CN101236530A - High speed cache replacement policy dynamic selection method - Google Patents
High speed cache replacement policy dynamic selection method Download PDFInfo
- Publication number
- CN101236530A CN101236530A CNA2008100571727A CN200810057172A CN101236530A CN 101236530 A CN101236530 A CN 101236530A CN A2008100571727 A CNA2008100571727 A CN A2008100571727A CN 200810057172 A CN200810057172 A CN 200810057172A CN 101236530 A CN101236530 A CN 101236530A
- Authority
- CN
- China
- Prior art keywords
- cache
- module
- replacement strategy
- cache replacement
- data
- 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
- 238000010187 selection method Methods 0.000 title claims abstract description 16
- 238000000034 method Methods 0.000 claims abstract description 42
- 230000015654 memory Effects 0.000 claims abstract description 25
- 238000004458 analytical method Methods 0.000 claims abstract description 7
- 230000006870 function Effects 0.000 claims description 28
- 230000015572 biosynthetic process Effects 0.000 claims description 16
- 238000005755 formation reaction Methods 0.000 claims description 16
- 238000003860 storage Methods 0.000 claims description 14
- 230000008859 change Effects 0.000 claims description 9
- 238000003780 insertion Methods 0.000 claims description 7
- 230000037431 insertion Effects 0.000 claims description 7
- 238000012423 maintenance Methods 0.000 claims description 4
- 230000008676 import Effects 0.000 claims description 3
- 238000007599 discharging Methods 0.000 claims description 2
- 230000008569 process Effects 0.000 description 17
- 230000004044 response Effects 0.000 description 12
- 238000012360 testing method Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000007630 basic procedure Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 238000000151 deposition Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000012467 final product Substances 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 239000003999 initiator Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
Images
Landscapes
- Memory System Of A Hierarchy Structure (AREA)
Abstract
The invention relates to a dynamic selection method for cache replacement policies, belonging to the memory system cache filed. The invention is characterized in that: the cache replacement policies are modularized through uniform interfaces, thereby online switching can be performed on any two cache replacement policies and novel cache replacement policies can be disposed; assess, record, acquisition and analysis are performed asynchronously, thereby CPU and memory overheads are small and affection on application is small; accurate decision-making results can be obtained as soon as possible through multi-run of policy selection, thereby not only the results are guaranteed to be brought into use quickly but also method overheads are reduced as much as possible.
Description
Technical field
The dynamic selection method of cache replacement strategy belongs to the storage system field, relates in particular to high-speed cache field wherein.
Background technology
The dynamic selection method of cache replacement strategy is meant: present existing cache replacement strategy is a lot, and different strategies has different effects in different loads; This method is selected the cache replacement strategy of the most suitable present load according to the change dynamics of load ground, and carries out online switching, thereby improves the hit rate of storage system high-speed cache, finally improves the overall performance of storage system.Special load or environment that traditional cache replacement strategy has at some feature, write (WOW) strategy at the relatively expensive equipment such as RAID5 of write operation as intelligence, and two locality strategy (DULO) is written as in order effect is preferably arranged in the main load.For load the unknown or that change, be difficult to from these strategies, select.The other cache replacement strategy can be adjusted self adaptively according to the variation of load, as the low load strategy (ARC) of self-adaptation, many formations strategies (MQ) etc., these strategies can be obtained general reasonable effect, but can't obtain best effect on certain loads.So someone has proposed the thought selected from a plurality of cache replacement strategies, self-adaptation multi-expert strategy (ACME) has proposed the selection that multi-expert system is used for cache replacement strategy, but its CPU and memory cost are all bigger, and lacks and realize checking.
The present invention proposes a kind of dynamic selection method of new cache replacement strategy, by asynchronous policy selection the process of policy selection and common cache access flow process are separated, thereby CPU and memory cost have been reduced, by the strategy of on-line switching selection result is come into operation immediately, to improve performance as early as possible.
Summary of the invention
The object of the present invention is to provide a kind of replacement policy system of selection that can be applicable to multiple cache systems, can from a plurality of candidate's cache replacement strategies, select the strategy of the most suitable present load and carry out online switching, thereby improve the performance of system.Reduce CPU and memory cost that selection course is brought simultaneously as far as possible.Modular design and the online switching that focuses on cache replacement strategy of the present invention, and the Dynamic Selection process of optimal cache replacement policy.
