CN109857680A - A kind of LRU flash cache management method based on dynamic page weight - Google Patents
A kind of LRU flash cache management method based on dynamic page weight Download PDFInfo
- Publication number
- CN109857680A CN109857680A CN201811394983.6A CN201811394983A CN109857680A CN 109857680 A CN109857680 A CN 109857680A CN 201811394983 A CN201811394983 A CN 201811394983A CN 109857680 A CN109857680 A CN 109857680A
- Authority
- CN
- China
- Prior art keywords
- page
- request
- weight
- workspace
- exchange area
- 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
- 238000007726 management method Methods 0.000 title claims abstract description 15
- 238000000034 method Methods 0.000 claims abstract description 32
- 238000003780 insertion Methods 0.000 claims abstract description 5
- 230000037431 insertion Effects 0.000 claims abstract description 5
- 230000015654 memory Effects 0.000 claims description 29
- 230000009471 action Effects 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 2
- 230000015572 biosynthetic process Effects 0.000 description 7
- 238000003786 synthesis reaction Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 5
- 230000003139 buffering effect Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000013508 migration Methods 0.000 description 2
- 230000005012 migration Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005265 energy consumption Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Landscapes
- Memory System Of A Hierarchy Structure (AREA)
Abstract
The LRU flash cache management method based on dynamic page weight that the invention discloses a kind of, step S1: page request in read request queue is simultaneously identified and is classified to page request type and locating region;Step S2: judgement insertion buffer state is determined to eliminate the page using the LRU, method based on dynamic page weight, adjusts buffer state, executes page request.Using technical solution of the present invention, buffer area type is divided into workspace and exchange area, the cold page, the hot page, containing dirty pages face, the clean page are distinguish, determine page request type and buffer zone, determines that eliminating the page completes page request in conjunction with the LRU, method based on dynamic page weight.Read-write consumption and time delay during flash cache read-write can be effectively reduced in technical solution of the present invention, while can also greatly promote the hit rate in flash cache read-write.
Description
Technical field
The present invention relates to storage system fields, and in particular to a kind of LRU flash cache management based on dynamic page weight
Method.
Background technique
High-performance that nand flash memory memory technology has by it, the characteristics such as small in size, low energy consumption, obtain in enterprise's application
It is widely applied.But with the continuous development of big data technology in recent years, the processing and analysis of mass data are to storage system
Data throughout and I/O delay more stringent requirements are proposed, in nand flash memory there is read-write consumption is big, asymmetry I/
The defects of O delay and block are wiped prevents it from substituting hard-disc storage completely.
By caching technology and storage equipment combine can be effectively reduced I/O delay, reduce it is asymmetric in different accumulation layers
Property.However buffer memory management methods most at present is optimized for hard disk storage devices, and the buffer area for flash memory is lacked
Management method.Traditional buffer memory management method, which is applied, has that hit rate is low in flash memory, and above-mentioned time delay is high, writes and disappears
Consuming high problem can not solve.
Therefore, in view of the above-mentioned problems, the present invention proposes a kind of technical solution, overcome technological deficiency of the existing technology.
Summary of the invention
For the deficiency of current flash cache area management method, the present invention proposes a kind of LRU based on dynamic page weight
Flash memory management method reduces the write operation of flash memory and the access office using memory device to improve the hit rate of buffer area
Domain property and time locality come the buffer area of maintenance efficient.
In order to solve technical problem of the existing technology, technical scheme is as follows:
A kind of LRU flash memory management method based on dynamic page weight is divided into two stages progress, specific to execute step such as
Under:
Step S1: page request in read request queue is simultaneously identified and is divided to page request type and locating region
Class;
Step S2: judgement insertion buffer state determines the superseded page using the LRU, method based on dynamic page weight,
Buffer state is adjusted, page request is executed.Wherein step S1 further will include:
Step S11: pre-processing page request queue, judges whether the corresponding page of page request is stored in buffering
Qu Zhong, if buffer area hit thens follow the steps S12, if buffer area miss thens follow the steps S2.The step is that S11 is further wrapped
It includes:
Step S111: page request queue is pre-processed, and sets page request as R, wherein including request number Rpid
And page request mode Ram∈{read,write}.Request queue S can be indicated with following equation:
S={ R1,R2,...,Rj,...,Rn},1≤j≤n
Request queue S is multiple page request RpidSet, wherein n represent the request in request queue S sum, j generation
Table page request number.
