CN105389135A - Solid-state disk internal cache management method - Google Patents
Solid-state disk internal cache management method Download PDFInfo
- Publication number
- CN105389135A CN105389135A CN201510919492.9A CN201510919492A CN105389135A CN 105389135 A CN105389135 A CN 105389135A CN 201510919492 A CN201510919492 A CN 201510919492A CN 105389135 A CN105389135 A CN 105389135A
- Authority
- CN
- China
- Prior art keywords
- data page
- node
- solid
- state disk
- page
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Abstract
The present invention discloses a solid-state disk internal cache management method. The method comprises: according to characteristics of asymmetry of flash memory reading and writing speeds, after a solid-state disk receives a command of an upper access data page, according to an operation type, assigning different weight values to read and write data pages in an internal cache of the solid-state disk, wherein the weight value of the writing data page is greater than that of the read data page; and in a system operation process, performing dynamic adjustment on the weight values of the data pages in the cache according to a hit condition of the data pages in the cache and the storage time of the data pages in the cache, and when the cache is full of data and needs to be replaced, a system searching data pages with the weight values of 0 in the cache for replacement. Compared with the prior art, the cache management method provided by the present invention improves the hit rate of the write data page on the premise of ensuring the total hit rate of the cache, thereby improving the performance of the solid-state disk and prolonging the service life of the solid-state disk.
Description
Technical field
The invention belongs to computer memory technical field, more specifically, relate to a kind of solid-state disk inner buffer management method.
Background technology
It is fast that flash memory (NANDFLASH) has reading and writing speed, and the feature that energy consumption is low and shock resistance is good, therefore, the solid-state disk based on flash memory has been widely used in high-end server and notebook computer.Deposit chip to be made up of a lot of data block, each data block comprises again multiple data page.In flash memory, the least unit of reading and writing operation is page, flash chip can any one page in random write flash memory, if but write operation time page object in before had data, then must first by the data erase of this page, then just new data can be write, the unit of Flash memory erase is block, i.e. each at least erasing one piece, simultaneously, to certain concrete flash block, its erasing times is limited; In flash memory, reading and writing operation required time is asymmetric, writes 8 ~ 12 times that one page required time normally reads one page required time in a flash memory.In order to improve the performance of the solid-state disk based on flash memory; based in the solid-state disk of flash memory; the data that the buffer memory that usually can add a high speed may reuse in the future for interim storage system; therefore, the buffer memory how effectively managing solid-state disk inside has great importance to raising solid-state disk performance and life-span.
Buffer storage managing algorithm conventional at present has LeastRecentlyUsed (LRU), 2queue (2Q), LowInter-referenceRecencySet (LIRS), AdaptiveReplacementCache (ARC) etc., and these conventional buffer storage managing algorithms are all the cache managements for host side.Because flash memory has the asymmetric feature of reading and writing, therefore direct they to be used in effect in the cache management of solid-state disk inside unsatisfactory.
Buffer storage managing algorithm for solid-state disk inside mainly contains for BPLRU (BlockPaddingLeastRecentlyUsed) algorithm of mixed-use developments FlashTranslationLayer (FTL) and CFLRU (Clean-FirstLRU) algorithm for page mapping, BPLRU algorithm is when having access to arbitrary page of certain block in flash memory, just its whole piece is called in buffer memory, swap out buffer memory time be also disposablely write back FLASH by whole piece, the advantage of this algorithm random read-write to be converted to order read-write, reduce during mixing stores simultaneously and merge (merger) operation, but BPLRU method is not suitable for the solid-state disk that page maps, because each logical page (LPAGE) can be mapped on the arbitrary Physical Page of solid-state disk in the solid-state disk of page mapping.
Page mapping FTL has simply, efficient feature, in current solid-state disk, the normal page that adopts maps FTL algorithm, map in FTL solid-state disk at page, the management of SSD inner buffer adopts CFLRU algorithm usually, when buffer memory write full data need replace time, as long as there is clean data page in buffer memory, just clean data page is replaced away in LRU mode; CFLRU method do not consider clean data page (read data) future system whether also can visit again, in extreme circumstances, the method can be degenerated to the lru algorithm only dirty data page (writing data) being carried out to buffer memory; In the load based on read operation, the hit rate of CFLRU can be very low, thus affect performance and the life-span of SSD; Meanwhile, CFLRU will cause a lot of write once data prolonged stay in the buffer, affect the performance of solid-state disk.
