CN1316377C - Write handling method for disc array arrangement - Google Patents
Write handling method for disc array arrangement Download PDFInfo
- Publication number
- CN1316377C CN1316377C CNB2004100371347A CN200410037134A CN1316377C CN 1316377 C CN1316377 C CN 1316377C CN B2004100371347 A CNB2004100371347 A CN B2004100371347A CN 200410037134 A CN200410037134 A CN 200410037134A CN 1316377 C CN1316377 C CN 1316377C
- Authority
- CN
- China
- Prior art keywords
- write
- cache
- itemize
- data
- 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.)
- Expired - Lifetime
Links
- 238000000034 method Methods 0.000 title claims description 17
- 230000008569 process Effects 0.000 claims description 9
- 241000748029 Chamaechaenactis scaposa Species 0.000 abstract 1
- 230000007423 decrease Effects 0.000 abstract 1
- 238000003672 processing method Methods 0.000 abstract 1
- 238000005516 engineering process Methods 0.000 description 5
- SBNFWQZLDJGRLK-UHFFFAOYSA-N phenothrin Chemical compound CC1(C)C(C=C(C)C)C1C(=O)OCC1=CC=CC(OC=2C=CC=CC=2)=C1 SBNFWQZLDJGRLK-UHFFFAOYSA-N 0.000 description 5
- 101100419716 Podospora anserina RPS9 gene Proteins 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 239000012141 concentrate Substances 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 230000008929 regeneration Effects 0.000 description 2
- 238000011069 regeneration method Methods 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004883 computer application Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000000875 corresponding effect Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
Images
Landscapes
- Memory System Of A Hierarchy Structure (AREA)
Abstract
The present invention relates to a writing processing method for a disc array structure. When writing requests of a host come, and if a cache is not full, data is directly written in the cache. Simultaneously, advance rewriting is carried out, and items in the cache on which the writing requests are placed and a plurality of subsequent items are filled in advance. When the cache is full, the writing requests of a host wait, and a plurality of areas are selected from areas where advance prewriting is completed for carrying out rewriting. After a area of which the rewriting is completed, a writing request is taken out from a writing request waiting queue, and the host writing is renewedly attempted. The present invention reduces hollow spaces so as to enhance the possibility of combination. The present invention decreases the quantity of writing requests, and further improves the rewriting efficiency of the cache in a memory system. The hollow spaces do not exist on the items, namely that the items are full stems, the difficulty degree of error processing can also be simplified, and the check data generation speed is quickened.
Description
Technical field
The present invention relates to the memory technology of data, relate in particular to the disposal route of writing in a kind of disc array structure.
Background technology
Along with constantly widening of computer application field, the user is increasing for the capacity requirement of storage system, and more and more higher to its reliability, availability requirement, rate request is also more and more faster.The processing speed of microprocessor improves very fast, but as the disk of main external memory since mechanical motion limit its speed improvement little (access time rests on Millisecond), so just limited the overall performance of storage system greatly.The disc array structure (RAID) of band Cache (buffer memory) that solves this bottleneck is efficient fast with it, high capacity, high reliability and flexible, inexpensive advantage are occupied an leading position in the high-performance computer storage.
Disk array technology is formed a disk array with common hard disk, and when main frame write data, the data decomposition that the RAID controller will write main frame was a plurality of data blocks, is written in parallel to disk array then; During the main frame reading of data, parallel the reading of RAID controller is dispersed in the data on each hard disk in the disk array, offers main frame after they are reconfigured.Owing to adopt the concurrent reading and concurrent writing operation, thereby improved the access degree of storage system.In addition, the RAID disk array can also adopt measures such as mirror image, parity checking, improves system survivability, guarantees the reliability of data.
Write-back is the process that write data among the Cache is write disk, write-back time=head positioning time (seek time and rotational latency)+data transmission period, be very long (millisecond rank) head positioning time comparatively speaking, and data transmission period is relevant with the size of write data, therefore will improve write-back efficient, mainly be to reduce head positioning time.In the disc array structure of band Cache, generally adopt quick writing technology (Fast Write) at present: when write request arrives, data block is write into promptly to make sound behind the Cache and should be finished signal, actual write back operations is asynchronous realization on the backstage, has reduced the write request response time.