Step S112: setting the page as P, page set U={ P1,P2,...,Pi,...,Pm},1≤i≤m≤n.Wherein m
The page sum in page set U is represented, i represents page number.Judge the page request R in request queue SjIt is corresponding
Page PiIt is no to save in the buffer, if buffer area hit thens follow the steps S12, if buffer area miss thens follow the steps S2.
Step S12: hit buffer area page P is setiPage request set ASi={ AS1,...,ASk,...,ASo}。
Wherein o represents request sum, and k represents request number.Judgement hit buffer area page PiIn workspace or
Exchange area, if PiS13 is thened follow the steps in workspace, if PiS2 is thened follow the steps in exchange area.
Step S13: the page will be transferred to region most-often used in workspace, and the content of pages of workspace is according to most normal
It is adjusted using page algorithm, ending request page algorithm.
Step S2: judgement insertion buffer state determines the superseded page using the LRU, method based on dynamic page weight,
Buffer state is adjusted, page request is executed.The step further comprises:
Step S21: judging whether the page has read, if it is unread then follow the steps S22 if the page has been read
Data are read in flash memory and execute step S23.
Step S22: judging whether workspace has expired, and workspace completely thens follow the steps S23, and workspace is less than then by the page
It is directly added into workspace and executes step S25.
Step S23: it calls to delete using renewal of the page algorithm recently in exchange area and eliminates the page, if the page carrys out self-exchange
Area thens follow the steps S24, is directly added into exchange area if the page is from flash memory and executes step S25;Step S23 is further wrapped
It includes:
Step S231: judging whether the page comes from exchange area, is, directly returns to the region the LRU parallel step in workspace
Otherwise S234 carries out S232.
Step S232: the power of every page in exchange area Page Range [0, w] is calculated according to dynamic page weight more new algorithm
Weight Pweight, and generate MapweightSave updated every page weight.Wherein w is page total size, PweightIt represents dynamic
The weight of every page, Map in state page weight algorithmweightFor dynamic page weight map.
Step S233: traversal MapweightIn every page, find and wherein the smallest one page of weight and be denoted as the superseded page,
It returns and eliminates page address.
Step S234: judging request page source, if request page thens follow the steps S24 from exchange area;If requested page
Face then expels in exchange area from flash memory and eliminates the page, and request page is inserted into execution step S25 in exchange area.
Step S24: calling dynamic page weight to select to eliminate the page in workspace, and the selected superseded page is expelled best friend
Area is changed, and the page need to be added, workspace is added.
Step S241: initialization sets target pages as Pi, request page address is Rpid, the kth request of request page
ForS is current all request RiSet.Page, which is calculated, according to following equation works as front PiTime limitation
TLI is set as PiThe equispaced of arrival can be evaluated whether P by TLIiTime locality, be based onIt can be pre-
The arrival time of other pages in page set U is surveyed,Be expressed as follows column formula:
Wherein m is buffer area page sum, and o is current page request number.
Step S242: read operation time-consuming LrWith write operation time-consuming LwBe expressed as follows column formula:
Wherein g represents current operation number, and q represents the Action number finally accessed.
Page delay degree is calculated using different algorithms for the clean page (CC, HC) and containing dirty pages face (CD, HD)
Set ASlatestFor PiIn the request signal that finally reaches, ASoFor the request signal currently reached, page freshnessSuch as following formula:
By acquiring page freshnessPage delay degreeAnd time limitationAccording to following equation meter
Calculate workspace page PiDynamic page weight
Wherein α is setting weight coefficient, PamFor page read-write state, o is current page request number, and m is page sum,
Q is recent visit operation.
The dynamic page weight of all pages is calculated in workspace, acquires minimum value
Step S243: willIt is denoted as the superseded page and expels to exchange area, execute step S25.
Step S25: current page request, which executes, to be terminated, and next request in page request queue is executed.
By adopting the above technical scheme, the present invention has following technical characterstic:
1. having used workspace and the page frame replacement region of exchange area two in this method, pass through access mode and access frequency
To distinguish the page.The page is divided into four seed types: the lonely reason page (cold-clean), cold containing dirty pages face (cold-dirty), heat are clear
Manage the page (hot-clean) and hot containing dirty pages face (hot-dirty).
2. distinguishing over conventional buffer queue management method, this method uses new page frame replacement method.In page frame replacement mistake
Cheng Zhong, buffer area not will receive the pollution of sequence of pages scanning, will have enough clean pages therefore will not be full of in buffer area
Containing dirty pages face.