Summary of the invention
For above defect or the Improvement requirement of prior art, the invention provides a kind of solid-state disk inner buffer management method, its object is to solve the low problem of existing solid-state disk inner buffer management method cache hit rate.
To achieve these goals, the invention provides a kind of solid-state disk inner buffer management method, specifically comprise the steps:
(1) when solid-state disk receives upper access data page X order, the initial value of setting data page X weights; And judge whether data page X is in queue A1in; If so, then access this data page, enter step (9); If not, then step (2) is entered;
(2) judge whether data page X is in ring AClock, if so, then enter step (3), if not, then enter step (5);
(3) whether the action judging visit data page X is read operation, if so, this data page X is moved on to the afterbody of ring AClock, enters step (9); If not, then show that the action of visit data page X is write operation, enter step (4);
(4) weights of this data page X are set to maximal value, and this data page X is moved on to the afterbody of ring AClock, enter step (9);
(5) be data page X allocation space in the buffer, enter step (6);
(6) judge that the mark of data page X is whether in queue A1out, if so, enters step (7); If not, then step (8) is entered;
(7) initial value of setting data page X weights, and this page is inserted into the afterbody of AClock, enter step (9);
(8) data page X is inserted into the head of queue A1in, enters step (9);
(9) read and write access is terminated;
Wherein, A1in refers to a fifo queue in solid-state disk in inner buffer; A1out refers to the recent minimum use queue of in solid-state disk inner buffer, i.e. LRU queue; AClock is the unidirectional ring that in solid-state disk inner buffer, couples together with pointer, and save two pointers in this unidirectional ring, head pointer and tail pointer, tail pointer is close to head pointer, and namely the next node of tail pointer points to head node;
A1in and AClock is all data cached for depositing; A1out is used for the mark of store data page; When a data page replaces out buffer memory from A1in, the mark of this data page is saved in A1out queue; When a page replaces out buffer memory from AClock ring, this page is abandoned.
Preferably, the method for the initial value of setting data page X weights is specific as follows:
A (), for write operation, the establishing method of data page X weights initial value is:
Time used for flash memory write data page is read the data page time used divided by flash memory, coefficient 0.6 is multiplied by the result obtained, then rounds, obtain and write weights initial value; If coefficient is too large, in buffer memory, write data page by what preserve a large amount of disposable access, if coefficient is too little, then much again accessedly may to write data page and may be replaced out buffer memory in the future, cause system performance to reduce, life-span reduction; Experimental result display system when coefficient is 0.6 can reach preferably result; Shown in following formula:
Wherein,
represent that logarithm value rounds;
B (), for read operation, data page weights initial value is 0.
Preferably, in above-mentioned steps (4), the weights of data page X are got 5; The expense 10 write is doubly to read operation; For write operation, if it is excessive to write weights, then extending the write once data page residence time in the buffer, being replaced out buffer memory, influential system overall performance by causing a lot of read data that in the future may be accessed again; If it is too small to write weights, then writes data page and will replace out buffer memory faster, the hit rate of write operation can be reduced.
Preferably, A1in queue length maximal value Kin is 20% ~ 30% of cache size; The size of AClock ring is 70% ~ 80% of buffer memory total amount; A1out queue saves the mark of each data page, in A1out queue, and 40% ~ 60% of energy store data page number that the maximum number Kout that can preserve mark is buffer memory.
