CN102081509A - Reading method and device for RAID1 equipment - Google Patents
Reading method and device for RAID1 equipment Download PDFInfo
- Publication number
- CN102081509A CN102081509A CN2009102533226A CN200910253322A CN102081509A CN 102081509 A CN102081509 A CN 102081509A CN 2009102533226 A CN2009102533226 A CN 2009102533226A CN 200910253322 A CN200910253322 A CN 200910253322A CN 102081509 A CN102081509 A CN 102081509A
- Authority
- CN
- China
- Prior art keywords
- read
- disk
- raid1
- reading
- request
- 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
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a method and a device for reading RAID1 equipment, wherein the device comprises: the information statistics module is used for counting the current state information of each member disk contained in the RAID1 equipment when a reading request of the RAID1 equipment data is received; the read block division module is used for dividing the size of the read block region of the member disk according to the size of the data requested to be read; and the read request processing module is used for distributing the read request to the member disks to execute parallel data reading according to the current state information of each member disk and the size of the read block interval and informing the idle member disks to read the next interval data into the cache in advance. The invention can improve the reading efficiency of the RAID1 equipment.
Description
Technical field
The present invention relates to disk array technology, relate in particular to a kind of read method and device of RAID1 equipment.
Background technology
The RAID1 disk array is generally realized by mirror image technology.A lot of companies use RAID1 to realize not having the back-up job of interrupting, so as to the protection critical data.When using RAID1, the data that write redundant arrays of inexpensive disks (RAID, redundant arrays of inexpensive disks) equipment will write each member's disk, and writing position is identical.Like this, the operate as normal that disk failures can not influence array appears in the disk array, even only stay a disk in the array, and the read-write requests on also can normal process RAID equipment.
RAID1 is applied in usually need provide the higher data redundancy protecting, to the less demanding occasion of read-write usefulness.When RAID equipment write data, data can write each member's disk, and this has just prolonged the write request response time.Many companies write usefulness by software and hardware optimization, but effect is not satisfactory.In the time of from RAID equipment reading of data, read performance is not high yet, the usefulness that general approaching single disk reads.Though this is because of data consistent on each member's disk, be difficult to concurrent reading between control member disk, generally be exactly that single disk is in work when reading.When reading RAID1 equipment, be to read same disk basically in disk array promptly, other member's disk is in idle state.
In existing RAID1 technology, some method is fixedly to read a disk, and some method considers that mainly disk current I/O position selects to read destination disk, but the general usefulness near single disk of read performance.Fig. 1 has described the process flow diagram that typically reads algorithm now, from the step 102 of this process flow diagram as can be seen to step 106, the each new current I/O position that all can check member's disk when reading request comes is read Request Processing and is finished renewal I/O position, back.Read under the situation in order like this, most read requests all can use same target disk, and other member's disk then is in idle state.
Summary of the invention
Purpose of the present invention is intended to one of solve the aforementioned problems in the prior at least.
For this reason, embodiments of the invention propose a kind of RAID1 equipment read method and device, realize the efficiently collaborative work of a plurality of member's disks, improve the read performance of RAID1 equipment.
According to an aspect of the present invention, the embodiment of the invention has proposed a kind of reading device of RAID1 equipment, described reading device comprises: the current state information of each member's disk that Information Statistics module, described Information Statistics module are comprised at the when request statistics RAID1 equipment that reads that receives the RAID1 device data; Read piece and divide module, describedly read the size that piece is divided the data that module reads according to described request, that divides member's disk reads the interval size of piece; And the read request processing module, described read request processing module is according to the current state information of each member's disk and read the interval size of piece, will the request of describedly reading be assigned to that the execution parallel data reads on member's disk.
The further embodiment according to the present invention, the current state information of described member's disk comprise the busy extent and/or the current I/O position of member's disk.Described current I/O position is determined according to the position of a last read-write requests, perhaps determines according to the update request of metadata.
