CN105824579B - A method of incidentally read-write reduction watt record writes amplification expense - Google Patents
A method of incidentally read-write reduction watt record writes amplification expense Download PDFInfo
- Publication number
- CN105824579B CN105824579B CN201610153718.3A CN201610153718A CN105824579B CN 105824579 B CN105824579 B CN 105824579B CN 201610153718 A CN201610153718 A CN 201610153718A CN 105824579 B CN105824579 B CN 105824579B
- Authority
- CN
- China
- Prior art keywords
- write
- read
- data block
- incidentally
- disc
- 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 - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 30
- 230000003321 amplification Effects 0.000 title claims abstract description 22
- 238000003199 nucleic acid amplification method Methods 0.000 title claims abstract description 22
- 230000009467 reduction Effects 0.000 title description 4
- 238000010586 diagram Methods 0.000 description 5
- 230000007246 mechanism Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000002776 aggregation Effects 0.000 description 3
- 238000004220 aggregation Methods 0.000 description 3
- 102000006463 Talin Human genes 0.000 description 2
- 108010083809 Talin Proteins 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 239000002699 waste material Substances 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000013341 scale-up Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/0601—Interfaces specially adapted for storage systems
- G06F3/0602—Interfaces specially adapted for storage systems specifically adapted to achieve a particular effect
- G06F3/061—Improving I/O performance
- G06F3/0611—Improving I/O performance in relation to response time
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/0601—Interfaces specially adapted for storage systems
- G06F3/0628—Interfaces specially adapted for storage systems making use of a particular technique
- G06F3/0638—Organizing or formatting or addressing of data
- G06F3/064—Management of blocks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/0601—Interfaces specially adapted for storage systems
- G06F3/0668—Interfaces specially adapted for storage systems adopting a particular infrastructure
- G06F3/0671—In-line storage system
- G06F3/0673—Single storage device
- G06F3/0674—Disk device
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Signal Processing For Digital Recording And Reproducing (AREA)
- Digital Magnetic Recording (AREA)
Abstract
The invention discloses a kind of incidentally read-writes to reduce the method that watt record writes amplification expense, comprising: step 1 assembles the non-sequential write request on multiple magnetic tracks；Step 2 reads the data block of multiple write requests covering across band within the time that disc rotates a circle, by the data block batch write-in disc on different magnetic tracks.This method can effectively reduce RMW bring when watt recording disc original place updates and write amplification expense.
Description
Technical field
The present invention relates to computer field of storage, in particular to a kind of incidentally read-write reduces a watt record (Shingled
Magnetic Recording, SMR) write the method for amplifying expense.
Background technique
The surface density of traditional magnetic disk as primary storage medium quickly will be close to 1TB/In2The limit also means that
The memory space of single-deck has been difficult to have greatly improved again.In order to solve this problem, in recent years, a kind of new disk skill
Art proposed in the industry, i.e. watt recording technique.Its storage that disk is further promoted by the way of partial stack adjacent track is close
Degree.However, it is this superposition magnetic track mode but make while increasing disk storage capacity disk can not directly original place more
Newly, data on its adjacent track can be covered because updating and writing the data on a tracks.For the damage for avoiding adjacent track data
It is bad, it needs to make reading modification and writes (Read-Modify-Write, RMW).As name word description, read from disk first
Then partial data modifies a part of data therein, modified and unmodified data is finally write back disk together again.