As shown in Figure 1, a kind of disposal route of writing has been proposed: when the main frame write request arrives among the patent US6327673 " STORAGE UNIT SUBSYSTEM " of HDS (data system company of Hitachi), judge that new write data need generate the information of new checking data whether in Cache, if do not exist entirely, then read in (comprising legacy data, old checking data) from disk, and generate checking data and write Cache, Cache adopts certain write-back strategy that data are write disk seclected time.Wherein, the main frame write request is just returned after writing Cache, writes to the main frame report and finishes, and above checking data correlated process and main frame read-write process are independent.This patent has improved write performance, for the fast literary sketch algorithm of RAID5 of routine, issuing write request to before the disk, in advance checking data is generated, and is kept among the Caehe, has reduced the time delay of write-back.In addition, because during each write request, it is full all the magnetic track at write request place to be looked ahead, and has improved the hit rate of read-write requests in Cache, simultaneously, the also quantity of completely writing when having increased write-back has greatly improved the performance of write-back.Each step in the corresponding diagram is described as follows:
(1) if write data in Cache recklessly in, be new write data need generate new checking data information not entirely in Cache, then the data from target data block to its place magnetic track ending are read the Old Data (legacy data) of Cache from data disks (Data Disk);
(2) new data that need write is set up among the Write Data (write data) that writes Cache in the process in Old Data data;
Need check whether data hit when (3) generating the check information of request msg, if do not hit, old checking data is read among the Old Parity (old checking data) of Cache;
(4) all ready if produce all required data of new checking data, these data be sent to disk adapter (disk adapter, DKA) data recover and regeneration (DataRecovery and Regeneration is DRR) in the circuit;
(5) finish when new checking data produces, DRR is delivered to it among the New Parity (new checking data) of Cache A and CacheB (New Parity is the same with the processing mode of the new data that main frame will be write).
Though technique scheme has improved write-back efficient to a certain extent, but because it is " data from target data block to its place magnetic track ending are read the Old Data ", if stripe unit of the less than of the new data that writes so own, according to this mode of looking ahead, the itemize at new data place will form one " cavity ", be both not have " dirty data " among the Cache (dirty data refers among the Cache and the inconsistent data of disk, be the data that main frame writes Cache, but these data are not synchronized to disk as yet), (legacy data refers to data consistent with disk among the Cache also not have " legacy data ", the Cache data sync that promptly main frame has been write is to the data of disk, or read the data of Cache from disk), that is to say, certain data block was never write by main frame in this itemize, or never covered by data in magnetic disk, promptly there be " cavity ".
Because it must be continuous being handed down to a request of disk, such as a zone writing disk physical address from 10 to 100.If write 10～50,60～100 two zones so, just can only issue 10～50 and 60～100 two requests respectively, and can't issue, because 50～60 these interval promptly so-called cavities do not allow to write according to a request.
As shown in Figure 2, because prior art is carried out pre-write-back according to magnetic track, look ahead is to its place magnetic track ending from target data block, thereby make the SU00 of itemize 0, the SU12 of itemize 1, the SU20 of itemize 2, more or less there are some cavities in the SU32 of itemize 3, A is the cavity among the Cache among the figure, B is the dirty data among the Cache, C is an existing legacy data among the Cache, D is the legacy data of looking ahead and obtaining among the Cache, E is the end position of magnetic track 1, F is end position (the supposition magnetic track 1 of magnetic track 2,2 addresses are continuous, so magnetic track 1 from the SU00 left margin to arrow E indication position, magnetic track 2 from magnetic track 1 end position to next arrow F indication position).
Because being handed down to the request of disk must be continuously, there be " cavity " in stripe unit SU20 in Cache, therefore, SU00, SU10, SU20, SU30 can't be merged into a write request during write-back, and have to split two write requests (SU00～SU10, SU20～SU30).In like manner, SU02, SU12, SU22, SU32 also have to split into two requests.
Because the Cache total volume is limited,,,, be difficult to the space that utilizes Cache limited better though simplified the write-back method because each write request all will read the data from target data block to its place magnetic track ending according to this write-back method.Like this, when the Cache write-back, exist data in the itemize in " cavity " will have to split into a plurality of write requests and issue, thereby increased the number of times of write request, reduced the efficient of write-back.
Summary of the invention
Technical matters to be solved by this invention is: overcome existing writing in the processing procedure, data in the itemize in existence cavity are had to split into a plurality of write requests and are issued, thereby increased the number of times of write request, reduced the deficiency of write-back efficient, the disposal route of writing in a kind of disc array structure is provided, reduce the number of times of write request, improve the write-back efficient of Cache in the storage system.