The further embodiment according to the present invention, described read piece divide module also according to the device type of member's disk and/or member's disk read in advance be provided with to determine that described to read piece interval big or small.
According to a further aspect in the invention, embodiments of the invention propose a kind of read method of RAID1 equipment, described read method may further comprise the steps: a) in when request of reading that receives the RAID1 device data, and the current state information of each member's disk that statistics RAID1 equipment is comprised; B) size of data that reads according to request is divided the interval size of reading of member's disk; And c), will the request of describedly reading be assigned to that the execution parallel data reads on member's disk according to the current state information of each member's disk and read interval size.
The further embodiment according to the present invention, the current state information of described member's disk comprise the busy extent and/or the current I/O position of member's disk.Described current I/O position is determined according to the position of a last read-write requests, perhaps determines according to the update request of metadata.
The further embodiment according to the present invention, described read interval size also further according to the device type of member's disk and/or member's disk read in advance be provided with to determine.
The further embodiment according to the present invention, described step c comprises: receiving described when request of reading, checking the busy extent of each member's disk; If when having idle member's disk, according to member's disk of a described free time of choice of location of reading request as first target disk to carry out the described data read that reads interval size; When carrying out described data read, check the busy extent of other member's disk once more; If when having idle member's disk, select another idle member's disk to carry out the described next interval data read that reads interval size, and put into buffer memory as second target disk; And receiving newly when reading request, if the position of the described new request of reading is positioned at the interval range that reads of described second target disk, then from described buffer memory, read corresponding data.
The present invention considers member's disk current I/O position, disk usefulness, and disk busy extent, user's read request continuity or the like various factors, thus control the concurrent collaborative work efficiently of a plurality of member's disks.Like this, will no longer include idle disk in the time of reading of data, especially under the big file situation of sequential read.Thereby improved the read performance of RAID1 equipment, shortened user's read request response time.
Aspect that the present invention adds and advantage part in the following description provide, and part will become obviously from the following description, or recognize by practice of the present invention.
Describe the present invention below in conjunction with the drawings and specific embodiments, but not as a limitation of the invention.
Description of drawings
Fig. 1 is the read method process flow diagram of the RAID1 equipment of prior art;
Fig. 2 is the reading device structure principle chart of the RAID1 equipment of the embodiment of the invention;
Fig. 3 is the read method process flow diagram of RAID1 equipment of the present invention;
Fig. 4 is the read method process flow diagram of the RAID1 equipment of one embodiment of the invention; And
Fig. 5 is the data read synoptic diagram of the embodiment of the invention.
Embodiment
Describe embodiments of the invention below in detail, the example of described embodiment is shown in the drawings, and wherein identical from start to finish or similar label is represented identical or similar assembly or the assembly with identical or similar functions.Below by the embodiment that is described with reference to the drawings is exemplary, only is used to explain the present invention, and can not be interpreted as limitation of the present invention.
At first, the figure illustrates the reading device structure principle chart of the RAID1 equipment of the embodiment of the invention with reference to figure 2.As shown in Figure 2, at by a plurality of member's disks, member's disk 1, member's disk 2, ... the RAID1 equipment 10 that member's disk n constitutes, the reading device 20 of the embodiment of the invention comprise Information Statistics module 22, read piece division module 26 and read request processing module 28.Information Statistics module 22 is in when request of reading that receives the RAID1 device data, the current state information of each member's disk 1～n that statistics RAID1 equipment is comprised.Read piece and divide the size of module 26 according to the corresponding data that read of request, that divides member's disk reads the interval size of piece, and 28 of read request processing modules are according to the current state information of each member's disk and read the interval size of piece, and the request of will reading is assigned to be carried out parallel data and read on member's disk.In read request processing module 28, handle current read request simultaneously, find to have idle disk in the controller, then will read next interval censored data in advance and enter buffer memory from idle disk.Like this, have higher performance under the sequential read situation.