3. proposing a kind of novel dynamic page Weight algorithm in this method, it considers time localities, operation
Cost and page timeliness.Operating cost during buffer queue management can postpone according to the read-write of different flash memories to carry out
Adjustment.
Compared with prior art, the present invention has following technical effect that
(1) low delay: the present invention distinguishes the cold page and the hot page.When buffer area is full, the hot page is expelled first,
The cold page executes delay expulsion, so that effectively increasing system execution efficiency reduces time delay.Compared with traditional LRU, method, this
Invention delay time reduces about 18.8%.
(2) high hit ratio: the present invention distinguishes containing dirty pages face and the clean page.Preferentially retain clean page in buffer area
Face avoids the case where containing dirty pages face is full of buffer area, to promote hit ratio.Compared with traditional LRU, method, the present invention hits ratio
Improve about 8.3%.
(3) low to write consumption: present invention improves over page migration methods.The superseded page when workspace is full, in workspace
Will expel to exchange area and and indirect expulsion, write consumption to reduce in system.Compared with traditional LRU, method, this
Consumption is write in invention reduces about 22.6%.
Detailed description of the invention
Fig. 1 is the LRU flash cache management method flow chart based on dynamic page weight;
Fig. 2 is that the LRU flash cache management method page based on dynamic page weight shifts figure;
Fig. 3 be flash memory in workspace with exchange section data interaction schematic diagram;
Fig. 4 is that synthesis path 1 reads and writes Expenditure Levels;
Fig. 5 is that synthesis path 2 reads and writes Expenditure Levels;
Fig. 6 is that synthesis path 3 reads and writes Expenditure Levels;
Fig. 7 is that synthesis path 4 reads and writes Expenditure Levels;
Fig. 8 is the present invention compared with the hit rate of other methods;
Fig. 9 be the present invention with other methods write consumption compared with;
Figure 10 is the present invention compared with the time delay of other methods.
Specific embodiment
Below with reference to accompanying drawing content to a kind of flash cache manager based on dynamic page weight provided by the invention
Method further illustrates.
Referring to Fig. 1, it is shown flow chart of the invention, a specific embodiment of the invention following steps:
Step S1: page request in read request queue is simultaneously identified and is divided to page request type and locating region
Class specifically comprises the following steps:
Step S11: page request queue is pre-processed, and sets page request as R, wherein including request number Rpid
And page request mode Ram∈{read,write}.Request queue S can be indicated with following equation:
S={ R1,R2,...,Rj,...,Rn},1≤j≤n
Request queue S is multiple page request RpidSet, wherein n represent the request in request queue S sum, j
Represent page request number.The page is set as P, page set U is indicated with following equation:
U={ P1,P2,...,Pi,...,Pm},1≤i≤m≤n
Page request is the set of multiple page P, and wherein m represents the sum of the page in page set U, and i represents page
Face number.Judge the page request R in request queue SjCorresponding page PiIt is no to save in the buffer, if buffer area hit
Step S12 is executed, if buffer area miss thens follow the steps S2.
Step S12: the page P in setting hit buffer areaiIn request set be ASiAs following equation indicates:
ASi={ AS1,AS2,...,ASk,...,ASo},1≤k≤o≤n
ASiFor to page PiThe set of request, wherein o represents request sum, and k represents request number.Judgement hit is slow
Deposit area page PiIn workspace or exchange area, if PiS13 is thened follow the steps in workspace, if PiIt is then executed in exchange area
Step S2.Page-hit and transfer figure are as shown in Figure 2.It include the logical address and read-write mode in request in upper layer request,
Judge whether request page is located at buffer area and updates buffer area in Hash bucket by hash function.It is divided into work in buffer area
Area and exchange area, the page is divided into the hot clean page (HC), hot containing dirty pages face (HD), the cold clean page (CC), cold containing dirty pages in buffer area
Four kinds of face (CD), according to different page types and present position is read or the superseded page.
Step S13: the region MRU that the page will be transferred in workspace updates the page status in workspace, sign-off sheet
It requests in person and asks.
Step S2: judgement insertion buffer state determines the superseded page using the LRU, method based on dynamic page weight,
Buffer state is adjusted, page request is executed.The step further comprises:
Step S21: judging whether page request R is read request, if not read request thens follow the steps S22, asks if reading
It asks, read data in a flash memory and executes step S25.