The present invention solves the problems of the technologies described above the technical scheme that is adopted to be:
Write disposal route in this disc array structure, the write-back process that data write disk among the Cache may further comprise the steps:
A, pre-write-back, described pre-write-back is base unit with the itemize, whether hit the situation of Cache according to write request, several itemizes that the itemize at write request place among the Cache is later or the itemize that comprises the write request place are looked ahead full, promptly the legacy data of these itemize address scopes is read in Cache, replace the legacy data of reading with the new data in these itemizes simultaneously from disk;
B, write phase, several zones of choosing from the zone of finishing described pre-write-back wherein write disk.
Further whether Cache is expired and judges, may further comprise the steps:
When A, main frame write request arrive, if Cache less than, directly data are write Cache, carry out pre-write-back simultaneously;
B, when Cache is full, the main frame write request is waited for, chooses the write phases in several zone beginning write-backs wherein from the zone of finishing pre-write-back;
C, after certain regional write-back is finished, from main frame write request waiting list, take out one, return steps A.
In the described steps A, after the main frame write request writes Cache, return main frame and write success, judge whether write request hits Cache, promptly whether have a blocks of data among the Cache with the identical starting and ending of current main frame write request address; If hit, judge again whether several itemizes after this write request place itemize have carried out pre-write-back, if there is not pre-write-back, then issue pre-write-back request, several itemizes after this write request place itemize are carried out pre-write-back; If miss, then issue pre-write-back request, with the itemize at write request place with and subsequent several itemizes carry out pre-write-back; Itemize quantity after the itemize of described write request place is read size according to the single maximum of disk tolerance, and the Cache remaining space is dynamically adjusted.If the Cache remaining space is less than given threshold value, then the itemize quantity after the itemize of write request place is taken as 0.
Among the described step B, when Cache is full, write request is inserted waiting list, judge whether to exist the zone of having finished pre-write-back then, if there is no, wait for that then the pre-write-back in certain zone finishes; If exist, then selection is wherein carried out write-back in several zones, and the value of write-back region quantity is capitalized size most according to the single of disk tolerance, and the write-back area size of configuration is dynamically adjusted.Described Cache is full of the valid data size that refers to Cache and surpasses the predefined size of user.
Beneficial effect of the present invention is: the present invention is divided into two stages with the write-back in write-back zone: pre-write back stage and write phase.By pre-write-back, the existing cavity of part that is about to write-back among the Cache is dropped to minimum, to improve the efficient in second stage of write-back.In addition, because write-back is divided into two stages independently, the previous stage is absorbed in and reads, stage of back is absorbed in and writes, and the situation of reading (because conventional write-back algorithm is just to begin to generate checking data when write-back) can not appear having in the sort of the writing of conventional write-back algorithm, increased the possibility that merges between request.The present invention is in Cache, write-back is carried out in zone of each selection, each write-back is when (containing pre-write-back), all fix and select 1 zone to N continuation address, and guarantee to have only a zone at write-back in a RAID group in the same time, with guarantee this section in the period the mobile of each magnetic disc head be fixed in the zone, effectively reduced head positioning time in the write-back, thereby greatly improved disk read-write efficient.
Prior art is because the least unit of main frame write request is uncertain, causes that the dirty data in the write-back itemize may be extremely discontinuous among the Cache, and RAID will issue other disk read-write request of BLOCK level for the write-back of this extremely discontinuous itemize, cause inefficiency.(BLOCK is a data block and the present invention helps RAID to mask the notion of BLOCK in Cache, it is the minimum data unit of disk operating, generally be 512Byte), the least unit that Cache offers RAID is an itemize, guaranteed read-write requests that RAID sends to IO all be the stripe unit level other, and the amount of looking ahead of the present invention is dynamically to adjust, be that the M value dynamically generates according to factors such as current C ache remaining spaces, rather than it is changeless, utilize the Cache space better by the amount of looking ahead that dynamic adjustment is each, thereby improved disk read-write efficient.
The present invention has reduced the cavity, to improve the possibility that merges, has reduced write request quantity, has further improved the write-back efficient of Cache in the storage system.And do not had the cavity on the itemize, promptly the full marks bar also can be simplified the fault processing difficulty, improves checking data and produces speed.