The current state information of each member's disk of 22 statistics of Information Statistics module in one embodiment, comprises: the busy extent of member's disk and/or current I/O position.Wherein, for statistics disk busy extent, can be according to disk read-write requests sum since start of operating system record, statistics disk read-write request situation of change, and dynamically update data.According to these information that operating system provides, can add up time period busy extent till now of disk.If this time ratio is shorter, for example preceding 10 seconds, 30 seconds historical data statistics etc., this busy extent can reflect the state when front disk.
In addition, the current I/O position statistics for member's disk in the RAID1 controller, comprises two types of read-write requests: the request of request that the user issues and controller self update metadata on member's disk.Generally preserve I/O position when front disk by writing down a last user writable request position, this mode is only applicable to the request that track user issues.Except the read-write requests that the user issues, also have the metadata updates request on member's disk, to write down current I/O position when handling these requests equally.In one embodiment of the invention, can take into account all I/O requests on member's disk, accurately indicate current I/O position.
The interval size of piece that reads that reads piece division module 28 division member disks is in order to realize concurrent the reading of a plurality of member's disks, to be provided with automatically by reading piece division module 28.When being provided with automatically, read piece and divide module 28 and enumerate some interval values, test these interval values respectively, select concurrent usefulness best (as 64k, 32k).In the time of the test interval value, order reads member's disk, read the data of certain interval size from member's disk, read the data of next interval size from another piece member disk, if adopt this interval size, the next interval censored data that another piece member disk reads is hit by major part in buffer memory, and two disks to read usefulness approaching, illustrate that to adopt this interval size proper.Therefore, can be in view of the above the piece interval of reading of member's disk be established size and is changed to this fixed value.
In addition, divide member's disk read the interval size of piece the time, read piece and divide module 28 and it is also conceivable that the disk unit type, dissimilar have different read-write usefulness.In RAID1, data can parallelly write each disk in the time of processing write requests, so data block (CHUNK) size is to writing not influence of usefulness; CHUNK is the least unit that reads from disk when handling read request, and the CHUNK size is very obvious to the equipment effectiveness influence, therefore appropriate C HUNK size need be set.Generally speaking, read piece during the bulk file read-write and divide module 28 and adopt the big CHUNK size system consumption can point, divide module 28 and adopt less CHUNK more suitable and read piece during the small documents read-write.
Secondly, it is also conceivable that disk unit read in advance be provided with, the hit rate when reading in order to improve in buffer memory, General System all can shift to an earlier date reads into buffer memory to certain interval censored data.Thereby reading that calculating member disk is suitable is interval big or small.Work as member's disk like this and read simultaneously, other interval could accurately be located and read to other member's disk.
Moreover, set appropriate C HUNK size, need take all factors into consideration file system, disk type, multiple factor such as system configuration.Therefore, read piece division module 28 and can also stay interface for the user is provided with the CHUNK size.That is, reading the interval size of piece also can be according to accessing characteristic, and a parameter of regulating as usefulness is provided with or appointment for the user.
28 of read request processing modules are taken all factors into consideration above-mentioned statistical information, and the request of reading is distributed on a plurality of member's disks, bear between each member's disk of mobile equilibrium, can reach to optimize the purpose that RAID1 reads usefulness.
Specifically, when reading request comes, read request processing module 28 judges at first whether the current C HUNK that request is positioned at is read by certain member's disk in advance, if just this member's disk is distributed in request at every turn.If do not read in advance, come the select target disk according to the busy extent of member's disk:, read the data read of interval size with execution as target disk according to member's disk of free time of choice of location of the request of reading if when having idle member's disk.When the select target disk read, read request processing module 28 was checked the busy extent of other disk once more.If there is idle disk, this idle disk is set to new target disk, notifies it to read next CHUNK data.