Preferably, described is the method for data page X allocation space in the buffer, specifically comprises the steps:
(5.1) to judge in buffer memory if so, then to enter whether available free space step (5.2), if not, then enter step (5.3);
(5.2) by data page X stored in the spatial cache node Z be assigned to, enter step (5.12);
(5.3) judge whether queue A1in length is greater than A1in queue length maximal value Kin, if so, then enter step (5.4); If not, then step (5.10) is entered;
(5.4) data block of A1in afterbody is replaced out buffer memory, the data of replacing out are designated as Y, enter step (5.5);
(5.5) data page X is put into the cache node Z replaced out, enter step (5.6);
(5.6) extract the zone bit of the data Y replaced out, enter step (5.7);
(5.7) judge whether queue A1out length is greater than the maximum number Kout that A1out queue can preserve identifier, if so, enter step (5.8); If not, then step (5.9) is entered;
(5.8) the afterbody node of A1out is deleted from A1out queue, enter step (5.9);
(5.9) zone bit of the above-mentioned data Y replaced out is put into new node L, the next node of L node is pointed to the head node of A1out, and using the head node of L node as A1out, enter step (5.12);
(5.10) obtain the weights of AClock head node data page, judge whether these weights are 0, if so, then enter step (5.12); If not, then step (5.11) is entered;
(5.11) these data page weights are subtracted 1, make tail pointer point to head page node, head pointer points to next data page node, enters step (5.10);
(5.12) head node is denoted as Z, makes head pointer point to the next node of Z node, make the next node of tail pointer point to head node, enter step (5.13);
(5.13) Z node is peeled off from AClock, data X is put into Z node, enter step (5.14);
(5.14) function returns Z node, and allocation space terminates;
In AClock, to read data page with write data page and give different weights time initial; In system operation, the weights dynamic change of each data page, replaces the page node that weights are 0 when carrying out replacement operation in AClock ring;
Weights be 0 data page have read data page and for a long time in the buffer and accessed do not write data page, because read data page is the clean page, do not need to operate flash memory during replacement, the expense of replacing read data page is less; When weights be 0 the page for writing data page time, then show that this page did not have in the long period accessed, according to principle of locality, replace weights become 0 write the hit rate that data page can improve buffer memory, thus improve system performance.
In general, the above technical scheme conceived by the present invention compared with prior art, different weights are given to write operation and read operation data page, make to write the time that time that data page stops in the buffer stops in the buffer than read data page longer, and according to the hit situation of data page, the weights of each page of dynamic adjustment, improve when making to be buffered in the higher total hit rate of maintenance and write hit rate, reach raising system performance, postpone the object in serviceable life; Specifically can obtain following listed beneficial effect:
(1) solid-state disk inner buffer management method provided by the invention, according to the feature of flash memory reading and writing speed asymmetry, gives different weights to reading and writing data page in solid-state disk inner buffer; The weights writing data page are greater than read data page; The time that in buffer memory, the weights of data page are deposited in the buffer according to data page hit situation in buffer memory and data page carries out dynamic conditioning, when buffer memory write full data need to replace time, in system searching buffer memory weights be 0 data page replace;
Larger than read data also weights owing to writing data page weights, write the time that the data page residence time in the buffer stops in the buffer than read data page long, write hit rate to improve buffer memory; Therefore, improve when ensureing buffer memory total hit rate (read hit rate+write hit rate) hit rate writing data page, thus improve the performance of solid-state disk, extend the life-span of solid-state disk simultaneously;
(2) solid-state disk inner buffer management method provided by the invention, there is total hit rate (reading to hit+write hit) that the buffer storage managing algorithm traditional with 2Q etc. is suitable, but owing to giving higher weights by writing data page than read data page in the present invention, make to write the time that time that data page stops in the buffer stops than read data page longer, improve and write hit rate; Experimental result shows, it write hit rate and can improve 30% at most than traditional algorithms such as 2Q; In addition, in Aclock ring, each page foundation principle of locality, dynamic conditioning weights, thus the system that ensure that has higher total hit rate;
(3) solid-state disk inner buffer management method provided by the invention, with the solid-state disk inner buffer management algorithm mapped based on page as CFLRU comparatively speaking, owing to employing two queues in this invention buffer memory, eliminate the pollution of disposable data to buffer memory; Make its total hit rate improve 10 ~ 35% than CFLRU, write hit rate also a little more than CFLRU;
(4) solid-state disk inner buffer management method provided by the invention, because total hit rate can reach buffer storage managing algorithm level outstanding at present, but compared to prior art, what improve buffer memory writes hit rate, therefore can reduce solid-state disk to operate the reading and writing of flash memory, particularly write operation, thus performance and life-span that solid-state disk can be improved.
Accompanying drawing explanation
Fig. 1 is the process flow diagram of the solid-state disk inner buffer management method based on weight provided by the invention;
Fig. 2 is the process flow diagram of data page X allocation space in the buffer;
Fig. 3 is the schematic diagram of the unidirectional ring of AClock.
