CN103246616B - A kind of globally shared buffer replacing method of access frequency within long and short cycle - Google Patents
A kind of globally shared buffer replacing method of access frequency within long and short cycle Download PDFInfo
- Publication number
- CN103246616B CN103246616B CN201310195427.7A CN201310195427A CN103246616B CN 103246616 B CN103246616 B CN 103246616B CN 201310195427 A CN201310195427 A CN 201310195427A CN 103246616 B CN103246616 B CN 103246616B
- Authority
- CN
- China
- Prior art keywords
- data
- access
- buffer
- long
- term
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 14
- 230000007774 longterm Effects 0.000 claims abstract description 27
- 238000012545 processing Methods 0.000 claims abstract description 25
- 238000009826 distribution Methods 0.000 claims abstract description 13
- 230000000737 periodic effect Effects 0.000 claims abstract description 8
- 238000011282 treatment Methods 0.000 claims abstract description 7
- 230000007246 mechanism Effects 0.000 claims description 10
- 238000013507 mapping Methods 0.000 claims description 9
- 238000003860 storage Methods 0.000 claims description 9
- 238000006243 chemical reaction Methods 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 4
- 230000006870 function Effects 0.000 claims description 4
- 238000004891 communication Methods 0.000 claims description 3
- 238000013523 data management Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 claims description 3
- 238000007726 management method Methods 0.000 claims description 3
- 238000005096 rolling process Methods 0.000 claims description 3
- 230000007423 decrease Effects 0.000 claims description 2
- 238000011160 research Methods 0.000 claims description 2
- 230000015556 catabolic process Effects 0.000 claims 1
- 238000006731 degradation reaction Methods 0.000 claims 1
- 238000000819 phase cycle Methods 0.000 claims 1
- 238000004321 preservation Methods 0.000 claims 1
- 238000010977 unit operation Methods 0.000 claims 1
- 230000008569 process Effects 0.000 abstract description 3
- 230000005540 biological transmission Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 238000004883 computer application Methods 0.000 description 1
- 239000008358 core component Substances 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 238000002955 isolation Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000009897 systematic effect Effects 0.000 description 1
Abstract
The present invention provides a kind of globally shared buffer replacing method of access frequency within long and short cycle, and host side mainly has two kinds of access modes to data:Long-term cycle access and short-term frequently access, long-term cycle access refers to data block in the longer time of interval by periodic repeated accesses, and it is traditional the data access of this type can not be recognized based on burin-in process replacement policy, cached when causing to be buffered in the such access of reply entirely ineffective.In order to solve the above-mentioned technical problem, the invention provides a kind of globally shared buffer replacing method for realizing access frequency within long and short cycle, this method includes map manager, caching replacement processing, cache distribution processing, five modules of mirror image processing and cache consistency treatments.
Description
Technical field
The present invention relates to computer application field, the globally shared caching of specifically a kind of access frequency within long and short cycle
Replacement method.
Background technology
Digital information on network is in explosive increase, and to the year two thousand twenty, global metadata amount will be enlarged by 50 times, simultaneously number
According to presenting at a high speed, multifarious feature, data diversity is mainly due to new many structured datas, and including network day
The data types such as will, social media, internet hunt, mobile phone call history and sensor network are caused.While the number of host side
Amount is also presented level several levels and increased, and different host sides is square in response time, transmission bandwidth, transmission granularity etc. to I/O process demands
Face is far different, along with the generation of these demands, and multi-control disk array is arisen at the historic moment, and multi-control disk array is used with the overall situation
Cache as core, the architecture interconnected between controller, buffer unit by cross matrix switch.Buffer unit is single thing
Part is managed, clog-free, high band wide data transmission with each controller are realized by the Crossbar link exchanges of redundancy.It is all slow
Memory cell constitutes globally shared caching, and visible to all controllers, and each controller has the control of equality to it.It is global
How shared buffer memory is managed optimization to meet different business need as the core component of multi-control disk array to it
One of systematic function critical bottleneck will be turned into by asking.