Description of drawings
Fig. 1 is a kind of treatment scheme synoptic diagram of writing of prior art;
The structural representation that has the cavity when Fig. 2 presses the pre-write-back of magnetic track for prior art;
There is not the structural representation in cavity during by the pre-write-back of itemize in Fig. 3 for the present invention;
Fig. 4 writes processing flow chart for the present invention.
Embodiment
With embodiment the present invention is described in further detail with reference to the accompanying drawings below:
In RAID, will be to the write request of disk array, write the stripe unit quantity X of dish according to needs in the itemize, (the stripe unit quantity Y/2 of X＞=this itemize of composition supposes itemize size=Y to be divided into capitalization
*The stripe unit size), small letter (the stripe unit quantity Y/2 of X＜this itemize of composition) and completely writing.In above-mentioned three kinds are write, most effective completely to write.Fundamental purpose of the present invention is how to improve to write dish efficient.Non-(upper case and lower case) step of completely writing of itemize comprises the sub-read request that issues legacy data earlier, and wait is returned, and does the sub-write request that issues new data behind the XOR again.If with pronouncing of all itemizes in the working set is to concentrate by disk in a stage to be handed down to the disk handling procedure, write also to concentrate and issue by disk as another stage, the disk handling procedure just can effectively merge the request of this disk and dispatch, and then improves write-back efficient.
Because that the prerequisite that two requests merge is a request type is identical (or all read, or all write), and the address is continuous between request.The present invention is divided into two stages with the write-back in write-back zone: read phase (being called pre-write back stage) and write phase.By pre-write-back, the existing cavity of part that is about to write-back among the Cache is dropped to minimum, to improve the efficient in second stage of write-back.In addition, because write-back is divided into two stages independently, the previous stage is absorbed in and reads, stage of back is absorbed in and writes, and the situation of reading (conventional write-back algorithm is just to begin to generate checking data when write-back) can not appear having in the sort of the writing of conventional write-back algorithm, like this, because the write-back flow process will be read and write independent processing respectively, make in certain time, otherwise read request is concentrated relatively, otherwise write request is concentrated relatively, so disk driver or RAID program, just discrete originally independently request may be merged into one, increase the possibility that merges between request.
During the pre-write-back of the present invention, be base unit, rather than to adopt of the prior art with the magnetic track method of base unit with the itemize, thereby guarantee among the Cache, so long as will there be any " cavity " in the itemize that main frame was once write, the quantity of completely writing when increasing the batch write-back as much as possible.As shown in Figure 3, the present invention like this, come what may, as long as the itemize that carried out looking ahead must be the full marks bar, and any cavity can not occur owing to look ahead according to itemize.Because there be not any " cavity " in all stripe units in the itemize 0～3, write so four itemizes can be merged into 3 bulks during write-back, issue simultaneously, thereby can reduce the quantity of write request, improve write performance.
Write request treatment scheme of the present invention is introduced several notions earlier as shown in Figure 4:
Cache is full: valid data (dirty data and the legacy data) size that refers to Cache surpasses the size that the user sets, and generally is about 80% of the total size of Cache.
Dirty data: refer among the Cache and the inconsistent data of disk that promptly the user writes Cache, but is not written back to the data of disk as yet.
Legacy data: refer to data consistent with disk among the Cache, the data of promptly the user being write Cache are written back to the data of disk.
Main frame can be thought certain client terminal, such as a computing machine that Win98 is installed; Cache is a memory headroom of storage system; Disk be storage system with the RAID disk array.
Pre-write-back: i.e. read phase, be used to finish the itemize at write request place and M itemize after this itemize are looked ahead " expiring ", make these itemizes become the full marks bar, so that there be not any " cavity " in these itemizes in Cache.Pre-write-back is one and reads the dish process, promptly from disk array the legacy data of the address realm at itemize place all read in Cache, simultaneously the legacy data of reading with the replacement of the new data in this itemize.By issuing prefetch request, one or more itemizes are become the full marks bar, it is full promptly to look ahead.To RAID5, after looking ahead completely, also need the checking data of these itemizes is calculated, so just finished the pre-write-back of this itemize.The value of M is to read factors such as size, current C ache remaining space according to the single maximum of disk tolerance dynamically to generate, rather than it is changeless, if remaining space is less than certain given threshold value, such as 5%, then M can be taken as 0, and itemize was all in same zone under the M maximal value must satisfy pre-write-back.Have from 10 such as the user and to write 20 request, the itemize start address at this request place is 5, the end address is 25, then 5～25 data are read from disk, synthetic with user's 10～20 new datas again, form 5～10,10～20,20～25 such blocks of data are promptly abandoned 10～20 partial data of reading from disk, do user data and change.Like this, the user to ask the itemize at place be prefetched full.