Embodiment
In order to make object of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.In addition, if below in described each embodiment of the present invention involved technical characteristic do not form conflict each other and just can mutually combine.
The solid-state disk buffer memory management method that embodiment provides, its flow process is as shown in Figure 1, specific as follows:
(1) when solid-state disk receives upper access data page X order, the weights of setting data page X, if be read operation to X page, then the weights of data page X are set to 0; If be write operation to data page X, then the weights of data page X are set to 5; And judge whether data page X is in solid-state disk in inner buffer queue A1in; If so, then access this data page, enter step (9); If not, then step (2) is entered;
(2) judge whether data page X is in solid-state disk inner buffer ring AClock, if so, then enters step (3); If not, then step (5) is entered;
(3) whether the action judging visit data page X is read operation, and if so, then the weights of this data page X are constant, this data page X moved on to the afterbody of ring AClock, enter step (9); If not, then show that the action of visit data page X is write operation, enter step (4);
(4) weights of this data page X are set to 5, and this data page X is moved on to the afterbody of ring AClock, enter step (9);
(5) be data page X allocation space in the buffer, enter step (6);
(6) judge that the mark of data page X is whether in queue A1out, if so, enters step (7); If not, then step (8) is entered;
(7) whether the action judging visit data page X is read operation, if so, then the weights of this data page is set to 0;
If not, then show that the action of visit data page X is write operation, the weights of this data page are set to 5, and this page is inserted into the afterbody of AClock, enter step (9);
(8) using the head node of data page X as A1in, be inserted into the head of queue A1in, enter step (9);
(9) read and write access is terminated.
The flow process that above-mentioned steps (5) is data page X distribution spatial cache is as shown in Figure 2, specific as follows:
(5.1) to judge in solid-state disk inner buffer if so, then to enter whether available free space step (5.2), if not, then enter step (5.3);
(5.2) for data page X distributes spatial cache, and by data page X stored in the spatial cache node Z be assigned to, step (5.12) is entered;
(5.3) judge whether queue A1in length is greater than A1in queue length maximal value Kin, if so, then enter step (5.4); If not, then step (5.10) is entered;
(5.4) taken from queue A1in by the data block Y of A1in afterbody, the tail pointer of A1in points to the previous node of data block Y, enters step (5.5);
(5.5) data page X is put into the cache node Z replaced out, enter step (5.6);
(5.6) extract the zone bit of the data Y taken from queue A1in, enter step (5.7);
(5.7) judge whether queue A1out length is greater than the maximum number Kout that A1out queue can preserve identifier, if so, enter step (5.8); If not, then step (5.9) is entered;
(5.8) the afterbody node of A1out is deleted from A1out queue, enter step (5.9);
(5.9) mark of the above-mentioned data block Y replaced out is put into new node L, the next node of L node is pointed to the head node of A1out, and using the head node of L node as A1out, enter step (5.12);
(5.10) obtain the weights of AClock head node data page, judge whether these weights are 0, if so, then enter step (5.12); If not, then step (5.11) is entered;
(5.11) these data page weights are subtracted 1, make tail pointer point to head page node, head pointer points to next data page node, enters step (5.10);
(5.12) head node is denoted as Z, makes head pointer point to the next node of Z node, make the next node of tail pointer point to head node, enter step (5.13);
(5.13) Z node is taken from AClock (namely the head node of AClock points to the next node of head node, and tail node points to new head node again), data X is put into Z node, enter step (5.14);
(5.14) function returns Z node, terminates allocation of space.
Figure 3 depicts the structure of Aclock ring.In Aclock, be provided with two pointers, head pointer head and tail pointer tail, the next of tail pointer points to head node head.Individual node all possesses different weights in addition.
The present embodiment, according to the feature of flash memory reading and writing asymmetry, gives the weights that the reading and writing pages is different, thus making to write time that the page stops in the buffer, comparatively to read the time that the page stops in the buffer longer, to improve the hit rate writing data also page; Meanwhile, according to principle of locality in system operation, the weights of each page of dynamic conditioning, thus the system that ensure that has the total hit rate of good buffer memory.