The invention is characterized in: described method is to realize according to the following steps successively on the digital integrated circuit in high-speed buffer storage system:
Step (1). initialization is provided with lower module:
The cache replacement strategy module, the cache service module, Visitor Logs collector module and Visitor Logs analyzer module, wherein:
1) cache replacement strategy module, set up with lower interface and describe logic when hitting and replacing:
A. initialization interface, for user's appointment or that determine when moving for the first time or switch after current cache policies module assignment internal memory, the initialization data structure;
B. discharge and use interface, discharge the cache replacement strategy module that is about to be replaced, comprise discharging data structure and internal memory, use during the cache replacement strategy module that Gong is replaced unloading;
C. upgrade and use interface, when cache data blocks is hit, upgrade the priority of this piece;
D. insert and use interface, when inserting new cache data blocks, the priority of this piece is set;
E. replace and use interface, get the minimum cache data blocks of priority and use when being replaced piece;
Total following five kinds of described cache replacement strategy module: be used deque, many formations recently, the low load of self-adaptation and the low nearest degree of correlation, a fixing entrance function register is respectively arranged, so that the interface of oneself is realized that registration is to described cache service module
In described each cache replacement strategy module, be in the current cache replacement strategy module that is of running status, all the other then are candidate's cache replacement strategy module;
2) cache service module, the metadata cache of the low speed storage device that comprises disk in internal memory, directly carry out internal storage access when the user prevents asking data, this cache service module is provided with: cached data, high-speed cache metadata and Hash table, wherein
A. cached data is the good data in magnetic disk of buffer memory, is divided into the cacheline of a plurality of fixed sizes, and the size of piece is 4KB,
B. the high-speed cache metadata is used for the described cached data of maintenance and management, comprising:
1) physical address of described each cacheline;
2) described each cacheline sign bit is used to indicate whether this piece has been written into disk;
3) one is used formation recently, is used to write down the priority orders of described cacheline;
C. Hash table, as key word, corresponding high-speed cache metadata is data, constitutes a data look-up table, whether is used for definite cache data blocks at described cache memory with the address of described cacheline;
Described cache service module realizes cache service successively in the following order:
A. when user accesses data, this cache service module is at first according to the address search Hash table of the given cacheline of data wherein;
B. if described address in Hash table, then high several cache hit use interface by upgrading, and call the function of hitting of described current cache replacement strategy module, direct access cache, sense data returns to the user;
C. if the address is not present in the Hash table, be cache miss then, judge whether cacheline is all occupied:
If all not occupied, then take out a unappropriated cacheline, by described insertion interface, call the insertion function of described current cache replacement strategy module, data are called in described unappropriated cacheline,
If cacheline uses up, then use interface by replacing, call the replacement function of described current cache replacement strategy module, find one to be replaced piece at the described tabulation medium priority minimum cacheline that is used recently, data are read in the cacheline of replacing out, return to the user again;
3) Visitor Logs collector module, the visit information that record is imported from described cache service module, comprising: disk number, reference address and size, be stored in earlier in the core buffer, when being filled with, these visit informations are written in the disk file;
4) Visitor Logs analyzer module, receive the information of the n bar Visitor Logs that collects setting from described Visitor Logs collector module after, each Visitor Logs is imported described each candidate's cache replacement strategy module successively, write down the hit-count of described each candidate's cache replacement strategy module, get the maximum candidate's cache replacement strategy module of hit-count for optimum, the described cache service module of result notification;
Step (2). carry out the dynamic selection method of cache replacement strategy successively according to the following steps:
Step (2.1). initialization cache service module:
Distribute and all cache data blocks of initialization;
Distribute and all data cached block descriptors of universal high speed of initialization, described cache data blocks is used to write down the address and the sign bit of this cache data blocks, and " general " two words are meant what each cache replacement strategy module all need be used;
Initialization Hash table and be used formation recently;
Step (2.2). the initialization of current cache replacement strategy module