Step S22: judging whether workspace has expired, and workspace completely thens follow the steps S23, and workspace is less than then by the page
It is directly added into workspace and executes step S24.
Step S23: being transferred to the region LRU for the superseded page in exchange area, if the page is thened follow the steps from exchange area
S24 is directly added into exchange area if the page is from flash memory and executes step S25.Workspace is handed over interval censored data is exchanged in flash memory
As shown in Figure 3, flash translation layer (FTL) retains page data with page address to mutual schematic diagram, and the superseded page in workspace is transferred to
It is eliminated behind the region MRU to flash memory, the superseded page of exchange area is eliminated after being transferred to the region LRU to flash memory.
Page selection process comprises the steps of:
Step S231: judging whether the page comes from exchange area, is, directly returns to the region LRU in workspace and executes step
Otherwise rapid S234 is carried out in next step.
Step S232: the power of every page in exchange area Page Range [0, w] is calculated according to dynamic page weight more new algorithm
Weight Pweight, and generate MapweightSave updated every page weight.Wherein w is page total size, PweightIt represents dynamic
The weight of every page, Map in state page weight algorithmweightFor dynamic page weight map.
Step S233: traversal MapweightIn every page, find and wherein the smallest one page of weight and be denoted as the superseded page,
It returns and eliminates page address.
Step S234: judging request page source, if request page thens follow the steps S24 from exchange area;If requested page
Face then expels in exchange area from flash memory and eliminates the page, and request page is inserted into execution step S25 in exchange area.
Step S24: calling dynamic page weight to select to eliminate the page in workspace, and the selected superseded page is expelled best friend
Area is changed, and the page need to be added, workspace is added.
Dynamic page Weight algorithm comprises the steps of:
Step S241: initialization sets target pages as Pi, request page address is Rpid, the kth request of request page
ForS is current all request RiSet.Page, which is calculated, according to following equation works as front PiTime limitation
TLI is set as PiThe equispaced of arrival can be evaluated whether P by TLIiTime locality, be based onIt can be pre-
The arrival time of other pages in page set U is surveyed,Be expressed as follows column formula:
Wherein m is buffer area page sum, and o is current page request number.
Step S242: read operation time-consuming LrWith write operation time-consuming LwBe expressed as follows column formula:
Wherein g represents current operation number, and q represents the Action number finally accessed.
Page delay degree is calculated using different algorithms for the clean page (CC, HC) and containing dirty pages face (CD, HD)
Set ASlatestFor PiIn the request signal that finally reaches, ASoFor the request signal currently reached, page freshnessSuch as following formula:
Pass through page freshnessPage delay degreeAnd time limitationWork is calculated according to following equation
Make area page PiDynamic page weight
Wherein α is setting weight coefficient, PamFor page read-write state, o is current page request number, and m is page sum,
Q is recent visit operation.
The dynamic page weight of all pages is calculated in workspace, acquires minimum value
Step S243: willIt is denoted as the superseded page and expels to exchange area, execute step (8).
Step S25: current page requests RjExecution terminates, and executes next request in page request queue S.
This method and the performance of other traditional LRU, methods are compared below:
Assessment of the invention is evaluated by the way of synthetic workload, experimental configuration is Intel's Duo
I7-8700K 4.3GHz processor, 32GB 3200mhz RAM and 64 Windows 10Pro operating systems.The present invention exists
The nand flash memory for simulating 128MB is carried out on Visual Studio 2015, the block size of simulation flash equipment is
128KB, every piece has page 64, and erasing delay is set as 1.5ms, and the duration is set as 100000P/E, read-write delay ratio setting
It is 1/8.Table 1 gives the detailed configuration of nand flash memory, and compares with physical device configuration.
The configuration of 1 nand flash memory of table
In order to simulate the access module in practical flash memory, experiment generates 4 synthesis paths based on zipf distribution.In table 2
The detailed features of four synthesis paths are described, wherein read-write ratio indicates to read and write proportion in all requests, read and write specific gravity
Indicate the read-write operation ratio executed to certain one page.The road of four paths in these cases is respectively shown in 4~7 in figure
As a result, the experimental results showed that after 2.5 ten thousand operations, the read-write operation of all tracking all becomes stable diameter.