As shown in Figure 4, when the main frame write request arrives:
1, whether judges Cache " expiring "
If " do not expire ", then the data that main frame need be write are directly write among the Cache, return main frame simultaneously and write success;
If " expire ", then allow host requests " wait ", as inserting in the main frame write request waiting list.
Have only when the Cache space less than, the write request that comes from main frame just may write Cache.
Otherwise the main frame write request can only wait for that Cache discharges segment space by write-back.
2, Cache " is expired " this situation, to insert waiting list to the write request of disk, (zone is made up of an integer itemize to judge whether to exist the zone of finishing pre-write-back then, such as T, just there is T * N itemize in N zone so, and area size depends on factors such as RAID type and segment depth).
If " do not have ", wait for that then certain pre-write-back in zone finishes;
If " have ", then select wherein N zone to carry out write-back, the value of N is decided by that the single of disk tolerance capitalizes size most, and the write-back area size (greater than 1M) of configuration, also is dynamically to adjust.
After the Cache utilization of space reaches some, then select one or more zones to carry out second stage, i.e. write phase in the write-back.It is all depleted that Cache completely is not equal to the Cache space, when the ratio of Cache remaining space and gross space during less than set-point P (such as 5%), promptly begins write-back.If carry out write-back all the time, the effect of Cache just not embodies, and writes the time that Cache spends to be far smaller than the time of writing disk certainly.The write-back strategy need be determined the N value, and for the RAID group of prior certain the RAID type created of a user, area size is fixed, if RAID group size is X, area size is Y, and then there is the zone of X/Y continuation address in this RAID group so.The value of N is capitalized size most according to the single of disk tolerance, and the write-back area size of configuration is chosen.
3, when certain regional write-back is finished, then from main frame write request waiting list, take out one, attempt finishing main frame again and write.In fact because after the write-back, Cache must vacate a part of space, write request generally can have been write direct Cache, unless the required data that write of this main frame write request are greater than the space (≤area size) vacateed of Cache, but the main frame write request is generally below 64K, so be difficult to occur this situation.
4, to Cache " less than " this situation, then the main frame write request is write Cache, after certain write request write Cache, whether " hitting " Cache that then whether judges write request promptly existed the blocks of data with the identical starting and ending of current main frame write request address among the Cache.
If " hit ", the data that show write request place itemize all in Cache, need not to issue the prefetch request at this itemize again.So, judge M itemize after this write request place itemize whether to have carried out pre-write-back (value of M is decided by that the single maximum that disk is tolerated reads size again, and Cache remaining space, if remaining space is less than certain given threshold value, such as 5%, then M can be taken as 0, and itemize was all in same zone under the M maximal value must satisfy pre-write-back), if there is not pre-write-back, then issue pre-write back command.
If " miss " then issues pre-write-back request, with the itemize at write request place with and subsequent the individual itemize of M (M 〉=0) look ahead full.
5, after the pre-write-back in certain zone is finished, judge again whether a Cache " expires " (because new main frame write request is still arriving in the pre-write-back process, and also can take a part of space of Cache owing to look ahead).If " expire ", then select regional write-back.Certainly owing to the pre-write-back in certain zone is finished, thereby the zone that exists pre-write-back to finish certainly this moment,
The present invention is divided into two stages with the write-back in write-back zone: read phase (being pre-write back stage) and write phase.In Cache, write-back is carried out in zone of each selection, each write-back is when (containing pre-write-back), all fix and select 1 zone to N continuation address, and guarantee to have only a zone at write-back in a RAID group in the same time, with guarantee this section in the period the mobile of each magnetic disc head be fixed in the zone, effectively reduced head positioning time in the write-back, thereby greatly improved disk read-write efficient.The present invention helps RAID to mask the notion of BLOCK in Cache, the least unit that Cache offers RAID is an itemize, guaranteed read-write requests that RAID sends to IO all be the stripe unit level other, and the amount of looking ahead of the present invention is dynamically to adjust, be that the M value dynamically generates according to factors such as current C ache remaining spaces, rather than changeless, utilize the Cache space better by the amount of looking ahead that dynamic adjustment is each, thereby improved disk read-write efficient.