Those skilled in the art will readily understand; the foregoing is only preferred embodiment of the present invention; not in order to limit the present invention, all any amendments done within the spirit and principles in the present invention, equivalent replacement and improvement etc., all should be included within protection scope of the present invention.
Claims (5)
1. a solid-state disk inner buffer management method, is characterized in that, described method specifically comprises the steps:
(1) when solid-state disk receives upper access data page X order, the initial value of setting data page X weights; And judge whether data page X is in queue A1in; If so, then access this data page, enter step (9); If not, then step (2) is entered;
(2) judge whether data page X is in ring AClock, if so, then enter step (3); If not, then step (5) is entered;
(3) whether the action judging visit data page X is read operation, if so, this data page X is moved on to the afterbody of ring AClock, enters step (9); If not, then step (4) is entered;
(4) weights of this data page X are set to maximum, and this data page X is moved on to the afterbody of ring AClock, enter step (9);
(5) be data page X allocation space in the buffer, enter step (6);
(6) judge that the mark of data page X is whether in queue A1out, if so, enters step (7); If not, then step (8) is entered;
(7) initial value of setting data page X weights, and this page is inserted into the afterbody of AClock, enter step (9);
(8) data page X is inserted into the head of queue A1in, enters step (9);
(9) access is terminated;
Described A1in refers to a fifo queue in solid-state disk in inner buffer, and AClock is the unidirectional ring that in solid-state disk inner buffer, couples together with pointer, and A1out refers to the recent minimum use queue of in solid-state disk inner buffer; When data page replaces out buffer memory from A1in, the mark of this data page is saved in A1out queue; When data page replaces out buffer memory from AClock ring, this data page is abandoned.
2. solid-state disk inner buffer management method as claimed in claim 1, it is characterized in that, the method to set up of data page X weights initial value is specific as follows:
A (), for write operation, the initial establishing method of data page X weights is:
Time used for flash memory write data page is read the data page time used divided by flash memory, coefficient 0.6 is multiplied by the result obtained, then rounds, obtain and write weights initial value; Shown in following formula:
Wherein,
represent that logarithm value rounds;
B (), for read operation, data page weights initial value is 0.
3. solid-state disk inner buffer management method as claimed in claim 2, is characterized in that, in described step (4), the weights of data page X are got 5.
4. solid-state disk inner buffer management method as claimed in claim 1 or 2, it is characterized in that, described A1in queue length maximal value Kin is 20% ~ 30% of cache size; The size of AClock ring is 70% ~ 80% of buffer memory total amount; In A1out queue, the maximum number Kout that can preserve mark can deposit 40% ~ 60% of page number for buffer memory.
5. solid-state disk inner buffer management method as claimed in claim 1 or 2, is characterized in that, described is the method for data page X allocation space in the buffer, specifically comprises the steps:
(5.1) to judge in buffer memory if so, then to enter whether available free space step (5.2), if not, then enter step (5.3);
(5.2) by data page X stored in the spatial cache node Z be assigned to, enter step (5.12);
(5.3) judge whether queue A1in length is greater than A1in queue length maximal value Kin, if so, then enter step (5.4); If not, then step (5.10) is entered;
(5.4) data block of A1in afterbody is replaced out buffer memory, the data of replacing out are designated as Y, enter step (5.5);
(5.5) data page X is put into the cache node Z replaced out, enter step (5.6);
(5.6) extract the zone bit of the data Y replaced out, enter step (5.7);
(5.7) judge whether queue A1out length is greater than the maximum number Kout that A1out queue can preserve identifier, if so, enter step (5.8); If not, then step (5.9) is entered;
(5.8) the afterbody node of A1out is deleted from A1out queue, enter step (5.9);