The partial data wherein read will guarantee the valid data that other positions on disk will not be covered when writing back.Therefore, the worst
In the case of, if disk is filled with valid data, a RMW may need to operate entire disk, cause it is serious write amplification, greatly
Reduction disk non-sequential write performance.Amplification is write in order to reduce, in a certain range by RMW control, reduces write performance expense,
Disk can be divided into continuous region, i.e. band (Band), as shown in Figure 1 (assuming that writing the farthest only meeting of a tracks (Track)
The data on next tracks, i.e. coverage rate K=2 are covered, each Band is made of two Track), the addressing between Band is
Continuously, but between Band reserved certain space (Gap) is needed, it is ensured that writing any position on Band1 will not all cover
Any data on Band2.In this manner it is possible to which amplification control will be write in the range of a Band, therefore writes magnification level and take
Certainly in the size of Band (Band Size).Due to needing the reserved space between adjacent Band with Band partitioning discs, magnetic is wasted
Disk capacity, to reduce disk size waste, Band Size cannot be too small, because Band Size is smaller, Band on disk
Quantity it is more, disk space waste it is more serious.Even if Band Size is set as tens MB or so, each of user
Write request may all cause RMW, need all to read the data on entire Band in the worst case, modified and write again
It returns, performance can also decline to a great extent.
Scale-up problem is write for watt recording technique RMW is serious, it is necessary to propose that a kind of method incidentally read and write is come into one
Step reduces RMW bring performance cost.
Summary of the invention
The purpose of the present invention is to provide a kind of incidentally read-writes to reduce the method that watt record writes amplification expense, can effectively reduce
RMW bring writes amplification expense when watt recording disc original place updates.
To achieve the goals above, the present invention provides a kind of method that incidentally read-write reduction watt record writes amplification expense, packet
It includes:
Step 1 assembles the non-sequential write request on multiple magnetic tracks；
The data block of multiple write requests covering across band was read, will be located within the time that disc rotates a circle by step 2
In the data block batch write-in disc on different magnetic tracks.
The incidentally read-write reduces the method that watt record writes amplification expense, wherein in the step 1, further includes:
All non-sequential write requests are saved in the buffer, and the non-sequential write request is divided in the following manner
Group:
The non-sequential write request in one group is that magnetic head is able to access that within the time that disc rotates a circle；
Magnetic head is capable of handling all non-sequential write requests in one group within the time that disc rotates a circle.
The incidentally read-write reduces the method that watt record writes amplification expense, wherein in the step 2, further includes:
Disk is according to packet transaction read-write requests.
The incidentally read-write reduces the method that watt record writes amplification expense, wherein in the step 2, further includes:
Data block is obtained in the position that the write request in corresponding group to be written and the position being capped in corresponding group；
Read the data block of capped position；
The position writing data blocks to be written according to data block, by read-out data block back.
The incidentally read-write reduces the method that watt record writes amplification expense, wherein
The writing data blocks after disc rotates a circle, the write back data block after disc second rotation.
The described incidentally read-write reduces the method that watt record writes amplification expense, wherein the time that disc rotates a circle is
10ms。
Compared with prior art, technological merit of the invention is:
The invention proposes a kind of incidentally read-writes to reduce the method that watt record writes amplification expense, and this method does user's request
Packet aggregation, non-sequential write request closer for physical distance are more by incidentally completing within the time that disc rotates a circle
Originally the time for needing to be multiple RMW is compressed to the time for being a RMW by a RMW operation, to promote the non-of watt recording disc
Sequence write performance.
Detailed description of the invention
Fig. 1 is existing disc space by band division schematic diagram；
Fig. 2 is incidentally read-write theory schematic diagram of the invention；
Fig. 3 is write request grouping schematic diagram of the invention.
Specific embodiment
Below in conjunction with the drawings and specific embodiments, the present invention will be described in detail, but not as a limitation of the invention.
As shown in Fig. 2, being incidentally read-write theory schematic diagram of the invention.
Below with reference to Fig. 2, detailed description of the present invention incidentally reads mechanism, incidentally writes mechanism.