2 synthesis path feature of table
In experimental evaluation, we compared other current state-of-the-art algorithms such as LRU, CF-LRU, LRU, AD-LRU, right
Dynamic page weight has carried out comprehensive analysis.The experimental results showed that, distinguished by using the clean page and containing dirty pages face in Fig. 8
Algorithm, compared with traditional LRU, method, hit rate of the invention about 8.3%;In Fig. 9 the experimental results showed that, by using work
Area with exchange section page migration method, compared with traditional LRU, method, the present invention, which writes consumption, reduces about 22.6%;In Figure 10
The experimental results showed that distinguishing algorithm by using the cold and hot page, compared with traditional LRU, method, delay time of the present invention is reduced
About 18.8%.
In the patent, we have proposed the buffer area the LRU replacement methods based on dynamic page weight, effective with holding one
Buffer area, improve the overall performance of flash memory.Lru algorithm based on dynamic page weight is local using the access of flash equipment
Property, according to access mode and the frequency separation page.Buffering divides into two regions as workspace and exchange area, and the page is divided into four
Seed type.Different from the common buffer area replacement method based on LRU, it is special that the LRU, method based on dynamic page weight has
Conversion of page method.A kind of new dynamic page Weight algorithm is proposed, which considers temporal locality, operating cost
With page timeliness to effectively improving buffering hit rate, reducing time delay and writing consumption.
The above description of the embodiment is only used to help understand the method for the present invention and its core ideas.It should be pointed out that pair
For those skilled in the art, without departing from the principle of the present invention, the present invention can also be carried out
Some improvements and modifications, these improvements and modifications also fall within the scope of protection of the claims of the present invention.
The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present invention.
Various modifications to these embodiments will be readily apparent to those skilled in the art, as defined herein
General Principle can be realized in other embodiments without departing from the spirit or scope of the present invention.Therefore, of the invention
It is not intended to be limited to the embodiments shown herein, and is to fit to and the principles and novel features disclosed herein phase one
The widest scope of cause.
Claims (1)
1. a kind of LRU flash cache management method based on dynamic page weight, which comprises the following steps:
Step S1: page request in read request queue is simultaneously identified and is classified to page request type and locating region;
Step S2: judgement insertion buffer state determines to eliminate the page, adjustment using the LRU, method based on dynamic page weight
Buffer state executes page request;
Wherein, step S1 further will include:
Step S11: pre-processing page request queue, judges whether the corresponding page of page request is stored in buffer area
In, if buffer area hit thens follow the steps S12, if buffer area miss thens follow the steps S2;Step S11 further comprises:
Step S111: page request queue is pre-processed, and sets page request as R, wherein including request number RpidAnd page
Face request mode Ram∈ { read, write }, request queue S can be indicated with following equation:
S={ R1,R2,...,Rj,...,Rn},1≤j≤n
Wherein, request queue S is multiple page request RpidSet, n represent the request in request queue S sum, j represent
Page request number;
Step S112: setting the page as P, page set U={ P1,P2,...,Pi,...,Pm},1≤i≤m≤n;Wherein m is represented
Page sum in page set U, i represent page number;Judge the page request R in request queue SjThe corresponding page
PiIt is no to save in the buffer, if buffer area hit thens follow the steps S12, if buffer area miss thens follow the steps S2;
Step S12: hit buffer area page P is setiPage request set ASi={ AS1,...,ASk,...,ASo};Wherein o generation
Table request sum, k represent request number;
Judgement hit buffer area page PiIn workspace or exchange area, if PiS13 is thened follow the steps in workspace, if PiPlace
S2 is thened follow the steps in exchange area;
Step S13: the page will be transferred to region most-often used in workspace, and the content of pages of workspace is according to most-often used
Page algorithm is adjusted, ending request page algorithm;
The step S2 further comprises:
Step S21: judging whether the page has read, if unread then follow the steps S22 if the page has been read in flash memory
Middle reading data simultaneously execute step S23;
Step S22: judging whether workspace has expired, and workspace completely thens follow the steps S23, and workspace is less than then direct by the page