The present invention has reduced the cavity, to improve the possibility that merges, has reduced write request quantity, has further improved the write-back efficient of Cache in the storage system.And do not had the cavity on the itemize, promptly the full marks bar also can be simplified the fault processing difficulty, improves checking data and produces speed.
Claims (8)
1, write disposal route in a kind of disc array structure, it is characterized in that, the write-back process that data write disk among the Cache may further comprise the steps:
A, pre-write-back, described pre-write-back is base unit with the itemize, whether hit the situation of Cache according to write request, several itemizes that the itemize at write request place among the Cache is later or the itemize that comprises the write request place are looked ahead full, promptly the legacy data of these itemize address scopes is read in Cache, replace the legacy data of reading with the new data in these itemizes simultaneously from disk;
B, write phase, several zones of choosing from the zone of finishing described pre-write-back wherein write disk.
2, write disposal route in the disc array structure according to claim 1, it is characterized in that, further judge, may further comprise the steps whether Cache is full:
When A, main frame write request arrive, if Cache less than, directly data are write Cache, carry out pre-write-back simultaneously;
B, when Cache is full, the main frame write request is waited for, chooses the write phases in several zone beginning write-backs wherein from the pressure territory of finishing pre-write-back;
C, after certain regional write-back is finished, from main frame write request waiting list, take out one, return steps A.
3, write disposal route in the disc array structure according to claim 2, it is characterized in that: in the described steps A, after the main frame write request writes Cache, return main frame and write success, judge whether write request hits Cache, promptly whether have a blocks of data among the Cache with the identical starting and ending of current main frame write request address;
If hit, judge again whether several itemizes after this write request place itemize have carried out pre-write-back, if there is not pre-write-back, then issue pre-write-back request, several itemizes after this write request place itemize are carried out pre-write-back;
If miss, then issue pre-write-back request, with the itemize at write request place with and subsequent several itemizes carry out pre-write-back.
4, write disposal route in the disc array structure according to claim 3, it is characterized in that: the itemize quantity after the itemize of described write request place is read size according to the single maximum of disk tolerance, and the Cache remaining space is dynamically adjusted.
5, write disposal route in the disc array structure according to claim 4, it is characterized in that: if the Cache remaining space less than given threshold value, then the itemize quantity after the itemize of write request place is taken as 0.
6, write disposal route in the disc array structure according to claim 2, it is characterized in that: among the described step B, when Cache is full, write request is inserted waiting list, judge whether to exist the zone of finishing pre-write-back then, if there is no, wait for that then certain pre-write-back in zone finishes; If exist, then selection is wherein carried out write-back in several zones.
7, write disposal route in the disc array structure according to claim 6, it is characterized in that: the value of described write-back region quantity is capitalized size most according to the single of disk tolerance, and the write-back area size of configuration is dynamically adjusted.
8, write disposal route in the disc array structure according to claim 6, it is characterized in that: described Cache is full of the valid data size that refers to Cache and surpasses the predefined size of user.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2004100371347A CN1316377C (en) | 2004-06-03 | 2004-06-03 | Write handling method for disc array arrangement |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2004100371347A CN1316377C (en) | 2004-06-03 | 2004-06-03 | Write handling method for disc array arrangement |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1704910A CN1704910A (en) | 2005-12-07 |
| CN1316377C true CN1316377C (en) | 2007-05-16 |
Family
ID=35577218
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2004100371347A Expired - Lifetime CN1316377C (en) | 2004-06-03 | 2004-06-03 | Write handling method for disc array arrangement |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN1316377C (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015081742A1 (en) * | 2013-12-04 | 2015-06-11 | 华为技术有限公司 | Data writing method and device |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100389383C (en) * | 2006-06-02 | 2008-05-21 | 杭州华三通信技术有限公司 | Method for solving writing cavity in magnetic disk redundancy array |