The concrete principle for incidentally reading mechanism is:
Within the time that disc rotates a circle, the data that multiple write requests across band can cover are read together, in Fig. 2
In, for Band1, whens writing data blocks 50,60, can cover data block 70,80.For Band2, when writing data blocks 90,100
Data block 110,120 can be covered, whens writing data blocks 130,140 can cover data block 150,160.Using incidentally reading mechanism, magnetic
Head first can arrive Track2 read block 70 by tracking, and then tracking arrives Track4 read block 120, and then tracking arrives Track6
Read block 150, then tracking arrives Track2 read block 80, and then tracking arrives Track4 read block 110, then
Track6 read block 160 is sought, the R in 3 RMW originally is completed within the time that disc turns around, it is time-consuming in total
10ms。
The concrete principle for incidentally writing mechanism is:
Within the time that disc rotates a circle, disc is written into the batch data on different magnetic tracks, as shown in Fig. 2,
In the W stage of RMW, magnetic head first seeks to Track1 and writes data block 50, then seeks Track3 and write data block 100, then seek
Track5 writes data block 130, then seeks Track1 and write data block 60, then seeks Track3 and write data block 90, then seek
Track5 writes data block 140.After first lap writes, then seeks Track1 and write data block 70, then seek Track4 and write data
Block 120, then seek Track6 and write data block 150, then seek Track2 and write data block 80, then seek Track4 and write data
Block 110, then seek Track6 and write data block 160, all data have been write within the time that disc rotation takes two turns, it is time-consuming
20ms。
Assume that disc rotates a circle time-consuming 10ms as described above.
As shown in figure 3, being write request grouping schematic diagram of the invention.The present invention include non-sequential write request packet aggregation and
Incidentally read and write two parts of process.
In conjunction with Fig. 3, the process of non-sequential write request packet aggregation is described in detail.
In Fig. 3, all non-sequential write requests are stored in buffer area (Buffer), and to the request in Buffer into
Row grouping (Grouping).Two principles of group basis: 1, the request in Group must be magnetic in time that disc rotates a circle
It is accessible to obtain；2, the not angular distance between the request on same magnetic track adjacent in Group cannot be too small, it is ensured that magnetic
Head disc within the time of switching magnetic track did not rotate the past, that is, ensured that magnetic head can be handled within the time that disc rotates a circle
All requests in Group.As shown in figure 3, data block 50,60,90,100,130,140 is Group1, and so on, data block
N-k to n forms Group N, wherein k≤n, and N≤n, n indicate that the request disk address to be accessed, N indicate the number of grouping, with
This analogizes the request tissue by that can merge RMW to together.
In conjunction with Fig. 3, the process incidentally read and write is described in detail.
In Fig. 3, after write request is grouped, disk carrys out processing write requests according to Group, and Fig. 2 is in processing Group1
The situation of request, it is assumed that K=2, disc sequential addressing since most outer course, address of the digital representation data block on disc,
In, data block 50,60,90,100,130,140 is the write request position to be written in Group1.Data block 70,80,110,
120,150,160 be the position that can be capped.In the method, magnetic head can first seek Track2, read block 70, then seek
Road is to Track4 read block 120, and then tracking arrives Track6 read block 150, then seeks Track2 and read data
Block 80, then Track4 read block 110 is sought, finally tracking arrives Track6 read block 160, turns around in disc
The R in 6 RMW originally is completed in time；Equally, in the W stage of RMW, magnetic head first seeks to Track1 and writes data block 50, then
It seeks Track3 and writes data block 100, then seek Track5 and write data block 130, then seek Track1 and write data block 60, then
It seeks Track3 and writes data block 90, then seek Track5 and write data block 140, then seek Track2 and write data block 70, then
It seeks Track4 and writes data block 120, then seek Track6 and write data block 150, then seek Track2 and write data block 80, then
It seeks Track4 and writes data block 110, finally tracking writes data block 160 to Track6, writes within the time that disc rotation takes two turns
Be over all data, entirely incidentally reads and writes within the time of disc triple turn and completes, it is assumed that and the every circle of disc rotation needs 10ms,
Then have time-consuming 30ms altogether.
The data of adjacent track can be covered due to writing data on watt recording disc, have to first will before data are written
Capped data are read out, for example, data block 70 can be covered by writing data block 50, so must assure that before writing data blocks 50
70 pieces of the data that can be capped are read, and pending data block 50 could write back data block 70 after being written, if write-in data
The data on adjacent track can be also covered when block 70, then are repeated the above process.Incidentally read-write is exactly rotated a circle using disk
The data that reading as much as possible can be capped in time achieve the purpose that shorten the RMW time.
