arrives after VMM19, first queue up in IO processor 16, IO processor 16 again according to certain strategy as first-in first-out strategy (FirstInFirstOut, FIFO) from queue, select a request, pass to VMCache.Write request is first kept in VMCache, then decides these data the need of the problem uploaded and when upload by VMCache.As write request data write VMCache, for client operating system 11, write request just completes, without the need to being concerned about when these data upload onto the server 20 again.So just original simultaneous operation can be become asynchronous operation, reduce the time of response.The request received can be encapsulated as network storage protocol message by network storage protocol client 18, sends to server 20 according to network storage protocol.After the result of device 20 to be serviced returns, successively can upload, return to client operating system 11, complete association requests.
Further, step S106 can also comprise:
Judge the state of server 20, if the state of server 20 is idle, then by the data upload of write request to server 20, if the state of server 20 is busy, then by write request data buffer storage in virtual machine buffer memory 17.
Particularly, client 10 according to the state of server 20 by write request data upload to server 20, when server 20 is busy, client operating system 11 submit to write request can first temporarily leave in virtual machine buffer memory 17, when device 20 to be serviced is idle, then by data upload.
In one embodiment, the VMM19 of client 10 according to states such as the CPU of server 20, internal memory, Disk State, network occupancies, can directly or indirectly determine that server 20 is busy or idle.For example, when the cpu busy percentage that the VMM19 of client 10 detects server 20 is 90%, the VMM19 of client 10 judges that obtaining this moment server 20 is in busy state, can continue by write request data buffer storage in virtual machine buffer memory 17, not by this write request data upload; Continue to detect server 20CPU utilance, when detect cpu busy percentage be reduced to certain value as 50% time, when client 10 judges to obtain this moment, server 20 is in idle condition, can by write request data upload to server 20.
It should be noted that, although exemplarily to describe write request data upload according to the state of server to server with cpu busy percentage, it will be appreciated by those skilled in the art that the present invention should be not limited thereto.In fact, user can according to the standard of practical application scene settings server state, and the server busy degree under different application scenarioss is not sought unity of standard, and specific targets should by system manager's self-defining.Such as can judge server state, as long as the effect judging service state can be reached by server memory utilance, server disk occupancy etc.
The virtual machine storage system implementation method of the present embodiment, after receiving write request, by by write request data buffer storage in virtual machine buffer memory, and according to the state of server by write request data upload to server, the congestion problems produced by writing burst in virtual machine storage system in cloud computing can be solved.
Fig. 2 illustrates the flow chart of the virtual machine storage system implementation method of another embodiment of the present invention.The step that Fig. 2 and Figure 1A, 1B are identical or assembly have identical function, for simplicity's sake, omit the detailed description to these steps or assembly.
As shown in Figure 2, shown in the method shown in Fig. 2 and Figure 1A, the main distinction of method is, can also comprise after step S201:
Step S202, by write request data buffer storage at nonvolatile memory.
Further, this nonvolatile memory comprises journal disk.
Particularly, what be operated in operating system lower floor writes the relevant data structure of metadata that the data of preserving in buffer memory may be topmost paper systems, daily record or transactional operation, lose these data and may affect the failover capabilities of file system, time serious, may file system crash be caused.If can not ensure non-volatile, the straight write operation semanteme of file system will lose efficacy, and this will mean that the data that user preserved in application software is as Word also may be lost.For user, this is unacceptable.By by the nonvolatile memory of write request data buffer storage in virtual machine buffer memory as journal disk, solid state hard disc, non-volatile read-only memory (Non-VolatileRandomAccessMemory, NVROM) etc., when unexpected power-off or system crash, the user data in buffer memory can not be lost.The problem that the data power down that can solve buffer memory is lost.
Further, can also comprise after step S201:
Step S204, according to the type decided of write request data whether by write request data upload to server.
Further, step S204 can comprise following situation:
Situation one, if write request data are user data, then by write request data upload to server.
Write request data if write request data are internal memory swap data and/or ephemeral data, are then stored in nonvolatile memory by situation two.
Wherein, if write request data are ephemeral data, ephemeral data is cached in internal memory; After at least one ephemeral data is merged into total ephemeral data, then be stored into nonvolatile memory.
Particularly, internal memory swap data (MemoryswapData, MD) is to support that the internal memory of client operating system exchanges (swap) and paging (paging) and produce write data in cloud computing system.It is a kind of method utilizing disk space exented memory space that modern operating system proposes to solve physical memory deficiency that internal memory exchanges with paging, the party's ratio juris is the expansion of a part of space in designated disk as physical memory, then take page frame replacement in disk the page infrequently used in physical memory or current inertia process, make to discharge physical memory space and may be used for other processes.Ephemeral data (TemporaryData, TD) comprises intermediate object program, cache file and the journal file etc. that produce in program operation process.User data (UserData, UD) refers to the file edited by user or download, and compared with first two data, the life cycle of this kind of data is long, needs to preserve for a long time.