| CN101655774B (en) * | 2009-09-01 | 2012-08-29 | 成都市华为赛门铁克科技有限公司 | Magnetic disc control method and system as well as relevant apparatus |
| CN102779017B (en) * | 2012-06-29 | 2016-04-13 | 华中科技大学 | The control method of data buffer area in a kind of solid-state disk |
| CN104169864B (en) * | 2014-03-13 | 2017-03-29 | 华为技术有限公司 | Redirect wiring method and device |
| WO2015165055A1 (en) * | 2014-04-30 | 2015-11-05 | 华为技术有限公司 | Data storage method, memory controller and central processing unit |
| CN104133642B (en) * | 2014-07-29 | 2018-07-13 | 浙江宇视科技有限公司 | SSD Cache fill methods and device |
| CN106293500A (en) * | 2015-06-23 | 2017-01-04 | 中兴通讯股份有限公司 | A kind of write operation control method, Apparatus and system |
| CN105260128B (en) * | 2015-09-11 | 2018-05-29 | 华为技术有限公司 | A kind of method and storage device for writing data into storage device |
| CN105630413B (en) * | 2015-12-23 | 2018-09-28 | 中国科学院深圳先进技术研究院 | A kind of synchronization write-back method of data in magnetic disk |
| CN106557437A (en) * | 2016-11-22 | 2017-04-05 | 上海联影医疗科技有限公司 | A kind of high speed storing method and system of raw data |
| CN111858399B (en) * | 2019-04-30 | 2023-11-24 | 贵州白山云科技股份有限公司 | Data transmission method, device, computer readable storage medium and computer equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5860091A (en) * | 1996-06-28 | 1999-01-12 | Symbios, Inc. | Method and apparatus for efficient management of non-aligned I/O write request in high bandwidth raid applications |
| US6629211B2 (en) * | 2001-04-20 | 2003-09-30 | International Business Machines Corporation | Method and system for improving raid controller performance through adaptive write back/write through caching |
| CN1499486A (en) * | 2002-11-07 | 2004-05-26 | 华为技术有限公司 | Method for reconstructing array of magnetic discs |
-
2004
- 2004-06-03 CN CNB2004100371347A patent/CN1316377C/en not_active Expired - Lifetime
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5860091A (en) * | 1996-06-28 | 1999-01-12 | Symbios, Inc. | Method and apparatus for efficient management of non-aligned I/O write request in high bandwidth raid applications |
| US6629211B2 (en) * | 2001-04-20 | 2003-09-30 | International Business Machines Corporation | Method and system for improving raid controller performance through adaptive write back/write through caching |
| CN1499486A (en) * | 2002-11-07 | 2004-05-26 | 华为技术有限公司 | Method for reconstructing array of magnetic discs |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015081742A1 (en) * | 2013-12-04 | 2015-06-11 | 华为技术有限公司 | Data writing method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN1704910A (en) | 2005-12-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1316377C (en) | Write handling method for disc array arrangement | |
| US5515500A (en) | Recording apparatus and method for an arrayed recording apparatus | |
| KR100373313B1 (en) | A method and system for managing a cache memory | |
| EP0483525B1 (en) | Workstation power management | |
| CN101566924B (en) | Data read-write method of disk array | |
| US6564285B1 (en) | Synchronous interface for a nonvolatile memory | |
| US7127549B2 (en) | Disk acceleration using first and second storage devices | |
| JPH08212018A (en) | Method and apparatus for dynamic extension of raid | |
| US5875455A (en) | Information recording and reproducing apparatus merging sequential recording requests into a single recording request, and method of data caching for such apparatus | |
| CN101976178B (en) | Method for constructing vertically-arranged and centrally-inspected energy-saving disk arrays | |
| JP3065481B2 (en) | Disk array device and data storage method | |
| CN101604226A (en) | A kind of method that makes up raising performance of storage system in dynamic buffering pond based on virtual RAID | |
| CN100426261C (en) | High-speed memory pre-read method and device | |
| CN101135952B (en) | Using an access log for disk drive transactions | |
| US6490651B1 (en) | Host-based virtual disk drive for improving the performance of a hard disk drive's input/output | |
| CN100428193C (en) | Data preacquring method for use in data storage system | |
| CN102087586A (en) | Data processing method and device | |
| JP2003076498A (en) | Disk storage device | |
| CN100470507C (en) | Method for rewriting in magnetic disc array structure | |
| JPH07225715A (en) | Method and system for accessing data in cache | |
| US8145839B2 (en) | Raid—5 controller and accessing method with data stream distribution and aggregation operations based on the primitive data access block of storage devices | |
| US5359611A (en) | Method and apparatus for reducing partial write latency in redundant disk arrays | |
| US6332196B1 (en) | Disk storage apparatus and power supply control method for the same | |
| JPH11224164A (en) | Magnetic disk sub system | |
| CN2603974Y (en) | Magnetic disc cache apparatus |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CX01 | Expiry of patent term |
Granted publication date: 20070516 |