And the RMW process as unit of band for being not optimised, each non-sequential write request is required to do primary
RMW, in this example, writing data blocks 50 need first to read data block 70, and writing data blocks 50 after disc rotates a circle, disc is again
Write back data block 70 after rotating a circle, in total time-consuming 20ms.And so on, 6 requests need 6*20ms=120ms that can just write in total
Enter disc.It can be seen that making 3 times of performance boost in this example using the method incidentally read and write.
The non-sequential write request closer to the writing position on watt recording disc assembles grouping, by the way of incidentally reading and writing
Multiple RMW steps are completed within the time that disc rotation is turned around, reduces total RMW expense, it is non-suitable to improve watt recording disc
Sequence write performance.
The present invention passes through the non-sequential write request on the multiple magnetic tracks that will be closer and assembles, when magnetic head rotates a circle
In, the data to be read in the corresponding multiple RMW operation of multiple write requests are read together.In write back stage, also in magnetic head
The data of reading are write back in the time to rotate a circle, i.e., the mode incidentally read and write together, achieve the effect that reduce by RMW number,
Reduce RMW bring when watt recording disc original place updates and writes amplification expense.
Certainly, the present invention can also have other various embodiments, without deviating from the spirit and substance of the present invention, ripe
It knows those skilled in the art and makes various corresponding changes and modifications, but these corresponding changes and change in accordance with the present invention
Shape all should fall within the scope of protection of the appended claims of the present invention.
Claims (5)
1. a kind of incidentally read-write reduces the method that watt record writes amplification expense characterized by comprising
Step 1 assembles the non-sequential write request on multiple magnetic tracks；
Step 2 reads the data block of multiple write requests covering across band within the time that disc rotates a circle, will be in not
With the data block batch write-in disc on magnetic track；
In the step 1, further include:
All non-sequential write requests are saved in the buffer, and the non-sequential write request is grouped in the following manner:
The non-sequential write request in one group is that magnetic head is able to access that within the time that disc rotates a circle；
Magnetic head is capable of handling all non-sequential write requests in one group within the time that disc rotates a circle.
2. incidentally read-write according to claim 1 reduces the method that watt record writes amplification expense, which is characterized in that the step
In rapid two, further include:
Disk is according to packet transaction read-write requests.
3. incidentally read-write according to claim 2 reduces the method that watt record writes amplification expense, which is characterized in that the step
In rapid two, further include:
Data block is obtained in the position that the write request in corresponding group to be written and the position being capped in corresponding group；
Read the data block of capped position；
The position writing data blocks to be written according to data block, by read-out data block back.
4. incidentally read-write according to claim 3 reduces the method that watt record writes amplification expense, which is characterized in that
The writing data blocks after disc rotates a circle, the write back data block after disc second rotation.