The content of the invention
It is an object of the invention to provide a kind of globally shared buffer replacing method of access frequency within long and short cycle.
The purpose of the present invention realized in the following manner, and block-based visiting frequency value is replaced to page, simultaneously
Long period and short-period type of data access have been taken into account, cache hit rate is effectively improved, system includes map manager
Module, caches replacement processing module, caching distribution processing module, mirror image processing module and cache consistency treatments module five
Module is wherein:
Map manager module, for the mapping of local memory and shared virtual memory address space, its function be by
Local memory is mapped in shared virtual memory address space and maintains the uniformity of data;
Replacement processing module is cached, the replacement for globally shared caching is handled, during address of cache, if in the page
The middle discovery page to be accessed then produces page faults not in internal memory, and when generation page faults, operating system must be
Internal memory selects a page to be moved out internal memory, to concede space for the page that will call in;
Caching distribution processing module, using two-stage memory management scheme, row address is entered using the mode of multi-level page-table and is turned
Change, tried one's best uniformly for the distribution that makes shared drive, the distribution of system shared drive is carried out by the way of cycle assignment;
Data, in order to prevent single-unit point failure, in reading and writing data, are written to two by mirror image processing module simultaneously
In independent region of memory, effectively prevent any node causes the loss of data in caching due to chance failure;
Cache consistency treatment modules, the implementation based on catalogue ensures same data in the multiple nodes of system
Multiple copies it is consistent, it is that every page in internal memory maintains a directory entry, and the directory entry, which is recorded, all currently holds this copy
Node and data block status information, include whether to be written, if exclusive or be total to by multiple nodes by some node
Enjoy, when a node is intended to toward a certain piece of write-in data and when data may be caused inconsistent, it just according to the content of catalogue to
Those nodes for holding the backup of this copy send nullified or more new signal to maintain the uniformity of data.
The globally shared caching system operation principles of access frequency within long and short cycle:
1)As shown in Fig. 2 all buffer units are organized into a global buffer pond, buffer unit because failure or liter
Cache pool is left or added to the reason for level, and system is realized using cache resources discovery mechanism and seamless increased or decreased global buffer
Capacity, it is ensured that storage service will not be influenceed by resource change, buffer unit passes through the active of periodic broadcasting resource declarations message
The availability oneself partly or completely cached is declared, new buffer unit is wide to all controllers, buffer unit when adding
Resource declarations message is broadcast, controller adds this element in available cache memory resource index after receiving;
2)As illustrated in fig. 2, it is assumed that six buffer units of A, B, C, D, E, F are included in common buffer pool, in order to ensure data
High Availabitity, caching mirror image is using annular High Availabitity strategy, i.e., every part data cached to be all written in two buffer units, have
The mirror image order of body is annular strategy.The business of storage system carrying is isolated by rolling up, and volume is to carry out Service Data Management
Base unit, be the buffer unit that each volume sets an owner type, certain is rolled up business datum all of the above and can be buffered
To owner buffer unit and the buffer unit backward buffer unit on.Assuming that volume 1 owner be buffer unit A, then its
Data are cached to buffer unit A and buffer unit B;If its owner is B, its is data cached be cached to buffer unit B and
Buffer unit C, by that analogy, so as to form data cached annular High Availabitity strategy.