Step (2.2.1). determine that a cache replacement strategy module is current cache replacement strategy module
When being specified by the user, acquiescence is got the last time of used module during end of run;
When moving for the first time, get the described policy module that is used recently;
Step (2.2.2). this current cache replacement strategy module of the described register function registration of cache service module invokes, again by the described current cache replacement strategy module of initialization interface initialization;
Step (2.3). again during user accesses data, the cache service module set by step (1) described method to insert user accesses data be cacheline;
Step (2.4). the Visitor Logs collector is gathered the user capture record;
Step (2.5). the Visitor Logs analysis
After the Visitor Logs collector collects described n bar Visitor Logs, a virtual cache service module of being set by oneself by the Visitor Logs analyzer module writes down the record hit-count when Visitor Logs is visited each candidate's cache replacement strategy module successively, and selecting the highest candidate's cache replacement strategy module of hit-count is the optimal strategy module;
Step (2.6). switch the cache replacement strategy module,
At first the cache service module discharges policy metadata by the current release that is in the cache replacement strategy module of running status with interface, and described policy metadata is meant the peculiar metadata of each cache replacement strategy module;
Then, the cache service module is distributed new policy metadata space by the initialization interface of that the highest new cache replacement strategy module of Visitor Logs hit-count;
At last, the cache service module imports new cache replacement strategy module to existing cache data blocks in the high-speed cache to rebuild new policy metadata by the insertion of new cache replacement strategy module successively with interface.
Described Visitor Logs times of collection n=1,000,000.After execution of step (2.6),, then return step (2.4) if epicycle is the first round.If the epicycle policy selection draw with last round of identical, then finish, otherwise, change step (2.4).After execution of step (2.6), carry out the cache service step of unloading.
Advantage of the present invention is as follows:
(1) the cache service module is used the cache replacement strategy module by fixed interface, can between any two cache replacement strategy modules, switch, also can dispose emerging cache replacement strategy module, thereby can utilize up-to-date scientific payoffs as early as possible.
(2) by asynchronous policy selection, Visitor Logs collection and analytic process and common cache access flow process are separated, reduce CPU and memory cost that the policy selection process is brought, thereby reduce the average response time of cache access.
(3) by many wheel policy selection, the time of equilibrium strategy selection course and accuracy reduce Visitor Logs and analyze the required time under the prerequisite that guarantees the selection accuracy, finally reduce the expense of policy selection process, and as early as possible the result is come into operation.
The present invention tests in department of computer science, Tsinghua university high-performance calculation technical institute.The result shows that the dynamic selection method of cache replacement strategy can be selected optimum cache replacement strategy according to loading condition, has effectively improved the hit rate of cache systems I/O visit, has reduced average response time.
Test to great magnitude of data hierarchical storage method is divided into simulation test and real system test two parts, the environment of simulation test is Xeon 700MHz*4, the 1GB internal memory, we have used Cello, SPC-1 Websearch, four kinds of trace of SPC-1 Financial and TPCC simulate, and test result sees Table 1, and this method can be selected the highest strategy of hit rate as can be seen.Fig. 5 and Fig. 6 have described the hit rate situation of change under two kinds of situations, and by policy selection, this method has been selected optimum strategy, shoot straight in other adaptive strategy as can be seen.
The environment of real system test is: machine has set up our cache systems in the above as the object machine end of iSCSI, it be configured to Xeon 2.8GHz CPU, 4GB internal memory and 6 IBM disks.A machine has been installed Windows 2000 and Microsoft iSCSI Initiator as the iSCSI front end.In trace player plays SPC-1 Websearch and two kinds of loads of TPCC that front end uses us to develop, test result sees Table 2, table 3 and Fig. 7, Fig. 8.Table 2 has provided the average response time that measures in the Websearch load, the average response time of this method is lower than other adaptive algorithm as can be seen, Fig. 7 has provided the average response time situation of change under a kind of situation, by policy selection, this method can reduce the average response time of cache systems as can be seen.Table 3 has provided the average response time that measures in the TPCC load, and Fig. 8 has gone out the average response time situation of change under a kind of situation, similar as can be seen trend.
Description of drawings
Fig. 1. overall architecture.
Fig. 2. interface and Data Structure Design.
Fig. 3. the cache access basic procedure.