Client can determine the state of server, writes caching mechanism by the VMCache in VMM, effectively can manage and write buffer memory.According to writing the data life period feature different with shared demand in cloud computing, VMCache can write data and be divided into ephemeral data, internal memory swap data and user data, for various data characteristics, client can take different write-in policies, only need by UD data write server, MD and TD data can be stored in VMM this locality, by distinguishing different data types, some unnecessary data of uploading onto the server are filtered, decrease the waste of server resource, thus decrease the read-write operation of server data; Further, for TD data, save bit by bit a certain amount of in the buffer, then write journal disk, thus avoid the frequent read-write operation of journal disk.
The virtual machine storage system implementation method of the present embodiment, by adding non-volatile magnetic disk, overcoming the congestion problems write and happen suddenly and cause, preventing data power down from losing; UD data by Data classification, thus only need be write server by client, MD and TD data can be stored in VMM this locality, thus decrease the read-write operation of server data; Further, for TD data, save bit by bit a certain amount of in the buffer, then write journal disk, thus avoid the frequent read-write operation of journal disk.
In one embodiment, show one virtual machine storage system implementation method example particularly, assembly identical with above embodiment in the present embodiment or step have identical function, for simplicity's sake, omit the detailed description to these assemblies or step.
Particularly, Fig. 3 A illustrates the structure chart of the VMCache of the virtual machine storage system implementation method of one embodiment of the invention, and as shown in Figure 3A, VMCache mainly comprises three main modular: memory cache 31, journal disk 32 and controller 30.Memory cache 31 can refer to the memory headroom independent of client operating system managed by VMCache.Repeat in order to avoid writing buffer memory with client operating system inside, memory cache 31 in VMCache writes data for what preserve TD type, Main Function is unified submission the to that scrappy TD write request gathered together, and reduces access times and the space junk of journal disk 32.
Journal disk 32 can be additional disk on the client, and also can be independently disk, this disk manages with log-structured form, and when running for cache client, the various of generation write data.Journal disk 32 has three functions: the first, for life cycle be the data of switching on and shutting down as TD, MD, will absorb all upload request writing data; The second, when server is busy, can by buffer memory UD data to reduce to write the impact of burst on systematic function; Three, when client or server meet accident restart time, help system can recover UD categorical data.
Controller 30 is cores of VMCache, and primary responsibility controls data batchmove, the memory block of locating read-write request and the memory space of managing internal memory buffer memory 31 and journal disk 32 between memory cache 31, journal disk 32 and server.Define relevant data structure in controller 30, comprise the block Hash table for data location and bitmap block, for daily record head, tail pointer and free block bitmap etc. that journal disk 32 manages.
VMCache can take different cache policies according to the feature of different types of data.Fig. 3 B illustrates the schematic flow sheet of the virtual machine storage system implementation method of one embodiment of the invention, as shown in Figure 3 B, when VMCache receives write request, grader 34 can first according to the type of write request data, by write request data dispatch to different waiting lists.Request in different queue can be treated with a certain discrimination according to respective feature: the data block that TD categorical data writes at every turn is less, it can first be cached in memory cache 31 by the memory cache manager 35 of VMCache, and then several write requests are merged into a write request (such as the data of 4 TD data being merged into write request data), submit to journal disk 32.MD categorical data can be work because client operating system memory headroom is not enough to maintain current all processes simultaneously, from the internal memory of client operating system management, isolate the buffer memory of a part as MD in VMM produce, its effect is little, the journal disk 32 so MD categorical data can write direct by VMCache.The life cycle of TD and MD can between switching on and shutting down, that is these data can empty after client restarts, and the writing (about 2G) of MD and TD in switching on and shutting down is more much smaller than the capacity of journal disk 32, can think that journal disk 32 can not be write full by MD and TD, therefore, TD and MD can be buffered in client by VMCache, need not upload onto the server.But the data of UD type are still meaningful after client restarts, based on the consideration to UD fail safe, VMCache can not in memory cache 31 buffer memory UD, but directly to be write in journal disk 32.According to server state, the UD being buffered in client can upload onto the server by VMM the most at last.
The virtual machine storage system implementation method of the present embodiment, after receiving write request, by by write request data buffer storage in virtual machine buffer memory, and whether upload according to the Determines of data type and server or when upload, the congestion problems produced by writing burst in virtual machine storage system in cloud computing can be solved, prevent data power down from losing, and decrease the read-write operation of server data, and for TD data, save bit by bit a certain amount of in the buffer, write journal disk again, thus avoid the frequent read-write operation of journal disk.
In one embodiment, the virtual machine buffer memory management method in a kind of virtual machine storage system implementation method is shown.Assembly same as the previously described embodiments in the present embodiment or step have identical function, for simplicity's sake, omit the detailed description to these assemblies or step.
Particularly, the cache management of VMCache can comprise: page-map management, memory cache management and journal disk management.Particularly, Fig. 4 A illustrates the schematic diagram of the page-map management of the virtual machine storage system implementation method of one embodiment of the invention, whether as shown in Figure 4 A, be updated according to the page in memory space, the page in virtual memory space can be divided into the renewal page and parent page by client.Wherein, upgrade the page and refer to and revised by client operating system 11, and at the page of client-cache, as the page 1 ~ 3 in Fig. 4 A.These pages are at page Hash table 42 (PageHashTable, PHT) registration in, wherein, page Hash table 42 is one of data structures important in controller, page Hash table 42 may be used for by the position of the page in client-cache of changing in record virtual memory space, can be found their position by inquiry PHT.Parent page refers to the page that client operating system is not modified or removed from client-cache, as the page in figure 4 ~ 5.These pages do not have corresponding list item in PHT, need access services device just can obtain.PHT and virtual memory space one_to_one corresponding, can be conducted interviews by the PHT pointer field of virtual disk table (VirtualDiskTable, VDT).