(5.9) zone bit of the described data Y replaced out is put into new node L, the next node of L node is pointed to the head node of A1out, and using the head node of L node as A1out, enter step (5.12);
(5.10) obtain the weights of AClock head node data page, judge whether these weights are 0, if so, then enter step (5.12); If not, then step (5.11) is entered;
(5.11) these data page weights are subtracted 1, make tail pointer point to head page node, head pointer points to next data page node, enters step (5.10);
(5.12) head node is denoted as Z, makes head pointer point to the next node of Z node, make the next node of tail pointer point to head node, enter step (5.13);
(5.13) Z node is peeled off from ring AClock, data X is put into Z node, enter step (5.14);
(5.14) function returns Z node, terminates allocation of space.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510919492.9A CN105389135B (en) | 2015-12-11 | 2015-12-11 | A kind of solid-state disk inner buffer management method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510919492.9A CN105389135B (en) | 2015-12-11 | 2015-12-11 | A kind of solid-state disk inner buffer management method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105389135A true CN105389135A (en) | 2016-03-09 |
| CN105389135B CN105389135B (en) | 2018-12-07 |
Family
ID=55421454
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510919492.9A Active CN105389135B (en) | 2015-12-11 | 2015-12-11 | A kind of solid-state disk inner buffer management method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105389135B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106775477A (en) * | 2016-12-19 | 2017-05-31 | 湖南国科微电子股份有限公司 | SSD master control BE Buffer, SSD master control, data transfer management device and method |
| CN108762664A (en) * | 2018-02-05 | 2018-11-06 | 杭州电子科技大学 | A kind of solid state disk page grade buffer queue management method |
| CN109213695A (en) * | 2017-06-30 | 2019-01-15 | 伊姆西Ip控股有限责任公司 | Buffer memory management method, storage system and computer program product |
| CN109407996A (en) * | 2018-11-07 | 2019-03-01 | 郑州云海信息技术有限公司 | A kind of buffer memory management method and relevant apparatus |
| CN109697033A (en) * | 2018-12-19 | 2019-04-30 | 中国人民解放军国防科技大学 | Tile record disk sensing storage caching method and system |
| CN111367627A (en) * | 2018-12-26 | 2020-07-03 | 北京奇虎科技有限公司 | Processing method and device for disk reading and writing task |
| CN111722797A (en) * | 2020-05-18 | 2020-09-29 | 西安交通大学 | SSD and HA-SMR hybrid storage system oriented data management method, storage medium and device |
| CN112559384A (en) * | 2020-12-17 | 2021-03-26 | 南昌航空大学 | Dynamic partitioning method for hybrid solid-state disk based on nonvolatile memory |
| CN114063921A (en) * | 2021-11-12 | 2022-02-18 | 福建师范大学 | Solid state disk mapping record cache management method based on reinforcement learning |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103150136A (en) * | 2013-03-25 | 2013-06-12 | 中国人民解放军国防科学技术大学 | Implementation method of least recently used (LRU) policy in solid state drive (SSD)-based high-capacity cache |
| CN103984736A (en) * | 2014-05-21 | 2014-08-13 | 西安交通大学 | Efficient buffer management method for NAND flash memory database system |
-
2015
- 2015-12-11 CN CN201510919492.9A patent/CN105389135B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103150136A (en) * | 2013-03-25 | 2013-06-12 | 中国人民解放军国防科学技术大学 | Implementation method of least recently used (LRU) policy in solid state drive (SSD)-based high-capacity cache |
| CN103984736A (en) * | 2014-05-21 | 2014-08-13 | 西安交通大学 | Efficient buffer management method for NAND flash memory database system |
Non-Patent Citations (2)
| Title |
|---|
| ARUL SELVAN RAMASAMY: "RFLRU:A Buffer Cache Management Algorithm for Solid State Drive to Improve the Write Performance on Mixed Workload", 《ENGINEERING LETTERS》 * |
| YANFEI LV等: "HAT: An Efficient Buffer Management Method for Flash-based Hybrid Storage Systems", 《FRONT.COMPUT.SCI RESEARCH ARTICALE》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106775477A (en) * | 2016-12-19 | 2017-05-31 | 湖南国科微电子股份有限公司 | SSD master control BE Buffer, SSD master control, data transfer management device and method |
| CN109213695B (en) * | 2017-06-30 | 2023-07-21 | 伊姆西Ip控股有限责任公司 | Cache management method, storage system and computer program product |
| CN109213695A (en) * | 2017-06-30 | 2019-01-15 | 伊姆西Ip控股有限责任公司 | Buffer memory management method, storage system and computer program product |