5. incidentally read-write according to claim 1-4 reduces the method that watt record writes amplification expense, feature exists
In the time that disc rotates a circle is 10ms.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610153718.3A CN105824579B (en) | 2016-03-17 | 2016-03-17 | A method of incidentally read-write reduction watt record writes amplification expense |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610153718.3A CN105824579B (en) | 2016-03-17 | 2016-03-17 | A method of incidentally read-write reduction watt record writes amplification expense |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105824579A CN105824579A (en) | 2016-08-03 |
| CN105824579B true CN105824579B (en) | 2019-01-11 |
Family
ID=56525264
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610153718.3A Expired - Fee Related CN105824579B (en) | 2016-03-17 | 2016-03-17 | A method of incidentally read-write reduction watt record writes amplification expense |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105824579B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106970764B (en) * | 2017-02-24 | 2019-08-16 | 中国科学院计算技术研究所 | It is a kind of based on band version watt record a RAID write ordered method and system |
| CN108804019B (en) * | 2017-04-27 | 2020-07-07 | 华为技术有限公司 | Data storage method and device |
| CN108376053B (en) * | 2018-01-23 | 2020-02-28 | 中国科学院计算技术研究所 | Tile record writing sequencing method and device supporting in-place updating |
| CN112181315B (en) * | 2020-10-30 | 2022-08-30 | 新华三大数据技术有限公司 | Data disk refreshing method and device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102789367A (en) * | 2011-02-14 | 2012-11-21 | 希捷科技有限公司 | Dynamic storage regions |
| CN102841855A (en) * | 2011-05-23 | 2012-12-26 | 日立环球储存科技荷兰有限公司 | Storage device with inline address indirection metadata storage |
| CN102999428A (en) * | 2012-11-01 | 2013-03-27 | 华中科技大学 | Four-stage addressing method for tile recording disk |
| CN104050975A (en) * | 2013-03-15 | 2014-09-17 | 希捷科技有限公司 | Sectional reading-correcting-writing operation |
| US8867161B2 (en) * | 2013-03-15 | 2014-10-21 | Seagate Technology Llc | Shingled magnetic recording with variable track spacing |
-
2016
- 2016-03-17 CN CN201610153718.3A patent/CN105824579B/en not_active Expired - Fee Related
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102789367A (en) * | 2011-02-14 | 2012-11-21 | 希捷科技有限公司 | Dynamic storage regions |
| CN102841855A (en) * | 2011-05-23 | 2012-12-26 | 日立环球储存科技荷兰有限公司 | Storage device with inline address indirection metadata storage |
| CN102999428A (en) * | 2012-11-01 | 2013-03-27 | 华中科技大学 | Four-stage addressing method for tile recording disk |
| CN104050975A (en) * | 2013-03-15 | 2014-09-17 | 希捷科技有限公司 | Sectional reading-correcting-writing operation |
| US8867161B2 (en) * | 2013-03-15 | 2014-10-21 | Seagate Technology Llc | Shingled magnetic recording with variable track spacing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105824579A (en) | 2016-08-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105824579B (en) | A method of incidentally read-write reduction watt record writes amplification expense | |
| US9129628B1 (en) | Data management for data storage device with different track density regions | |
| US9263088B2 (en) | Data management for a data storage device using a last resort zone | |
| CN105934749B (en) | Garbage collection and data relocation for data-storage system | |
| US9720860B2 (en) | System and method for efficient processing of queued read commands in a memory system | |
| US20160005423A1 (en) | Data management for a data storage device with zone relocation | |
| US9460751B2 (en) | Binding shingled recording bands | |
| CN106293543B (en) | It is a kind of that ordered method and device is write based on watt record RAID incidentally recycled | |
| US9036284B1 (en) | Isolated shingled bands of fractional tracks | |
| US10140180B1 (en) | Segment-based outer code recovery | |
| CN106970764B (en) | It is a kind of based on band version watt record a RAID write ordered method and system | |
| US8879183B1 (en) | Segmenting of read-modify-write operations | |
| US8862856B2 (en) | Implementing remapping command with indirection update for indirected storage | |
| US8953269B1 (en) | Management of data objects in a data object zone | |
| KR20160055886A (en) | Dynamic band boundaries | |
| US8370564B2 (en) | Access control device, information processing device, access control program and access control method | |
| US9377956B1 (en) | Rewrite operation for recording bands | |
| JP6148996B2 (en) | Apparatus and method for partitioned read-modify-write operations | |
| US20130198586A1 (en) | Data storage control apparatus, data storage apparatus and data storage method in the same | |
| US10216624B1 (en) | Two-pass logical write for interlaced magnetic recording | |
| US10867633B1 (en) | Reduced adjacent track erasure from write retry | |
| US20160103621A1 (en) | Multi-Block Data Storage Using Data Compression | |
| JPH05210915A (en) | Disk device | |
| US8375177B2 (en) | Apparatus, system, and method for efficiently verifying writes | |
| US9875055B1 (en) | Check-pointing of metadata |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190111 |