Fig. 4 B illustrates the schematic diagram of the memory cache layout of the virtual machine storage system implementation method of one embodiment of the invention, and as shown in Figure 4 B, memory cache is made up of buffer memory head (RAMHeader) and some pages (Pages, PGs).Wherein, buffer memory head comprises head and the afterbody of vdisk_id, inverted list and fifo queue.In VMCache, can set and only have the virtual disk preserving TD categorical data just can have memory cache, can not shared drive buffer memory between different virtual disks.Wherein, vdisk_id refers to the virtual disk id having this memory cache example; Inverted list is with the page in memory cache for index, the mapping relations of the page in each page and virtual memory space in record buffer memory; FIFO head, tail pointer then marked the position that the page distributes and reclaims respectively.
Fig. 4 C illustrates the journal disk structural representation of the virtual machine storage system implementation method of one embodiment of the invention, and as shown in Figure 4 C, journal disk memory space 40 can be divided into a superblock and two subregions.The configuration information of journal disk can be preserved, as the initial of section size, subregion 0 and subregion 1 and termination page number etc. in superblock.Subregion 0 primary responsibility buffer memory TD and MD categorical data, the responsible buffer memory UD data of subregion 1.The structure of two subregions is identical, is all made up of subregion head and some sections, and subregion head have recorded the pointer pointing to control section.Section is made up of paragraph header and some pages, and paragraph header comprises vdisk_id, page inverted list, page bitmap and segment type 4 part, and first two section is identical with the meaning of the corresponding data structure in memory cache head, the effective page in page bitmap essential record section.Segment type has four selectable value, be respectively dispensing section (AssignmentSeGment, ASG), submit section (CommittingSeGment, CSG), exhausting section (RecycleSeGment to, and data segment (DataSeGment, DSG) RSG).First three type is collectively referred to as control section, has mark, management function, can be used in identification data section state, is responsible for the data segment of segmentation different conditions.Except control section, other all sections all belong to data segment.Data segment does not possess management function, only for data cached.
For convenience of description, four kinds of states can be defined to section: idle (Free): this data segment is in idle condition, writes this block without any data; Take (Occupied): this data segment is data cached but not yet submit to; Submit to (Committing): the content of this data segment is submitted to server, but not yet receives the confirmation of server; Available (valid): receive server and submit to and successfully to confirm and content in data segment is consistent with the corresponding page on server.
Because subregion 0 does not exist the problem of upload server, therefore section wherein only may locate first two state, namely idle or take.Otherwise, subregion 1 stage casing then may be in one of above-mentioned 4 kinds of states.
Fig. 4 D illustrates the schematic diagram of the section state transitions of the virtual machine storage system implementation method of one embodiment of the invention, as shown in Figure 4 D, section State Transferring specifically can comprise: write (write), write data in journal disk from memory cache or client operating system; Submit to (commit): upload the data to server, server returns the message received and confirms; Submit to and confirm (committedack): server returns confirmation, shows that data write in the disk of server end by server; Rewrite (rewrite): the page related in data segment is revised by last layer.
Fig. 4 E illustrates the schematic diagram of the segment layout of the virtual machine storage system implementation method of one embodiment of the invention, as shown in Figure 4 E, journal disk subregion 1 comprises different conditions section, and in the ordinary course of things, the section of equal state flocks together and can form continuous print memory space.The Duan Ze of different conditions is separated by control section; The free segment that ASG is free segment and takies on segment boundary, CSG is the section of taking and the section of taking submitted on segment boundary, and RSG submits section and the available or borderline submission section of free segment to.It should be noted that and may comprise data cached in control section yet, is not uniquely DSG unlike the value of the segment type field of control section with data segment.So when data segment state changes, control section needs to adjust accordingly.Restart if system meets accident, as long as fault recovery routine scans journal disk, just can return to the data segment state before startup.Simultaneously to needing the data segment uploaded to distinguish, ensure that the fail safe of UD data.
Because the time overhead scanning whole journal disk searching control section is larger, usually need tens seconds even several minutes.In order to reduce this expense, the pointer pointing to control section can be preserved in subregion head.Like this, the pointer in subregion head during startup, only need be inquired about, just can positioning control section.But it is when every secondary segment state changes that this method there will be a problem, all needs synchronized update control section pointer, this just makes each write operation all need to rewrite 2 diverse locations on disk, thus causes declining to a great extent of write performance.The strategy of regular update can be taked to address this problem.First these three pointers are saved in internal memory exactly, at regular intervals by disk write once.In fact the method is exactly by postponing more to newly arrive to decrease the write number of times of disk and the probability of random write.When restarting when meeting accident, the control section pointer in subregion head may be out-of-date, cannot indicate the tram of control section, now still needs to find control section by the method for scanning disk.But, because the position of control section should near control section pointer pointed location, therefore carrying out scanning disk according to the prompting of control section pointer also can very fast finding, without the need to scanning whole disk.