Like this, in fact just there are two disks to read concurrent, and continuous before and after the interval position of reading.The data that new target disk reads out will be put into buffer memory, and can note the order of new target disk in array and the position of reading.Like this, when reading request comes, read request processing module 28 judges whether new request is positioned at the interval range that new target disk reads out when new.If be positioned at, these requests are sent to new target disk, because these data have been read in buffer memory, read request will be obtained data and return very soon from buffer memory.If the position of the new request of reading not the reading in the interval range of new target disk, then continue according to the busy extent of member's disk select idle member's disk as target disk to distribute this new request of reading.Read request processing module 28 is this to be read the current goal disk and seeks new target disk action simultaneously and can continue.
In one embodiment, read request processing module 28 can also be further combined with the target disk of the nearest current I in the position that idle member's disk middle distance reads request/O choice of location correspondence.That is to say that if the I/O position is the most approaching, and disk is not in busy state, select when front disk as target disk; If the I/O position is the most approaching, but disk is in busy state, select current I in other disk/O position more approaching as target disk.
Fig. 3 has provided the read method process flow diagram of RAID1 equipment of the present invention, as shown in the figure, may further comprise the steps: in when request of reading that receives the RAID1 device data, and the current state information (step 202) of each member's disk that statistics RAID1 equipment is comprised; Then, according to the size of data that request is read, divide reading of member's disk interval big or small (step 204); At last, according to the current state information of each member's disk and read interval size, the request of will reading is assigned to be carried out parallel data and reads (step 208) on member's disk.
The current state information of each member's disk of being added up in one embodiment, comprises: the busy extent of member's disk and/or current I/O position.Member's disk busy extent can be according to disk read-write requests sum since start of operating system record, statistics disk read-write request situation of change, and dynamically update data.According to these information that operating system provides, can add up time period busy extent till now of disk.
In addition, the current I/O position statistics for member's disk in the RAID1 controller, can comprise two types of read-write requests: the request of request that the user issues and controller self update metadata on member's disk.Therefore, the current I of member's disk/O position can be determined according to the position of a last read-write requests.When having the metadata updates request, current I/O position is then determined according to the update request of metadata.
Divide member's disk to read the interval size of piece be in order to realize concurrent the reading of a plurality of member's disks, can be provided with automatically by the RAID1 controller.When being provided with automatically, enumerate some interval values, test these interval values respectively, select concurrent usefulness best (as 64k, 32k).In the time of the test interval value, order reads member's disk, read the data of certain interval size from member's disk, read the data of next interval size from another piece member disk, if adopt this interval size, the next interval censored data that another piece member disk reads is hit by major part in buffer memory, and two disks to read usefulness approaching, illustrate that to adopt this interval size proper.Therefore, can be in view of the above the piece interval of reading of each member's disk be established size and is changed to this fixed value.
In addition, divide each member's disk read the interval size of piece the time, it is also conceivable that the disk unit type, dissimilar have different read-write usefulness.CHUNK is the least unit that reads from disk when handling read request, and the CHUNK size is very obvious to the equipment effectiveness influence, therefore appropriate C HUNK size need be set.Generally speaking, adopt the big CHUNK size system consumption can point during the bulk file read-write, and adopt less CHUNK more suitable during the small documents read-write.
Secondly, it is also conceivable that disk unit read in advance be provided with, the hit rate when reading in order to improve in buffer memory, General System all can shift to an earlier date reads into buffer memory to certain interval censored data.Thereby reading that calculating member disk is suitable is interval big or small.Work as member's disk like this and read simultaneously, other interval could accurately be located and read to other member's disk.
Moreover, set appropriate C HUNK size, need take all factors into consideration file system, disk type, multiple factor such as system configuration.Therefore, this method can also be included as the user step that the CHUNK size stays interface is set.That is, reading the interval size of piece also can be according to accessing characteristic, and a parameter of regulating as usefulness is provided with or appointment for the user.