3)It is globally shared to be cached with three kinds of operational modes, it is normal mode, degraded mode and direct write pattern respectively, caching is single
First redundancy scheme uses the state-detection mechanism of the software and hardware combining in multi-controller redundancy scheme to realize, reliable in order to improve
Property, its buffer service only will be taken over by the forward direction node in annular High Availabitity after the failure of certain buffer unit, while cache pool enters
Degraded mode, in normal mode and degraded mode, global buffer pond uses and writes back pattern data cached write-in back-end physical
Disk, and tolerate that any one buffer unit breaks down in synchronization;Under extreme case, one buffer unit of system spare
During operation, globally shared caching will be taken under data direct write pattern, direct write pattern, and buffer unit failure will not be to data integrity
There is any influence with uniformity;
4)Buffer unit equally includes logical address mapping mechanism and long-distance inner access protocol, realizes and cache client
The logical address mapping and data communication of module, buffer unit are divided into index area, three parts in data field and MIRROR SITE, indexed
Area is used for carrying out image and the conversion of data, preserves all data field pages of this buffer unit, the index of the MIRROR SITE page, number
The read-write cache data of storage service are deposited according to area, MIRROR SITE deposits the mirror image data that other controllers are sended over, Yi Jiqi
The index data of his buffer unit, realizes read-write data, the full mirror image of index data;
Access frequency within long and short cycle replacement method and step:
1)Host side has two kinds of access modes to data:Long-term cycle access and short-term frequently access, long-term cycle access
Refer to that data block is being spaced in longer time by periodic repeated accesses, it is short-term frequent using including daily record, timed backup
Access refers to that data block is frequently accessed in the short period of time, should be tried one's best in cache hit rate, global buffer to maximize
Preserve the data that are accessed frequently in short term, if but the short-term data frequently accessed do not accessed always, then they also should be first
Global buffer is replaced out in the data of long-term cycle access, this is accomplished by taking into account long-term and short-term access situation, research
Global buffer replacement processing based on shot and long term visiting frequency, as shown in figure 1, increasing data access for the data block of global buffer
Frequency.
2)Each byte data block access frequency value is arranged from high to low, is represented in the cycle of interval from short to long, number
Situation is accessed according to block, the value of n-th of byte i-th bit from high to low is represented with Bn [i], if in unit interval T, data block
Accessed, then it is 1 to put B0 [0], and it is 0 otherwise to put B0 [0], every the 8n*T times, using following Policy Updates data block access frequency
Angle value:
Bn[i+1]= Bn [i] (i=0,1,……,6)
Bn +1[i+1]= Bn +1[i] (i=0,1,……,6)
Bn +1[0]= Bn [0] || Bn [1] || … … || Bn [6] || Bn [7]
Above rule ensures, if access module is long-term cycle access, then will have the multidigit to be continuously in low byte
1, if access module frequently accesses to be short-term, then when will have the multidigit to be continuously 1 in upper byte, and low byte can be with few
Probability occur multidigit continuously be 1 situation;
3)When replacing the data in global buffer, select data block visiting frequency value is minimum to be replaced every time, this
It is that there is larger visiting frequency value because of the short-term data frequently accessed, if but be not accessed within longer one period, it is short
The data access frequency value that phase frequently accesses will gradually be less than the data of long-term cycle access, so can guarantee that meeting in global buffer
The short-term data frequently accessed are preserved as far as possible, if but be not accessed in longer one period, they will be prior to long-term cycle access
Data and be replaced.
Brief description of the drawings
Fig. 1 is data access frequency value schematic diagram;
Fig. 2 is globally shared cache pool topological diagram.
Embodiment
Technical scheme is set forth in below in conjunction with accompanying drawing and example.
Host side mainly has two kinds of access modes to data:Long-term cycle access and short-term frequently access, the long-term cycle visit
Ask and refer to that data block is being spaced in longer time by periodic repeated accesses, and it is traditional based on burin-in process replacement policy
The data access of this type can not be recognized, is cached when causing to be buffered in the such access of reply entirely ineffective.
In order to solve the above-mentioned technical problem, the globally shared slow of access frequency within long and short cycle is realized the invention provides a kind of
Deposit replacement method, including map manager, caching replacement processing, caching distribution processing, mirror image processing and cache consistency treatments
Five modules, wherein:
Map manager module, for the mapping of local memory and shared virtual memory address space, its function be by
Local memory is mapped in shared virtual memory address space and maintains the uniformity of data.