The virtual machine storage system implementation method of the present embodiment, VMCache introduces the double-deck buffer structure of internal memory and journal disk combination, and based on the journal disk management method of control section, on the one hand, the data of writing not needing to upload can be retained in client, reduce read-write flow; When ensure that system is surprisingly restarted on the other hand, user data is not lost; Further, for TD data, save bit by bit a certain amount of in the buffer, then write journal disk, thus avoid the frequent read-write operation of journal disk; VMCache meets cloud computing demand, effectively can promote the write performance of all kinds data.
In one embodiment, the example of the reading/writing method in virtual machine storage system implementation method is shown.The assembly that the present embodiment is same as the previously described embodiments or step have identical function, for simplicity's sake, omit the detailed description to these assemblies or step.
Particularly, this virtual machine storage system implementation method can also comprise:
Receive the read request of client operating system.
In virtual machine buffer memory, data corresponding to read request are searched according to read request.
If find the data that read request is corresponding in virtual machine buffer memory, then read data, if do not find data corresponding to read request in virtual machine buffer memory, then read request is sent to server.
Further, Fig. 5 A illustrates the flow chart reading method in the virtual machine storage system implementation method of one embodiment of the invention, and as shown in Figure 5A, this flow process reading method comprises:
Step S302, client operating system initiates read request, and the parameter of read request can be <vdisk_id, init_sec, sec_no>, wherein init_sec is the initial sector of access storage areas, and sec_no is the quantity of access sector.
Step S304, according to this read request, inquiry paging, the algorithm adopted during inquiry is PGi (i=1 ~ n).Access storage areas, in units of the page, is divided into several pages PG1 ~ PGn by VMCache, and each like this requests for page just can be described by <vdisk_id, page_id, sec_bitmap> tlv triple.Sec_bitmap mainly may be used for describing and needing the situation of access sector when access storage areas size is less than the page in the page.A sector in each bit representing pages in sec_bitmap, 1 represents that corresponding sector needs are accessed.VMCache can read each page successively.
Step S306, to the PGi of any one page, with the page_id of PGi for index, the list item that retrieval is corresponding in PHT [vdisk_id].If do not find, then show that desired page is not in client-cache, then perform step S307; If found, then perform step S308.
Step S307, to the server request page.
Step S308, reads the page from memory cache or journal disk.Such as, if the page retrieved is PGk, from internal memory or journal disk, PGk is read.
Step S310, compares sector bitmap.Can the effective sector can compared in PGk " cover " sector will accessed in PGi, can its formula be (PGk->sec_bitmap^PGi->sec_bit map) & PGi->sec_bitmap>0?); If can cover, then read this page, and perform step S314; If can not cover, then perform step S312.
Step S312, to the server request page.According to the judged result of step S310, can not cover if the effective sector in PGk " can not cover " sector will accessed in PGi, then download PGi from server.
Step S316, by sector bitmap merge page, upgrades the page in buffer memory.According to sec_bitmap, the downloading page on server is merged.
Step S314, judges whether read request completes.Namely by algorithm ifi=n? judge whether that reading completes, if judged result is affirmative, represent that reading completes; If judged result is negative, then continue to read the next page.Said process will be repeated, until all pages read complete when reading the next page.
Further, Fig. 5 B illustrates the flow chart of write method in the virtual machine storage system implementation method of one embodiment of the invention, as shown in Figure 5 B, suppose that client operating system initiates write request, the parameter of write request can be <vdisk_id for parameter, init_sec, sec_no>, write request operation can perform different workflows according to the difference writing data type.Specifically can comprise the following steps:
Step S402, client operating system initiates write request, and the parameter of read request can be <vdisk_id, init_sec, sec_no>, wherein init_sec is the initial sector of access storage areas, and sec_no is the quantity of access sector.
Step S404, judges whether these write request data are TD data.Can its evaluation algorithm be ifVDT [vdisk_id]->type=TD? if the determination result is YES, then perform step S406, otherwise perform step S415.
Step S406, VMCache carry out paging to access storage areas.
Step S408, judges whether this page exists in internal memory.If current page is PGi, the VMCache page that search PGi is corresponding in PHT [vdisk_id], if find, then illustrates that this page exists in internal memory, then perform step S410, upgrade target pages, revise sec_bitmap value; Otherwise perform step S413.
Step S413, to judge in buffer memory if so, then to perform whether available free space step S414, distribute a free page and write; Otherwise representation space is full, operation to be committed re-executes page write operation after reclaiming space.
Step S411, judges whether write algorithm completes.Wherein, by algorithm i=n? judge whether that executing this writes algorithm.If so, then perform step S412, write operation completes, otherwise continues to write one page.
Step S412, VMCache carry out segmentation to access storage areas, write journal disk successively.Segmentation algorithm is segmentation SGj (j=1...m).
Step S416, judges whether write request data type is MD data.If so, then step S417 is performed, otherwise, represent that these data are UD data, perform step S419, UD categorical data can be write direct in journal disk subregion 1.
Step S417, write journal disk subregion 0.Specifically can comprise and segmentation is carried out to access storage areas, and then write each data segment successively, upgrade PHT table.
Step S419, judges whether available free section of journal disk subregion 1.If the determination result is YES, then show journal disk 1 less than, then perform step S420; If judged result is no, show that journal disk 1 is full, then, after space is reclaimed in operation to be committed, re-execute write algorithm and perform step S420.
Step S420, distributes a free segment, by write request data write journal disk subregion 1.Specifically can comprise and segmentation is carried out to access storage areas, and then write each data segment successively, upgrade PHT table.