Read the interval by member's disk of taking all factors into consideration above-mentioned statistical information and suitable size, the request of reading is distributed on a plurality of member's disks, bear between each member's disk of mobile equilibrium, can reach and optimize the purpose that RAID1 reads usefulness.
Specifically, when reading request comes at every turn, for the data of not read in advance, come the select target disk according to the busy extent of member's disk:, read the data read of interval size with execution as target disk according to member's disk of free time of choice of location of the request of reading if when having idle member's disk.When the select target disk reads, check the busy extent of other disk once more.If there is idle disk, this idle disk is set to new target disk, notifies it to read next CHUNK data.
Like this, two concurrent reading of disk, and continuous before and after the interval position of reading.The data that new target disk reads out will be put into buffer memory, and can note the order of new target disk in array and the position of reading.Like this, when reading request comes, at first judge whether new request is positioned at the interval range that new target disk reads out when new.If be positioned at, these requests are sent to new target disk, because these data have been read in buffer memory, read request will be obtained data and return very soon from buffer memory.If the position of the new request of reading not the reading in the interval range of new target disk, then continue according to the busy extent of member's disk select idle member's disk as target disk to distribute this new request of reading.This current goal disk that reads is sought new target disk action simultaneously and can be continued.
Current I/O position concerns to come the select target disk: if the I/O position is the most approaching, and disk is not in busy state, select when front disk as target disk; If the I/O position is the most approaching, but disk is in busy state, select current I in other disk/O position more approaching as target disk.
Fig. 4 has provided the read method process flow diagram of the RAID1 equipment of one embodiment of the invention, and in this embodiment, read method of the present invention has been considered current state informations such as the current I/O position of member's disk and busy extent simultaneously.
Specifically, RAID1 equipment at first receives new read request to read data (step 302) wherein, can judge afterwards whether the data of read request are read (step 304) in advance by certain member's disk, reading in advance here is metadata cache read operation at there being read request before, when asking first, this step then can be omitted in the read request that receives.
If do not read in advance, then come the select target disk to handle read request (step 306): if the I/O position is the most approaching according to the position of read request and member's disk current I/O position relation and the current busy extent of member's disk, and disk is not in busy state, selects to work as front disk as target disk; If the I/O position is the most approaching, but disk is in busy state, select current I in other disk/O position more approaching as target disk.
Then, when the select target disk reads in step 306, check the busy extent of other member's disk once more.If there is idle disk, this idle disk is set to new target disk, and reads the next interval censored data (step 308) adjacent with this request from this free time member's disk, and the data that read out will be put into buffer memory.
At last, the direct reading of data of read request correspondence or cache read are fetched data return (step 310).Repeat above-mentioned steps, promptly can constantly handle each RAID1 data read request.
Fig. 5 has provided and has utilized the present invention to the example that the RAID1 that is made up of two member's disks carries out data read, has wherein stored identical data A respectively on member's disk 30,40, B, C, D.Suppose that corresponding one of each data read the interval, these four data are stored continuously, when the user needs reading of data A, at first select member's disk 30 as target disk reading of data A, read the state of checking member's disk 40 simultaneously.
If equipment is in idle state, then select member's disk 40 as new target disk, notify its reading of data B, when the user needs reading of data B, as target disk,, can directly return member's disk 40 like this to the user because data are read.This action can continue until finishing corresponding read request.
Because RAID1 adopts mirror image technology, the data consistent of preserving on member's disk, therefore the present invention takes all factors into consideration many-sided factor, member's disk current I/O position for example, disk usefulness, the disk busy extent, user's read request continuity or the like, thus control the concurrent collaborative work efficiently of a plurality of member's disks.Like this, will no longer include idle disk in the time of reading of data, especially under the big file situation of sequential read.Thereby improved the read performance of RAID1 equipment, shortened user's read request response time.