Replacement processing module is cached, the replacement for globally shared caching is handled, during address of cache, if in the page
The middle discovery page to be accessed then produces page faults not in internal memory.When occurring page faults, operating system must be
Internal memory selects a page to be moved out internal memory, and to concede space for the page that will call in, this replacements processing is based on data
Access type, identification long period data and the data that frequently access of short cycle, according to this strategy carry out caching replacement.
Distribution processing is cached, using two-stage memory management scheme, address conversion is carried out using the mode of multi-level page-table.For
The distribution of shared drive is set to try one's best uniformly, the distribution of system shared drive is carried out by the way of cycle assignment.
Data, in order to prevent single-unit point failure, in reading and writing data, are written to two by mirror image processing module simultaneously
In independent region of memory, effectively prevent any node causes the loss of data in caching due to chance failure.
Cache consistency treatment modules, the implementation based on catalogue ensures same data in the multiple nodes of system
Multiple copies it is consistent, it is that every page in internal memory maintains a directory entry, and the directory entry, which is recorded, all currently holds this copy
Node and the status information of data block (such as whether being written, monopolized by some node or by multiple nodes sharings
Deng).When a node is intended to write number and when data may be caused inconsistent toward a certain piece, it just can according to the content of catalogue only to
Those nodes for holding the backup of this copy send nullified or more new signal to maintain the uniformity of data.
All buffer units are organized into a global buffer pond.Buffer unit is because the reason such as failure or upgrading is left
Or cache pool is added, system uses cache resources discovery mechanism, realizes and seamless increases or decreases global buffer capacity, it is ensured that storage
Service will not be influenceed by resource change.Buffer unit by periodic broadcasting resource declarations message actively declare oneself part or
The availability of the whole caching of person.Message is stated to all controllers, buffer unit broadcast resource when new buffer unit is added,
Controller adds this element in available cache memory resource index after receiving.Assuming that in common buffer pool comprising A, B, C, D, E,
Six buffer units of F, in order to ensure data High Availabitity, caching mirror image using annular High Availabitity strategy, i.e., every part it is data cached all
It can be written in two buffer units, specific mirror image order is annular strategy.The business of storage system carrying is by involving in
Row isolation, volume is the base unit for carrying out Service Data Management, is the buffer unit that each volume sets an owner type, certain
Rolling up business datum all of the above can be cached on the backward buffer unit of owner buffer unit and the buffer unit.It is false
If the owner of volume 1 is buffer unit A, then its data is cached to buffer unit A and buffer unit B;If its owner is B,
Its is data cached to be cached to buffer unit B and buffer unit C, by that analogy, so as to form data cached annular High Availabitity
Strategy.
It is globally shared to be cached with three kinds of operational modes, it is normal mode, degraded mode and direct write pattern respectively.Buffer unit
Redundancy scheme uses the state-detection mechanism of the software and hardware combining in multi-controller redundancy scheme to realize.In order to improve reliability,
Its buffer service only will be taken over by the forward direction node in annular High Availabitity after the failure of certain buffer unit, degraded while cache pool enters
Pattern.In normal mode and degraded mode, global buffer pond writes back-end physical disk using the pattern that writes back data cached,
And it can tolerate that any buffer unit breaks down in synchronization;Under extreme case, one buffer unit fortune of system spare
During row, globally shared caching will take data direct write pattern.Under direct write pattern, buffer unit failure will not to data integrity and
Uniformity has any influence.
Buffer unit equally includes logical address mapping mechanism and long-distance inner access protocol, realizes and cache client mould
The logical address mapping and data communication of block.Buffer unit is divided into index area, three parts in data field and MIRROR SITE.Index area
For carrying out image and the conversion of data, all data field pages of this buffer unit, the index of the MIRROR SITE page are preserved.Data
Deposit the read-write cache data of storage service in area.MIRROR SITE deposits the mirror image data that other controllers are sended over, and other
The index data of buffer unit, realizes read-write data, the full mirror image of index data.