Step S418, judges whether write algorithm completes.Wherein, by algorithm j=m? judge whether that executing this writes algorithm.If so, then perform step S412, write operation completes, otherwise continues to write one section.
The virtual machine storage system implementation method of the present embodiment, after receiving read-write requests, whether the Determines according to data type and server is uploaded or when is uploaded, and can solving the congestion problems produced by writing burst in virtual machine storage system in cloud computing, preventing data power down from losing.
In one embodiment, the example of the commit method in virtual machine storage system implementation method is shown.The assembly that the present embodiment is same as the previously described embodiments or the identical step of label have identical function, for simplicity's sake, omit the detailed description to these assemblies or step.
Particularly, submit operation refers to and " dirty " data is write low layer buffer memory or memory from high-rise buffer memory, then reclaims the process of high-rise spatial cache.The data that VMCache is implemented in memory cache and journal disk by background thread are submitted to and space reclamation.In submission strategy, strategy can be submitted to according to determining on submission opportunity and frequency.Possible submission strategy actively submits a strategy to, namely frequently submits to, and its advantage is that Information Security is better, and high-rise buffer memory is often in idle condition, effectively can process and write burst; When shortcoming is underload, high-rise buffer memory not only effectively can not reduce and writes flow, adds one deck data copy on the contrary.Another kind of possible submission strategy is passive submission strategy, and its advantage is the repetition write operation that can absorb the same page, is convenient to be optimized for write operation content; Its shortcoming is that Information Security is poor, because high-rise buffer memory is in normal completely state, therefore not obvious to the lifting of burst write performance.The factor of strategy is submitted in impact to, and to mainly contain two: one be the ratio that in high-rise buffer memory, occupied cache blocks accounts for total capacity, namely the degree that " expires " of buffer memory, can referred to as degree; Another is the fail safe of data, and namely data can referred to as reliable susceptibility to the sensitivity of reliability.
When considering degree and reliable susceptibility, trigram models <t_low, t_high, a T> can be adopted to describe and to submit strategy to.Wherein t_low and t_high is two threshold values (t_low<t_high) representing degree, can represent that degree is on the impact submitting strategy to t_low and t_high.If Co represents current cache degree, as C0<t_low, do not perform submission; As t_low<C0<t_high, submit to when performing idle; As C0>t_high, perform " doing one's best " and submit to.
Different submission strategies realizes by adjustment t_low and t_high.Such as t_low is larger, submits to frequency lower, submits to strategy more passive.In extreme circumstances, as t_low=1, system can perform submit operation after buffer memory completely.On the contrary, t_high is less, submits to frequency lower, submits to strategy more positive, and as t_high=0, the carrying out all can " done one's best " after each write buffer memory is submitted to.If be in the intermediate zone of t_low and t_high, then illustrate that the degree in this interval is little on submitting the impact of strategy to, system can according to busy situation, determines submission opportunity voluntarily.
Last element T in tlv triple represents the maximum at adjacent twice submission time interval.When last time, submission time was more than T for distance, system will start submit operation, to prevent loss of data.T embodies reliable susceptibility to the impact submitting strategy to.In VMCache, the caching design of TD and UD relates to the problem of submission.Wherein TD can automatically empty after system restart, so reliable susceptibility is not high, T value can be set to infinity; The reliable susceptibility of UD is higher, but UD is buffered in the buffer memory of journal disk, and in the unexpected power-off of system or after restarting, the data in buffer memory can not be lost, so T value is also too little without the need to arranging.But, if data cached prolonged stay is in client, also there is certain risk.Such as: if the customer end A that user logs in is when closing, the UD data in buffer memory fail to upload onto the server, and so user is when other customer end B log in, and just cannot have access to latest data.Therefore T can not be set to infinity.Therefore, according to engineering experience, the value of T can be set to 5 ~ 10 minutes.
The virtual machine storage system implementation method of the present embodiment, after receiving read-write requests, whether the Determines according to data type and server is uploaded or when is uploaded, and can solving the congestion problems produced by writing burst in virtual machine storage system in cloud computing, preventing data power down from losing.
In one embodiment, the example of the fault recovery method in virtual machine storage system implementation method is shown.The assembly that the present embodiment is same as the previously described embodiments or the identical step of label have identical function, for simplicity's sake, omit the detailed description to these assemblies or step.
Particularly, client and server are when occurring surprisingly restarting, and VMCache can ensure fail safe or the restorability of UD categorical data.
Such as, if when client computer is surprisingly restarted, UD data can be recovered from client.Fig. 6 illustrates the flow chart of data reconstruction method in the virtual machine storage system implementation method of one embodiment of the invention, and as shown in Figure 6, the method specifically can comprise:
Step S602, reads the control section of preserving in journal disk;
Step S604, scans journal disk from control section, the data segment finding dispensing section, submit to section to identify;
Step S606, the data segment uploaded dispensing section successively, submit to section to identify;
Step S608, after uploading, normally starts client operating system.
Particularly, VMCache records the data mode before restarting by the control section in journal disk.After client restarts, VMCache according to following algorithm, will upload and not preserve UD data: 1) read the control section pointer that subregion 1 is preserved in subregion head; 2) from control section pointer, scan journal disk, find ASG, CSG and RSG; 3) data segment between ASG to RSG is uploaded successively; 4) be recovered, normal start up system.Like this, can server or client meet accident restart time, can ensure that data are not lost.