Fig. 4. the policy selection process.
Fig. 5. (W3,4G) the hit rate situation of change under:
-■-be used recently (LRU);
--self-adaptation is hanged down load (ARC);
-★-low nearest the degree of correlation (LIRS);
-◇-this method.
Fig. 6. (T2,256M) the hit rate situation of change under:
-■-be used recently (LRU);
--self-adaptation is hanged down load (ARC);
-★-low nearest the degree of correlation (LIRS);
-◇-this method.
Fig. 7. (W3, average response time 3G):
-■-be used recently (LRU);
--self-adaptation is hanged down load (ARC);
-◇-this method;
Fig. 8. (T1, average response time 512M):
-■-be used recently (LRU);
--self-adaptation is hanged down load (ARC);
-◇-this method;
Embodiment
The present invention can realize in the secondary storage architecture of being made up of any high speed and low speed storage device.High speed storing equipment is generally internal memory, perhaps solid magnetic disc, and it is fast to have speed, the price height, capacity is little, the characteristics that outage is easily lost.According to the principle of data access locality, their buffer memorys are the data of visit recently, thereby improve the performance of total system.Low speed storage device is generally ordinary magnetic disc, and it is slow to have speed, and price is low, and capacity is big, non-volatile characteristics.They have stored all data.The dynamic selection method of cache replacement strategy is mainly by the cache service module, the cache replacement strategy module, and Visitor Logs acquisition module and Visitor Logs analysis module are formed.The architecture of this method as shown in Figure 1.
Module (1). the cache replacement strategy module
The cache replacement strategy module has been described the logic when cache systems hits and replaces.The cache service module is by the logic function of a unified interface accessing cache replacement strategy module.The interface of cache replacement strategy module and Data Structure Design are as shown in Figure 2.The cache replacement strategy module has a fixing entrance function register, and it realizes that registration is to the cache service module to oneself interface by this function.
It provides five kinds of functions to use for the cache service module
1) initialization, this function is called when the cache systems initialization;
2) unloading, this function is called when cache module unloads;
3) cacheline inserts, and is called when new cacheline adds fashionable this function;
4) cacheline hits, and this function is called when cacheline hits;
5) cacheline is replaced, and this function is called when new cacheline is replaced old cacheline.
Being used for the data of management of cache is called the high-speed cache metadata, comprises that the cache data blocks descriptor is used for recording data blocks corresponding address and sign bit; Hash table is used for writing down a data block whether in high-speed cache; Be used the priority orders that formation is used for representing data block recently.This method is divided into data cached block descriptor of universal high speed and tactful cache data blocks descriptor with the cache data blocks descriptor, is combined into universal element data and policy metadata respectively.The universal element data refer to the metadata that basic All Policies all will be used, and policy metadata refers to each tactful distinctive metadata.The universal element data are by cache service module assignment and maintenance, and policy metadata is safeguarded voluntarily by the cache replacement strategy module and used.
This method has realized being used recently (LRU), deque (2Q), the low nearest degree of correlation (LIRS), self-adaptation is hanged down load (ARC) and five kinds of cache replacement strategy modules of many formations (MQ), wherein be in the current cache replacement strategy module that is of running status, other be candidate's cache replacement strategy module, this method can dynamically switch to current cache replacement strategy module with candidate's cache replacement strategy module
The process that strategy switches is as follows: at first the unloading function of the old cache replacement strategy of cache service module invokes discharges policy metadata, the initialization function of the new cache replacement strategy of cache service module invokes distributes new policy metadata then, and the insertion function of the last new cache replacement strategy of cache service module invokes imports existing data block in the high-speed cache successively new cache replacement strategy module and sets up policy metadata.
Module (2). the cache service module
The cache service module provides cache service for the user, and it in internal memory, just directly carries out the metadata cache of low-speed devices such as disk internal storage access, thereby improve access speed greatly when data are visited once more.The cache service module utilizes the cacheline replacement function of current cache replacement strategy module to determine which data block can be placed in the high-speed cache.The cache service module can select the highest cache replacement strategy cross cutting of hit rate to be changed to current cache replacement strategy module in candidate's cache replacement strategy module according to the analysis result of Visitor Logs analyzer module.The cache service module mainly contains following ingredient and logic:
Cached data: the data in magnetic disk of buffer memory is divided into the cacheline of a plurality of fixed sizes.The size of piece can be specified by the user, is defaulted as 4KB.