Host side mainly has two kinds of access modes to data:Long-term cycle access and short-term frequently access.The long-term cycle visits
Ask and refer to that data block is being spaced in longer time by periodic repeated accesses, typical case's application includes daily record, timed backup etc..
Short-term frequently access refers to that data block is frequently accessed in the short period of time.To maximize cache hit rate, global buffer
In should try one's best and preserve the data that are accessed frequently in short term, if but the short-term data frequently accessed do not accessed always, then it
Also should be replaced out global buffer prior to the data of long-term cycle access.This is accomplished by taking into account long-term and short-term access feelings
Condition, studies the global buffer replacement processing based on shot and long term visiting frequency, as shown in figure 1, being the data block increase of global buffer
Data access frequency,
Each byte data block access frequency value is arranged from high to low, is represented in the cycle of interval from short to long, data
Block is accessed situation.The value of n-th of byte i-th bit from high to low is represented with Bn [i], if in unit interval T, data block quilt
Access, then it is 1 to put B0 [0], it is 0 otherwise to put B0 [0].Every the 8n*T times, using following Policy Updates data block visiting frequency
Value:
Bn[i+1]= Bn [i] (i=0,1,……,6)
Bn +1[i+1]= Bn +1[i] (i=0,1,……,6)
Bn +1[0]= Bn [0] || Bn [1] || … … || Bn [6] || Bn [7]
Above rule can ensure, if access module is long-term cycle access, then will have multidigit company in low byte
Continue for 1, if access module frequently accesses to be short-term, then when there will be the multidigit to be continuously 1 in upper byte, and low byte can be with
There is the situation that multidigit is continuously 1 in few probability.
When replacing the data in global buffer, select data block visiting frequency value is minimum to be replaced every time.This is
Because the short-term data frequently accessed have larger visiting frequency value, if but be not accessed within longer one period, in short term
The data access frequency value frequently accessed will gradually be less than the data of long-term cycle access.It can so ensure use up in global buffer
Amount preserves the short-term data frequently accessed, if but be not accessed in longer one period, they will be prior to long-term cycle access number
According to and be replaced.
It is the known technology of those skilled in the art in addition to the technical characteristic described in specification.
Claims (1)
1. a kind of globally shared buffer replacing method of access frequency within long and short cycle, it is characterised in that block-based visiting frequency
Value is replaced to page, while having taken into account long period and short-period type of data access, is effectively improved cache hit
Rate, system include map manager module, cache replacement processing module, caching distribution processing module, mirror image processing module and
Five modules of cache consistency treatments module, wherein:
Map manager module, for the mapping of local memory and shared virtual memory address space, its function is will be local
Internal memory is mapped in shared virtual memory address space and maintains the uniformity of data;
Replacement processing module is cached, the replacement for globally shared caching is handled, during address of cache, if being sent out in the page
The existing page to be accessed then produces page faults not in internal memory, and when occurring page faults, operating system must be in internal memory
One page of selection is moved out internal memory, to concede space for the page that will call in;
Caching distribution processing module, using two-stage memory management scheme, carries out address conversion using the mode of multi-level page-table, is
The distribution of shared drive is set to try one's best uniformly, the distribution of system shared drive is carried out by the way of cycle assignment;
Data, in order to prevent single-unit point failure, in reading and writing data, are written to two independences by mirror image processing module simultaneously
Region of memory in, effectively prevent any node caused due to chance failure caching in loss of data;
Cache consistency treatment modules, the implementation based on catalogue ensures the multiple of same data in the multiple nodes of system