Further, for preventing server from surprisingly closing, guaranteeing that server data is still retained in client, the present invention proposes a kind of delayed submission confirmation method.Fig. 7 illustrates the flow chart of delayed submission confirmation method in the virtual machine storage system implementation method of another embodiment of the present invention, and as shown in Figure 7, the method can comprise:
Step S702, is submitted to server by the data segment in journal disk.
Step S704, when the confirmation of receipt receiving server and return, changes the control section of this data segment for submitting section to.
Step S706, confirm when receiving the submission that server returns, the control section changing this data segment is available segments.
Step S708, reclaims the data segment of this available segments mark.
Particularly, when the data upload of client computer is after server, data buffer storage is first carried out improving performance by server in internal memory.But, when server meet accident restart time, the data in internal memory just may be lost.In order to prevent the generation of this situation, the mechanism that VMCache can adopt delayed submission to confirm: after a data segment is on a client uploaded onto the server, server only returns confirmation of receipt to client.After client receives confirmation, then the state of data segment is adjusted to submit state, wherein, the data segment of submit state must retain, and does not allow to reclaim.After server writes disk the data segment that client computer is submitted to, can send and submit confirmation to, at this moment the state of data segment just can shift as available.When system needs, can reclaim the data segment of upstate.This delayed acknowledgement mechanism when server meets accident closedown, can guarantee that the data in server memory are still kept in the data segment of client.Therefore, after fault recovery, only need requesting client again to upload the data segment being in submit state, just can recover data.
The virtual machine storage system implementation method of the present embodiment, VMCache introduces the double-deck buffer structure of internal memory and journal disk combination, and based on the journal disk management method of control section, on the one hand, the data of writing not needing to upload can be retained in client, reduce read-write flow; When ensure that system is surprisingly restarted on the other hand, user data is not lost; Further, for TD data, save bit by bit a certain amount of in the buffer, then write journal disk, thus avoid the frequent read-write operation of journal disk; VMCache meets cloud computing demand, effectively can promote the write performance of all kinds data.
Fig. 8 illustrates the structured flowchart of the virtual machine storage system implement device of one embodiment of the invention.As shown in Figure 8, this virtual machine storage system implement device 800 mainly comprises:
Receiver module 802, for receiving the write request of client operating system;
Cache module 804, is connected with receiver module 802, for by write request data buffer storage in virtual machine buffer memory;
Control module 806, is connected with cache module 804, for according to the state of server by write request data upload to server.
The virtual machine storage system implement device of the present embodiment, after receiving write request, by by write request data buffer storage in virtual machine buffer memory, and according to the state of server by write request data upload to server, the congestion problems produced by writing burst in virtual machine storage system in cloud computing can be solved.
In one embodiment, control module 806 also for:
Judge the state of server, if the state of server is idle, then by the data upload of write request to server, if the state of server is busy, then by write request data buffer storage in virtual machine buffer memory.
In one embodiment, cache module 804 also for:
By write request data buffer storage at nonvolatile memory.
In one embodiment, control module 806 also for:
According to the type of write request data by write request data upload to server.
In one embodiment, control module 806 also for:
If write request data are user data, then by write request data upload to server; And/or
If write request data are internal memory swap data and/or ephemeral data, then write request data are stored in nonvolatile memory.
In one embodiment, control module 806 also for:
If write request data are ephemeral data, ephemeral data is cached in internal memory;
After at least one ephemeral data is merged into total ephemeral data, be stored into nonvolatile memory.
In one embodiment, control module 806 also for:
Data segment in journal disk is submitted to server;
When the confirmation of receipt receiving server and return, the control section of change data segment is for submitting section to;
Confirm when receiving the submission that server returns, the control section of change data segment is available segments;
Reclaim the data segment of available segments mark.
In one embodiment, nonvolatile memory comprises journal disk.
In one embodiment, if client or server meet accident when restarting, control module 806 also for:
Read the control section of preserving in journal disk;
Journal disk is scanned, the data segment finding dispensing section, submit to section to identify from control section;
The data segment uploaded dispensing section successively, submit to section to identify;
After uploading, normally start client operating system.
In one embodiment, receiver module 802 also for: receive the read request of client operating system.Control module 806 also for: in virtual machine buffer memory, search data corresponding to read request according to read request; If find the data that read request is corresponding in virtual machine buffer memory, then read data, if do not find data corresponding to read request in virtual machine buffer memory, then read request is sent to server.
In one embodiment, present invention also offers a kind of virtual machine storage system, comprising: the client of configuring virtual machine storage system implement device; Server, communicates with client, provides service for the user for client.
Description of the invention provides in order to example with for the purpose of describing, and is not exhaustively or limit the invention to disclosed form.Many modifications and variations are obvious for the ordinary skill in the art.Selecting and describing embodiment is in order to principle of the present invention and practical application are better described, and enables those of ordinary skill in the art understand the present invention thus design the various embodiments with various amendment being suitable for special-purpose.