High-speed cache metadata: the data that are used for the maintenance and management high-speed cache, the physical address that comprises each cacheline, whether this piece of sign bit sign of each cacheline has been written into disk, and one is used the priority that formation is used to write down cacheline recently.
Hash table: the address of using cacheline is as key word, and corresponding high-speed cache metadata is used for determining that as the data look-up table of data whether a data block is at high-speed cache.
The basic procedure of cache service module as shown in Figure 3, when the application access data, at first check that by reference address corresponding data block is whether in high-speed cache, if words be cache hit, directly from high-speed cache, visit and get final product, if not words be cache miss, at this moment check and whether also have vacant cache data blocks, directly accessed piece is read in this cache data blocks if any, from all data blocks of high-speed cache, select one and be replaced piece if it's not true, then accessed piece is read in this replace block, at this moment visit data piece from high-speed cache again.
Module (3). Visitor Logs collector module
When user capture cache service module, the information of the current visit of Visitor Logs collector module records comprises disk number, reference address and size.Visitor Logs collector module at first is recorded in these information in the core buffer, when buffer zone is full, these information is written in the disk file.
Module (4). the Visitor Logs analyzer module
When Visitor Logs collector module collects 1,000, behind 000 record, the Visitor Logs analyzer module is promptly started working, and whole process as shown in Figure 4.The Visitor Logs analyzer module is at first set up a virtual cache service module, and it is not except depositing the cached data, and other are all identical with the cache service module.Then the Visitor Logs analyzer module is specified a cache policies module for the virtual cache service module, then according to Visitor Logs accesses virtual cache service module successively, and the record hit-count.By all candidate's cache policies modules are repeated said process, the Visitor Logs analyzer module can obtain the hit-count of all candidate's cache policies modules, and to select the highest strategy of hit-count be optimal strategy and carry out strategy switching.If this strategy of picking out is with last round of identical, then the policy selection process finishes, otherwise the Visitor Logs collector is gathered 1,000,000 Visitor Logs again, is analyzed by the Visitor Logs analyzer module again.Many wheels policy selection process so both can guarantee the accuracy analyzed, also can avoid too much Visitor Logs collection and analytic process.
After many wheels policy selection finished, cache systems entered normal operating phase, and this stage does not have the process of Visitor Logs collection and analysis, did not almost have additional CPU and memory cost.The time T that Visitor Logs is analyzed
a, the time T that Visitor Logs is gathered
cTime T with normal operating phase
rFollowing relation: T is arranged
a<<T
c<<T
r, this explanation normal operating phase has accounted for the most of the time, has guaranteed that the CPU of total system and memory cost are smaller.
The dynamic selection method of described cache replacement strategy contains following steps successively:
Step 1: cache service module initialization
The initialization of cache service module comprises:
1) distribution and all cache data blocks of initialization;
2) distribution and all data cached block descriptors of universal high speed of initialization;
3) initialization Hash table
4) initialization is used formation recently
Step 2: current cache replacement strategy module initialization
The cache service module determines that at first a cache replacement strategy module is the current strategies module, and this can be specified by the user, and acquiescence is got the module of using during end of run last time, is used policy module recently if operation is for the first time got.The register function of cache service module invokes cache replacement strategy module is registered this module then, calls this policy module of initialization interface function initialization of this policy module at last.
Step 3: application access
The cache service module is at first checked Hash table, if the address of corresponding data piece in high-speed cache, direct access cache then.If not in high-speed cache, check whether high-speed cache full, if less than directly it would be called in high-speed cache and visit, if full then select earlier and be replaced piece, then accessed called in and visits.
Step 4: Visitor Logs collection
The Visitor Logs acquisition module writes down the information of application access, comprises reel number, start address and the size of visit.The Visitor Logs acquisition module at first records the information in the core buffer, again it is write disk file when core buffer is full.