Copy it is consistent, it is that every page in internal memory maintains a directory entry, and the directory entry records all sections for currently holding this copy
The status information of point and data block, includes whether to be written, if exclusive or by multiple nodes sharings by some node, when
One node is intended to toward a certain piece of write-in data and when data may be caused inconsistent, and it is just according to the content of catalogue to holding this
Those nodes of copy backup send nullified or more new signal to maintain the uniformity of data, access frequency within long and short cycle it is complete
Office's shared buffer memory system operation principle:
1)All buffer units are organized into a global buffer pond, buffer unit because leave the reason for failure or upgrading or
Cache pool is added, system uses cache resources discovery mechanism, realizes and seamless increases or decreases global buffer capacity, it is ensured that storage clothes
Business will not be influenceed by resource change, buffer unit by periodic broadcasting resource declarations message actively declare oneself partly or
The availability of whole cachings, message, control are stated when new buffer unit is added to all controllers, buffer unit broadcast resource
Device processed adds this element in available cache memory resource index after receiving;
2)Assuming that including six buffer units of A, B, C, D, E, F in common buffer pool, in order to ensure data High Availabitity, mirror image is cached
Using annular High Availabitity strategy, i.e., every part data cached to be all written in two buffer units, and specific mirror image order is
Annular strategy, the business of storage system carrying is isolated by rolling up, and volume is the base unit for carrying out Service Data Management, is every
Individual volume sets the buffer unit of an owner type, and certain, which rolls up business datum all of the above, can be cached to owner caching list
On the backward buffer unit of member and the buffer unit, it is assumed that the owner of volume 1 is buffer unit A, then its data is cached to caching
Unit A and buffer unit B;If its owner is B, its is data cached to be cached to buffer unit B and buffer unit C, with such
Push away, so as to form data cached annular High Availabitity strategy;
3)It is globally shared to be cached with three kinds of operational modes, it is normal mode, degraded mode and direct write pattern respectively, buffer unit is superfluous
Remaining mechanism uses the state-detection mechanism of the software and hardware combining in multi-controller redundancy scheme to realize, in order to improve reliability, certain
Its buffer service only will be taken over by the forward direction node in annular High Availabitity after buffer unit failure, while cache pool enters degradation mould
Formula, in normal mode and degraded mode, global buffer pond uses the pattern that writes back data cached write-in back-end physical disk, and
And tolerate that any one buffer unit breaks down in synchronization;Under extreme case, during one buffer unit operation of system spare,
Globally shared caching will be taken under data direct write pattern, direct write pattern, buffer unit failure will not to data integrity with it is consistent
Property has any influence;
4)Buffer unit equally includes logical address mapping mechanism and long-distance inner access protocol, realizes and cache client module
Logical address mapping and data communication, buffer unit is divided into index area, three parts in data field and MIRROR SITE, and index area is used
To carry out image and the conversion of data, all data field pages of this buffer unit, the index of the MIRROR SITE page, data field are preserved
The read-write cache data of storage service are deposited, MIRROR SITE deposits the mirror image data that other controllers are sended over, and other delay
The index data of memory cell, realizes read-write data, the full mirror image of index data;
Access frequency within long and short cycle replacement method and step:
1)Host side has two kinds of access modes to data:Long-term cycle access and short-term frequently access, long-term cycle access refer to
Data block, by periodic repeated accesses, using including daily record, timed backup, is frequently accessed in short term in the longer time of interval
Refer to that data block is frequently accessed in the short period of time, to maximize the preservation that should be tried one's best in cache hit rate, global buffer