Claims (17)

1. a virtual machine storage system implementation method, is characterized in that, comprising:
Receive the write request of client operating system;
By described write request data buffer storage in virtual machine buffer memory;
According to the state of server by described write request data upload to described server.
2. method according to claim 1, is characterized in that, described write request data upload comprises to described server by the state according to server:
Judge the state of described server, if the state of described server is idle, then by the data upload of described write request to described server, if the state of described server is busy, then by described write request data buffer storage in described virtual machine buffer memory.
3. method according to claim 1, is characterized in that,
Described write request data buffer storage is comprised at virtual machine buffer memory: by described write request data buffer storage at nonvolatile memory;
And/or
Described method also comprises:
According to the type of described write request data by described write request data upload to described server.
4. method according to claim 3, is characterized in that,
Described write request data upload comprises to described server by the type according to described write request data:
If described write request data are user data, then by described write request data upload to described server; And/or
If described write request data are internal memory swap data and/or ephemeral data, then described write request data are stored in described nonvolatile memory;
And/or
Described method also comprises:
If described write request data are ephemeral data, described ephemeral data is cached in internal memory, after at least one ephemeral data is merged into total ephemeral data, is stored into described nonvolatile memory.
5. method according to any one of claim 1 to 4, is characterized in that, described nonvolatile memory comprises journal disk.
6. method according to claim 5, is characterized in that, if client or server meet accident when restarting, also comprises:
Read the control section of preserving in journal disk;
Described journal disk is scanned, the data segment finding dispensing section, submit to section to identify from described control section;
The data segment uploaded described dispensing section successively, submit to section to identify;
After uploading, normally start client operating system.
7. method according to claim 5, is characterized in that, also comprises:
Data segment in journal disk is submitted to server;
When the confirmation of receipt receiving described server and return, change the control section of described data segment for submitting section to;
Confirm when receiving the submission that described server returns, the control section changing described data segment is available segments;
Reclaim the data segment of described available segments mark.
8. method according to any one of claim 1 to 4, is characterized in that, also comprises:
Receive the read request of client operating system;
In virtual machine buffer memory, data corresponding to described read request are searched according to described read request;
If find data corresponding to described read request in described virtual machine buffer memory, then read data, if do not find data corresponding to described read request in described virtual machine buffer memory, then described read request is sent to server.
9. a virtual machine storage system implement device, is characterized in that, comprising:
Receiver module, for receiving the write request of client operating system;
Cache module, is connected with described receiver module, for by described write request data buffer storage in virtual machine buffer memory;
Control module, is connected with described cache module, for according to the state of server by described write request data upload to described server.
10. device according to claim 9, it is characterized in that, described control module is also for the state that judges described server, if the state of described server is idle, then by the data upload of described write request to described server, if the state of described server is busy, then by described write request data buffer storage in described virtual machine buffer memory.
11. devices according to claim 9, is characterized in that,
Described cache module is used for described write request data buffer storage at nonvolatile memory;
And/or
Described device also comprises:
Described control module also for according to the type of described write request data by described write request data upload to described server.
12. devices according to claim 11, is characterized in that,
Described control module also for:
If described write request data are user data, then by described write request data upload to described server; And/or
If described write request data are internal memory swap data and/or ephemeral data, then described write request data are stored in described nonvolatile memory;
And/or
Described device also comprises:
If described control module is also ephemeral data for described write request data, described ephemeral data is cached in internal memory, after ephemeral data described at least one is merged into total ephemeral data, is stored into described nonvolatile memory.
13. devices according to any one of claim 9 to 12, it is characterized in that, described nonvolatile memory comprises journal disk.
14. devices according to claim 13, is characterized in that, if client or server meet accident when restarting, described control module also for:
Read the control section of preserving in journal disk;
Described journal disk is scanned, the data segment finding dispensing section, submit to section to identify from described control section;
The data segment uploaded described dispensing section successively, submit to section to identify;
After uploading, normally start client operating system.
15. devices according to claim 13, is characterized in that, described control module also for:
Data segment in journal disk is submitted to server;
When the confirmation of receipt receiving described server and return, change the control section of described data segment for submitting section to;
Confirm when receiving the submission that described server returns, the control section changing described data segment is available segments;
Reclaim the data segment of described available segments mark.
16. devices according to any one of claim 9 to 12, is characterized in that, also comprise:
Described receiver module is used for: the read request receiving client operating system;
Described control module also for:
In virtual machine buffer memory, data corresponding to described read request are searched according to described read request;
If find data corresponding to described read request in described virtual machine buffer memory, then read data, if do not find data corresponding to described read request in described virtual machine buffer memory, then described read request is sent to server.
17. 1 kinds of virtual machine storage systems, is characterized in that, comprising:
Be configured with at least one client as the virtual machine storage system implement device as described in arbitrary in claim 9 to 16;
Server, communicates with described client, provides service for the user for client.
CN201410390308.1A 2014-08-11 2014-08-11 Virtual machine storage system and its implementation and device Active CN105376269B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410390308.1A CN105376269B (en) 2014-08-11 2014-08-11 Virtual machine storage system and its implementation and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410390308.1A CN105376269B (en) 2014-08-11 2014-08-11 Virtual machine storage system and its implementation and device