Workspace is added and executes step S25;
Step S23: it calls to delete using renewal of the page algorithm recently in exchange area and eliminates the page, if the page is from exchange area
Step S24 is executed, be directly added into exchange area if the page is from flash memory and executes step S25;Step S23 further comprises:
Step S231: judging whether the page comes from exchange area, is, directly returns to the region the LRU parallel step in workspace
Otherwise S234 carries out S232;
Step S232: the weight of every page in exchange area Page Range [0, w] is calculated according to dynamic page weight more new algorithm
Pweight, and generate MapweightSave updated every page weight;Wherein w is page total size, PweightRepresent dynamic page
The weight of every page, Map in the Weight algorithm of faceweightFor dynamic page weight map;
Step S233: traversal MapweightIn every page, find and wherein the smallest one page of weight and be denoted as the superseded page, return
Eliminate page address;
Step S234: judging request page source, if request page thens follow the steps S24 from exchange area;If request page is come
It is then expelled in exchange area from flash memory and eliminates the page, and request page is inserted into execution step S25 in exchange area;
Step S24: it calls dynamic page weight to select to eliminate the page in workspace, the selected superseded page is expelled to exchange
Area, and the page need to be added, workspace is added, wherein the step S24 further comprises:
Step S241: initialization sets target pages as Pi, request page address is Rpid, request page kth request beS is current all request RiSet;Page, which is calculated, according to following equation works as front PiTime limitation Pi TLI:
TLI is set as PiThe equispaced of arrival can be evaluated whether P by TLIiTime locality, be based on Pi TLIIt can predict page
The arrival time of other pages in the set U of face,Be expressed as follows column formula:
Wherein m is buffer area page sum, and o is current page request number;
Step S242: read operation time-consuming LrWith write operation time-consuming LwBe expressed as follows column formula:
Wherein g represents current operation number, and q represents the Action number finally accessed;
Page delay degree P is calculated using different algorithms for the clean page (CC, HC) and containing dirty pages face (CD, HD)i EC:
Set ASlatestFor PiIn the request signal that finally reaches, ASoFor the request signal currently reached, page freshness Pi RESuch as
Following formula:
ASo≤ASlatest
By acquiring page freshness Pi RE, page delay degree Pi ECAnd time limitation Pi TLI, according to following equation calculating work space
Page PiDynamic page weight Pi w:
Wherein α is setting weight coefficient, PamFor page read-write state, o is current page request number, and m is page sum, and q is
Recent visit operation;
The dynamic page weight of all pages is calculated in workspace, acquires minimum value
Step S243: willIt is denoted as the superseded page and expels to exchange area, execute step S25;
Step S25: current page request, which executes, to be terminated, and next request in page request queue is executed.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811394983.6A CN109857680B (en) | 2018-11-21 | 2018-11-21 | LRU flash memory cache management method based on dynamic page weight |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811394983.6A CN109857680B (en) | 2018-11-21 | 2018-11-21 | LRU flash memory cache management method based on dynamic page weight |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109857680A true CN109857680A (en) | 2019-06-07 |
CN109857680B CN109857680B (en) | 2020-09-11 |
Family
ID=66890140
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811394983.6A Active CN109857680B (en) | 2018-11-21 | 2018-11-21 | LRU flash memory cache management method based on dynamic page weight |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109857680B (en) |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103257935A (en) * | 2013-04-19 | 2013-08-21 | 华中科技大学 | Cache management method and application thereof |
CN104145252A (en) * | 2012-03-05 | 2014-11-12 | 国际商业机器公司 | Adaptive cache promotions in a two level caching system |
CN104699422A (en) * | 2015-03-11 | 2015-06-10 | 华为技术有限公司 | Determination method and determination device of cache data |
CN106527966A (en) * | 2015-09-10 | 2017-03-22 | 蜂巢数据有限公司 | Garbage collection in ssd drives |
CN107391398A (en) * | 2016-05-16 | 2017-11-24 | 中国科学院微电子研究所 | A kind of management method and system in flash cache area |
CN107590084A (en) * | 2017-08-22 | 2018-01-16 | 浙江万里学院 | A kind of page level buffering area improved method based on classification policy |
US20180095896A1 (en) * | 2013-04-03 | 2018-04-05 | International Business Machines Corporation | Maintaining cache consistency in a cache for cache eviction policies supporting dependencies |