The data being accessed frequently in short term, if but the short-term data frequently accessed do not accessed always, then they also should be prior to length
The data of phase cycle access and be replaced out global buffer, this is accomplished by taking into account long-term with short-term access situation, and research is based on
The global buffer replacement processing of shot and long term visiting frequency, is the data block increase data access frequency of global buffer;
2)Each byte data block access frequency value is arranged from high to low, is represented in the cycle of interval from short to long, data block
Accessed situation, the value of n-th of byte i-th bit from high to low is represented with Bn [i], if in unit interval T, data block is interviewed
Ask, then it is 1 to put B0 [0], it is 0 otherwise to put B0 [0], every the 8n*T times, using following Policy Updates data block visiting frequency value:
Bn[i+1]= Bn [i] (i=0,1,……,6)
Bn +1[i+1]= Bn +1[i] (i=0,1,……,6)
Bn +1[0]= Bn [0] || Bn [1] || … … || Bn [6] || Bn [7]
Above rule ensures, if access module is long-term cycle access, then it is continuously 1 that will have multidigit in low byte, if
Access module frequently accesses to be short-term, then when will have the multidigit to be continuously 1 in upper byte, and low byte can be with few general
There is the situation that multidigit is continuously 1 in rate;
3)When replacing the data in global buffer, select data block visiting frequency value is minimum to be replaced every time, this be because
There is larger visiting frequency value for the data that frequently access in short term, if but not being accessed within longer one period, short-term frequency
The data access frequency value of numerous access will gradually be less than the data of long-term cycle access, and so can guarantee that in global buffer to try one's best
The short-term data frequently accessed are preserved, if but be not accessed in longer one period, they will be prior to long-term cycle access data
And be replaced.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310195427.7A CN103246616B (en) | 2013-05-24 | 2013-05-24 | A kind of globally shared buffer replacing method of access frequency within long and short cycle |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310195427.7A CN103246616B (en) | 2013-05-24 | 2013-05-24 | A kind of globally shared buffer replacing method of access frequency within long and short cycle |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103246616A CN103246616A (en) | 2013-08-14 |
CN103246616B true CN103246616B (en) | 2017-09-26 |
Family
ID=48926144
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310195427.7A Active CN103246616B (en) | 2013-05-24 | 2013-05-24 | A kind of globally shared buffer replacing method of access frequency within long and short cycle |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103246616B (en) |
Families Citing this family (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104268101B (en) * | 2014-09-22 | 2017-12-05 | 无锡城市云计算中心有限公司 | A kind of memory allocation method and device |
CN104331352B (en) * | 2014-11-19 | 2018-03-09 | 浪潮(北京)电子信息产业有限公司 | Detection method and device are read outside cache uniformity chip address band |
CN104461935B (en) * | 2014-11-27 | 2018-03-13 | 华为技术有限公司 | A kind of method, apparatus and system for carrying out data storage |
US10019373B2 (en) * | 2014-12-19 | 2018-07-10 | Mediatek Inc. | Memory management method for supporting shared virtual memories with hybrid page table utilization and related machine readable medium |
GB2539383B (en) * | 2015-06-01 | 2017-08-16 | Advanced Risc Mach Ltd | Cache coherency |
CN107810490A (en) * | 2015-06-18 | 2018-03-16 | 华为技术有限公司 | System and method for the buffer consistency based on catalogue |
WO2017117734A1 (en) * | 2016-01-06 | 2017-07-13 | 华为技术有限公司 | Cache management method, cache controller and computer system |
CN107291635B (en) * | 2017-06-16 | 2021-06-29 | 郑州云海信息技术有限公司 | Cache replacement method and device |
CN107329696B (en) * | 2017-06-23 | 2019-05-14 | 华中科技大学 | A kind of method and system guaranteeing data corruption consistency |
CN109376022B (en) * | 2018-09-29 | 2021-12-14 | 中国科学技术大学 | Thread model implementation method for improving execution efficiency of Halide language in multi-core system |
CN111158578B (en) * | 2018-11-08 | 2022-09-06 | 浙江宇视科技有限公司 | Storage space management method and device |