Publications (2)

Publication Number Publication Date
CN105376269A true CN105376269A (en) 2016-03-02
CN105376269B CN105376269B (en) 2019-11-26

Family

ID=55378073

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410390308.1A Active CN105376269B (en) 2014-08-11 2014-08-11 Virtual machine storage system and its implementation and device

Country Status (1)

Country Link
CN (1) CN105376269B (en)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106325772A (en) * 2016-08-23 2017-01-11 惠州市拉维尼科技有限公司 Data processing method
CN106406981A (en) * 2016-09-18 2017-02-15 深圳市深信服电子科技有限公司 Disk data reading/writing method and virtual machine monitor
CN107239319A (en) * 2016-03-29 2017-10-10 阿里巴巴集团控股有限公司 The date storage method and device of a kind of virtual machine
CN107395732A (en) * 2017-07-29 2017-11-24 安徽云帮邦网络技术有限公司 A kind of business data storage inquiry cloud platform
CN109426623A (en) * 2017-08-29 2019-03-05 深圳市中兴微电子技术有限公司 A kind of method and device reading data
CN109617736A (en) * 2018-12-26 2019-04-12 杭州和利时自动化有限公司 A kind of data transmission method and relevant apparatus

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020116404A1 (en) * 2000-06-07 2002-08-22 Transact In Memory, Inc. Method and system for highly-parallel logging and recovery operation in main-memory transaction processing systems
CN1955939A (en) * 2006-10-13 2007-05-02 清华大学 Backup and recovery method based on virtual flash disk
CN102713828A (en) * 2011-12-21 2012-10-03 华为技术有限公司 Multi-device mirror images and stripe function-providing disk cache method, device, and system
CN103699063A (en) * 2013-11-28 2014-04-02 歌尔声学股份有限公司 Device and method for acquiring off-line data of MES (Manufacturing Execution Method) system
CN103765406A (en) * 2011-06-30 2014-04-30 亚马逊科技公司 Methods and apparatus for remotely updating executing processes

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020116404A1 (en) * 2000-06-07 2002-08-22 Transact In Memory, Inc. Method and system for highly-parallel logging and recovery operation in main-memory transaction processing systems
CN1955939A (en) * 2006-10-13 2007-05-02 清华大学 Backup and recovery method based on virtual flash disk
CN103765406A (en) * 2011-06-30 2014-04-30 亚马逊科技公司 Methods and apparatus for remotely updating executing processes
CN102713828A (en) * 2011-12-21 2012-10-03 华为技术有限公司 Multi-device mirror images and stripe function-providing disk cache method, device, and system
CN103699063A (en) * 2013-11-28 2014-04-02 歌尔声学股份有限公司 Device and method for acquiring off-line data of MES (Manufacturing Execution Method) system

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107239319A (en) * 2016-03-29 2017-10-10 阿里巴巴集团控股有限公司 The date storage method and device of a kind of virtual machine
CN107239319B (en) * 2016-03-29 2021-05-28 阿里巴巴集团控股有限公司 Data storage method and device of virtual machine
TWI735542B (en) * 2016-03-29 2021-08-11 香港商阿里巴巴集團服務有限公司 Data storage method and device of virtual machine
CN106325772A (en) * 2016-08-23 2017-01-11 惠州市拉维尼科技有限公司 Data processing method
CN106406981A (en) * 2016-09-18 2017-02-15 深圳市深信服电子科技有限公司 Disk data reading/writing method and virtual machine monitor
CN107395732A (en) * 2017-07-29 2017-11-24 安徽云帮邦网络技术有限公司 A kind of business data storage inquiry cloud platform
CN109426623A (en) * 2017-08-29 2019-03-05 深圳市中兴微电子技术有限公司 A kind of method and device reading data
CN109617736A (en) * 2018-12-26 2019-04-12 杭州和利时自动化有限公司 A kind of data transmission method and relevant apparatus

Also Published As

Publication number Publication date
CN105376269B (en) 2019-11-26

Similar Documents

Publication Publication Date Title
CN105376269A (en) Virtual machine storage system, virtual machine storage system realizing method and virtual machine storage system realizing device
CN105549905B (en) A kind of method that multi-dummy machine accesses distributed objects storage system
US10725878B2 (en) Storage apparatus, storage system, and control method of storage system for dynamically securing free space when a storage apparatus is disused
CN102667772B (en) File level hierarchical storage management system, method, and apparatus
CN107391774B (en) The rubbish recovering method of log file system based on data de-duplication
US20180107593A1 (en) Information processing system, storage control apparatus, storage control method, and storage control program
US10565075B2 (en) Storage device and block storage method based on the storage device
KR20120090965A (en) Apparatus, system, and method for caching data on a solid-state strorage device
US8843602B2 (en) Network boot system
CN108733306B (en) File merging method and device
CN102511044A (en) Method for deleting the data and device thereof
CN101387987A (en) Storage device, method and program for controlling storage device
CN110945486B (en) Storage fragment management method and terminal
US7251716B2 (en) Method and system for data processing with recovery capability
US20100023532A1 (en) Remote file system, terminal device, and server device
CN103268291B (en) The method of persistence index metadata is postponed in flash-memory storage system
CN110162396A (en) Method for recovering internal storage, device, system and storage medium
CN110955488A (en) Virtualization method and system for persistent memory
CN104917788A (en) Data storage method and apparatus
CN115617264A (en) Distributed storage method and device
CN115098447A (en) File recovery method and device, electronic equipment and readable storage medium
CN110245129B (en) Distributed global data deduplication method and device
CN108958660B (en) Distributed storage system and data processing method and device thereof
KR101463536B1 (en) Memory management apparatus, method and system
JP4189342B2 (en) Storage apparatus, storage controller, and write-back cache control method

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
TR01 Transfer of patent right

Effective date of registration: 20220125

Address after: 100007 room 205-32, floor 2, building 2, No. 1 and No. 3, qinglonghutong a, Dongcheng District, Beijing

Patentee after: Tianyiyun Technology Co.,Ltd.

Address before: No.31, Financial Street, Xicheng District, Beijing, 100033

Patentee before: CHINA TELECOM Corp.,Ltd.

TR01 Transfer of patent right