| CN108762664A (en) * | 2018-02-05 | 2018-11-06 | 杭州电子科技大学 | A kind of solid state disk page grade buffer queue management method |
| CN108762664B (en) * | 2018-02-05 | 2021-03-16 | 杭州电子科技大学 | Solid state disk page-level cache region management method |
| CN109407996A (en) * | 2018-11-07 | 2019-03-01 | 郑州云海信息技术有限公司 | A kind of buffer memory management method and relevant apparatus |
| CN109407996B (en) * | 2018-11-07 | 2022-02-18 | 郑州云海信息技术有限公司 | Cache management method and related device |
| CN109697033A (en) * | 2018-12-19 | 2019-04-30 | 中国人民解放军国防科技大学 | Tile record disk sensing storage caching method and system |
| CN109697033B (en) * | 2018-12-19 | 2022-01-07 | 中国人民解放军国防科技大学 | Tile record disk sensing storage caching method and system |
| CN111367627A (en) * | 2018-12-26 | 2020-07-03 | 北京奇虎科技有限公司 | Processing method and device for disk reading and writing task |
| CN111367627B (en) * | 2018-12-26 | 2024-02-13 | 三六零科技集团有限公司 | Method and device for processing read-write disk task |
| CN111722797A (en) * | 2020-05-18 | 2020-09-29 | 西安交通大学 | SSD and HA-SMR hybrid storage system oriented data management method, storage medium and device |
| CN112559384A (en) * | 2020-12-17 | 2021-03-26 | 南昌航空大学 | Dynamic partitioning method for hybrid solid-state disk based on nonvolatile memory |
| CN112559384B (en) * | 2020-12-17 | 2022-04-15 | 南昌航空大学 | Dynamic partitioning method for hybrid solid-state disk based on nonvolatile memory |
| CN114063921B (en) * | 2021-11-12 | 2023-06-27 | 福建师范大学 | Solid state disk mapping record cache management method based on reinforcement learning |
| CN114063921A (en) * | 2021-11-12 | 2022-02-18 | 福建师范大学 | Solid state disk mapping record cache management method based on reinforcement learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105389135B (en) | 2018-12-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105389135A (en) | Solid-state disk internal cache management method | |
| CN107193646B (en) | High-efficiency dynamic page scheduling method based on mixed main memory architecture | |
| CN110134514B (en) | Extensible memory object storage system based on heterogeneous memory | |
| CN106547703B (en) | A kind of FTL optimization method based on block group structure | |
| CN103425600B (en) | Address mapping method in a kind of solid-state disk flash translation layer (FTL) | |
| CN102981963B (en) | A kind of implementation method of flash translation layer (FTL) of solid-state disk | |
| CN107463447B (en) | B + tree management method based on remote direct nonvolatile memory access | |
| CN107391398B (en) | Management method and system for flash memory cache region | |
| CN108108128A (en) | A kind of abrasion equilibrium method and SSD for mixing SSD | |
| CN103440207A (en) | Caching method and caching device | |
| CN101477492A (en) | Circulating rewriting flash memory equalization method used for solid state disk | |
| CN110413537B (en) | Flash translation layer facing hybrid solid state disk and conversion method | |
| CN111708481B (en) | Solid State Disk (SSD) double-area wear leveling method based on super block | |
| CN108845957B (en) | Replacement and write-back self-adaptive buffer area management method | |
| CN104899154B (en) | The page management method hosted is mixed based on embedded system | |
| CN108762671A (en) | Mixing memory system and its management method based on PCM and DRAM | |
| CN109240944B (en) | Data reading and writing method based on variable-length cache line | |
| CN110888600B (en) | Buffer area management method for NAND flash memory | |
| CN106354658B (en) | A method of it reducing mapping table memory source in mixed-use developments algorithm and occupies | |
| CN110888886A (en) | Index structure, construction method, key value storage system and request processing method | |
| CN111580754B (en) | Write-friendly flash memory solid-state disk cache management method | |
| CN111352593B (en) | Solid state disk data writing method for distinguishing fast writing from normal writing | |
| CN110262982A (en) | A kind of method of solid state hard disk address of cache | |
| CN107133369A (en) | A kind of distributed reading shared buffer memory aging method based on the expired keys of redis | |
| CN107908358A (en) | A kind of method of reduction NVMe solid state disk writes amplification |
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 |