CN109582895A (en) * | 2018-12-04 | 2019-04-05 | 山东浪潮通软信息科技有限公司 | A kind of cache implementing method |
CN110750507B (en) * | 2019-09-30 | 2022-09-20 | 华中科技大学 | Persistent client caching method and system under global namespace facing DFS |
CN111273860B (en) * | 2020-01-15 | 2022-07-08 | 华东师范大学 | Distributed memory management method based on network and page granularity management |
WO2022021158A1 (en) * | 2020-07-29 | 2022-02-03 | 华为技术有限公司 | Cache system, method and chip |
CN112612727B (en) * | 2020-12-08 | 2023-07-07 | 成都海光微电子技术有限公司 | Cache line replacement method and device and electronic equipment |
CN115374046B (en) * | 2022-10-21 | 2023-03-14 | 山东云海国创云计算装备产业创新中心有限公司 | Multiprocessor data interaction method, device, equipment and storage medium |
CN117032596B (en) * | 2023-10-09 | 2024-01-26 | 苏州元脑智能科技有限公司 | Data access method and device, storage medium and electronic equipment |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1996944A (en) * | 2006-11-09 | 2007-07-11 | 华中科技大学 | A method for global buffer management of the cluster storage system |
CN101609432A (en) * | 2009-07-13 | 2009-12-23 | 中国科学院计算技术研究所 | Shared buffer memory management system and method |
CN102262512A (en) * | 2011-07-21 | 2011-11-30 | 浪潮(北京)电子信息产业有限公司 | System, device and method for realizing disk array cache partition management |
CN102609362A (en) * | 2012-01-30 | 2012-07-25 | 复旦大学 | Method for dynamically dividing shared high-speed caches and circuit |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7529888B2 (en) * | 2004-11-19 | 2009-05-05 | Intel Corporation | Software caching with bounded-error delayed update |
-
2013
- 2013-05-24 CN CN201310195427.7A patent/CN103246616B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1996944A (en) * | 2006-11-09 | 2007-07-11 | 华中科技大学 | A method for global buffer management of the cluster storage system |
CN101609432A (en) * | 2009-07-13 | 2009-12-23 | 中国科学院计算技术研究所 | Shared buffer memory management system and method |
CN102262512A (en) * | 2011-07-21 | 2011-11-30 | 浪潮(北京)电子信息产业有限公司 | System, device and method for realizing disk array cache partition management |
CN102609362A (en) * | 2012-01-30 | 2012-07-25 | 复旦大学 | Method for dynamically dividing shared high-speed caches and circuit |
Also Published As
Publication number | Publication date |
---|---|
CN103246616A (en) | 2013-08-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103246616B (en) | A kind of globally shared buffer replacing method of access frequency within long and short cycle | |
CN107526546B (en) | Spark distributed computing data processing method and system | |
US9348527B2 (en) | Storing data in persistent hybrid memory | |
US9361236B2 (en) | Handling write requests for a data array | |
US10169232B2 (en) | Associative and atomic write-back caching system and method for storage subsystem | |
CN106126112B (en) | Multiple stripe memory with multiple read ports and multiple write ports per cycle | |
CN102117248A (en) | Caching system and method for caching data in caching system | |
CN103106286B (en) | Method and device for managing metadata | |
US20120102273A1 (en) | Memory agent to access memory blade as part of the cache coherency domain | |
CN103019948A (en) | Working set exchange using continuously-sorted swap files | |
CN100383792C (en) | Buffer data base data organization method | |
CN106066890B (en) | Distributed high-performance database all-in-one machine system | |
US11093410B2 (en) | Cache management method, storage system and computer program product | |
CN104145252A (en) | Adaptive cache promotions in a two level caching system | |
CN105138292A (en) | Disk data reading method | |
CN110196818A (en) | Data cached method, buffer memory device and storage system | |
CN102763091A (en) | Integrating a flash cache into large storage systems | |
CN106446268A (en) | Database lateral extension system and method | |
CN105897859A (en) | Storage system | |
CN107341114A (en) | A kind of method of directory management, Node Controller and system | |
CN110297787A (en) | The method, device and equipment of I/O equipment access memory | |
CN104298574A (en) | Data high-speed storage processing system | |
US10705977B2 (en) | Method of dirty cache line eviction | |
US20140297966A1 (en) | Operation processing apparatus, information processing apparatus and method of controlling information processing apparatus | |
CN106126434B (en) | The replacement method and its device of the cache lines of the buffer area of central processing unit |
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 |