Certainly; the present invention also can have other various embodiments; under the situation that does not deviate from spirit of the present invention and essence thereof; those of ordinary skill in the art work as can make various corresponding changes and distortion according to the present invention, but these corresponding changes and distortion all should belong to the protection domain of the appended claim of the present invention.
Claims (10)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009102533226A CN102081509A (en) | 2009-11-30 | 2009-11-30 | Reading method and device for RAID1 equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009102533226A CN102081509A (en) | 2009-11-30 | 2009-11-30 | Reading method and device for RAID1 equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102081509A true CN102081509A (en) | 2011-06-01 |
Family
ID=44087493
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2009102533226A Pending CN102081509A (en) | 2009-11-30 | 2009-11-30 | Reading method and device for RAID1 equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102081509A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102207831A (en) * | 2011-07-04 | 2011-10-05 | 成都市华为赛门铁克科技有限公司 | Data reading-writing method and device of magnetic disk array |
| CN102426553A (en) * | 2011-11-11 | 2012-04-25 | 中国科学技术大学 | Method and device for transmitting data to user based on double-cache pre-reading |
| CN103970484A (en) * | 2013-02-06 | 2014-08-06 | 华为技术有限公司 | Method, device and system for controlling data storage |
-
2009
- 2009-11-30 CN CN2009102533226A patent/CN102081509A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102207831A (en) * | 2011-07-04 | 2011-10-05 | 成都市华为赛门铁克科技有限公司 | Data reading-writing method and device of magnetic disk array |
| CN102426553A (en) * | 2011-11-11 | 2012-04-25 | 中国科学技术大学 | Method and device for transmitting data to user based on double-cache pre-reading |
| CN102426553B (en) * | 2011-11-11 | 2014-05-28 | 中国科学技术大学 | Method and device for transmitting data to user based on double-cache pre-reading |
| CN103970484A (en) * | 2013-02-06 | 2014-08-06 | 华为技术有限公司 | Method, device and system for controlling data storage |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP4694333B2 (en) | Computer system, storage device, system management device, and disk device power control method | |
| US8392648B2 (en) | Storage system having a plurality of flash packages | |
| US7539816B2 (en) | Disk control device, disk control method | |
| JP5958020B2 (en) | Storage system | |
| US7321955B2 (en) | Control device, control method and storage medium recording a control program for controlling write-back schedule of data from cache memory to a plurality of storage devices | |
| US10649667B2 (en) | Mitigating GC effect in a RAID configuration | |
| EP1193600A2 (en) | Memory control apparatus and its control method | |
| US20100070656A1 (en) | System and method for enhanced load balancing in a storage system | |
| US20090210611A1 (en) | Storage system and data write method | |
| EP2927779B1 (en) | Disk writing method for disk arrays and disk writing device for disk arrays | |
| US20080140908A1 (en) | Storage system, and method and program for selecting memory region | |
| US8769195B2 (en) | Storage apparatus and storage apparatus control method | |
| US10802731B1 (en) | Power saving mechanisms for a dynamic mirror service policy | |
| CN101174198B (en) | Data storage system and data access method thereof | |
| US8769199B2 (en) | Methods and systems of distributing RAID IO load across multiple processors | |
| US10152242B1 (en) | Host based hints | |
| JP2002157091A (en) | Storage subsystem and storage device used in the system | |
| CN102081509A (en) | Reading method and device for RAID1 equipment | |
| JP2021135538A (en) | Storage controller and storage control program | |
| CN116583827A (en) | Data access method, device, disk controller, disk and data storage system | |
| US20150067285A1 (en) | Storage control apparatus, control method, and computer-readable storage medium | |
| US8364890B2 (en) | RAID control apparatus and control method therefor | |
| KR101135313B1 (en) | Nand flash memory apparatus and method for scheduling requested operations | |
| US10803893B1 (en) | Data transfer scheduling for fairness and balance | |
| CN101840313B (en) | LUN mirror image processing method and equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20110601 |