US20180300257A1 (en) * | 2016-05-31 | 2018-10-18 | International Business Machines Corporation | Using an access increment number to control a duration during which tracks remain in cache |
US20180314432A1 (en) * | 2016-06-06 | 2018-11-01 | International Business Machines Corporation | Invoking input/output (i/o) threads on processors to demote tracks from a cache |
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 |
CN108845957A (en) * | 2018-03-30 | 2018-11-20 | 杭州电子科技大学 | It is a kind of to replace and the adaptive buffer management method of write-back |
-
2018
- 2018-11-21 CN CN201811394983.6A patent/CN109857680B/en active Active
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104145252A (en) * | 2012-03-05 | 2014-11-12 | 国际商业机器公司 | Adaptive cache promotions in a two level caching system |
US20180095896A1 (en) * | 2013-04-03 | 2018-04-05 | International Business Machines Corporation | Maintaining cache consistency in a cache for cache eviction policies supporting dependencies |
CN103257935A (en) * | 2013-04-19 | 2013-08-21 | 华中科技大学 | Cache management method and application thereof |
CN104699422A (en) * | 2015-03-11 | 2015-06-10 | 华为技术有限公司 | Determination method and determination device of cache data |
CN106527966A (en) * | 2015-09-10 | 2017-03-22 | 蜂巢数据有限公司 | Garbage collection in ssd drives |
CN107391398A (en) * | 2016-05-16 | 2017-11-24 | 中国科学院微电子研究所 | A kind of management method and system in flash cache area |
US20180300257A1 (en) * | 2016-05-31 | 2018-10-18 | International Business Machines Corporation | Using an access increment number to control a duration during which tracks remain in cache |
US20180314432A1 (en) * | 2016-06-06 | 2018-11-01 | International Business Machines Corporation | Invoking input/output (i/o) threads on processors to demote tracks from a cache |
CN107590084A (en) * | 2017-08-22 | 2018-01-16 | 浙江万里学院 | A kind of page level buffering area improved method based on classification policy |
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 |
CN108845957A (en) * | 2018-03-30 | 2018-11-20 | 杭州电子科技大学 | It is a kind of to replace and the adaptive buffer management method of write-back |
Non-Patent Citations (1)
Title |
---|
刘恒: "并行计算框架Spark中一种新的RDD分区权重缓存替换算法", 《小型微型计算机系统》 * |
Also Published As
Publication number | Publication date |
---|---|
CN109857680B (en) | 2020-09-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107193646B (en) | High-efficiency dynamic page scheduling method based on mixed main memory architecture | |
Jin et al. | AD-LRU: An efficient buffer replacement algorithm for flash-based databases | |
CN104572491B (en) | A kind of read buffer management method and device based on solid state hard disc | |
Wang et al. | Adaptive placement and migration policy for an STT-RAM-based hybrid cache | |
EP3414665B1 (en) | Profiling cache replacement | |
Jiang et al. | S-FTL: An efficient address translation for flash memory by exploiting spatial locality | |
US8601216B2 (en) | Method and system for removing cache blocks | |
CN106528454B (en) | A kind of memory system caching method based on flash memory | |
CN108762664B (en) | Solid state disk page-level cache region management method | |
Lv et al. | Operation-aware buffer management in flash-based systems | |
CN103795781B (en) | A kind of distributed caching method based on file prediction | |
CN104794064A (en) | Cache management method based on region heat degree | |
CN103136121A (en) | Cache management method for solid-state disc | |
CN110888600B (en) | Buffer area management method for NAND flash memory | |
CN106569732B (en) | Data migration method and device | |
CN108845957B (en) | Replacement and write-back self-adaptive buffer area management method | |
CN106775466A (en) | A kind of FTL read buffers management method and device without DRAM | |
Wu et al. | APP-LRU: A new page replacement method for PCM/DRAM-based hybrid memory systems | |
CN100428193C (en) | Data preacquring method for use in data storage system | |
CN110297787A (en) | The method, device and equipment of I/O equipment access memory | |
Sun et al. | Co-active: A workload-aware collaborative cache management scheme for NVMe SSDs | |
Wang et al. | A novel buffer management scheme based on particle swarm optimization for SSD | |
CN110795363A (en) | Hot page prediction method and page scheduling method for storage medium | |
CN109478164A (en) | For storing the system and method for being used for the requested information of cache entries transmission | |
CN102521161B (en) | Data caching method, device and server |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right |
Effective date of registration: 20240112 Address after: 230000 floor 1, building 2, phase I, e-commerce Park, Jinggang Road, Shushan Economic Development Zone, Hefei City, Anhui Province Patentee after: Dragon totem Technology (Hefei) Co.,Ltd. Address before: 310018 Xiasha Higher Education Zone, Hangzhou, Zhejiang Patentee before: HANGZHOU DIANZI University |
|
TR01 | Transfer of patent right |