Step 5: Visitor Logs analysis
When Visitor Logs collector module collects 1,000, behind 000 record, the Visitor Logs analyzer module is at first set up a virtual cache service module.Then the Visitor Logs analyzer module is specified a cache policies module for the virtual cache service module, then according to Visitor Logs accesses virtual cache service module successively, and the record hit-count.By all candidate's cache policies modules are repeated said process, the Visitor Logs analyzer module can obtain the hit-count of all candidate's cache policies modules, and selecting the highest strategy of hit-count is optimal strategy.
Step 6: strategy switches
Cache service module invokes unloading interface function discharges old policy module, call the register and the initialization interface function of optimal strategy module then, at last existing data block in the high-speed cache is imported new cache replacement strategy module by inserting interface function, finish handoff procedure.
Step 7: take turns policy selection more
If the optimal strategy that the epicycle policy selection obtains is with last round of identical, then finish, otherwise jump procedure 4 is carried out the next round Visitor Logs and is gathered and analyze.If epicycle is exactly certain jump procedure 4 of the first round.
Step 8: policy selection is restarted
When the hit rate of cache service module discovery strategy significantly descends, restart the policy selection process, promptly carry out step 4-7 again.
Step 9: cache service module unloading
1) data block that all is not write disk writes disk
2) discharge policy module, discharge its internal memory and data structure
3) discharge all universal element data, comprise the data cached block descriptor of universal high speed, Hash table and be used formation recently
4) discharge all cache data blocks
Hit rate situation under various strategies of table 1 and the load
Trace | LRU | ARC | MQ | 2Q | LIRS | This method | |||||
0.2 | 0.3 | 0.01 | 0.1 | 0.2 | 0.3 | Select | |||||
SPC-1 Financial | |||||||||||
F1,4M | 66.4 | 66.4 | 67.3 | 65 | 66.1 | 57.6 | 65.6 | 66.2 | 66.4 | 67.2 | MQ |
F1,8M | 68.3 | 68.3 | 68.9 | 68.5 | 69.1 | 62.4 | 68 | 68.2 | 68.2 | 68.7 | 2Q(0.3) |
F1,16M | 70.1 | 70.2 | 70.3 | 71 | 71.4 | 67.1 | 70 | 70.1 | 70.1 | 71 | 2Q(0.3) |
F1,32M | 71.8 | 72 | 72 | 72.7 | 72.9 | 70.6 | 71.9 | 71.9 | 71.9 | 72.7 | 2Q(0.3) |
F1,64M | 73.2 | 73.3 | 73.2 | 73.8 | 74 | 72.6 | 73.2 | 73.3 | 73.3 | 73.3 | URS(0.3) |
F2,4M | 53.5 | 54.2 | 56.5 | 56.1 | 56.1 | 49.6 | 53.4 | 54 | 54.1 | 56.1 | MQ |
F2,8M | 61.9 | 62.3 | 64.3 | 64.6 | 64.5 | 59.1 | 61.9 | 62.1 | 62.2 | 64.1 | 2Q(0.3) |
F2,16M | 71.2 | 70.8 | 72.3 | 73 | 73.2 | 68.5 | 70.3 | 70.7 | 70.9 | 72.7 | 2Q(0.3) |
F2,32M | 78.2 | 77.7 | 78.6 | 80.1 | 80.1 | 76 | 77.5 | 78.1 | 78.3 | 79.5 | 2Q(0.2) |
F2,64M | 83.2 | 83.1 | 83.3 | 84.6 | 84.5 | 81.5 | 83.3 | 83.5 | 83.5 | 84.2 | 2Q(0.2) |
SPC-1 Websearch | |||||||||||
W1,1G | 0.3 | 0.7 | 5.4 | 1.3 | 1.3 | 13.2 | 12.2 | 11.1 | 10 | 10.5 | LIRS(001) |
W1,2G | 11.2 | 20.6 | 17.9 | 19.9 | 18.7 | 30.8 | 28.8 | 26.5 | 24.1 | 27.5 | LIRS(0.01) |
W1,4G | 48.9 | 50.7 | 49.9 | 44.2 | 44.2 | 53.5 | 51.7 | 50.6 | 50.1 | 51.6 | LIRS(0.01) |
W2,1G | 0.4 | 2 | 12.2 | 4.2 | 4.2 | 14.8 | 13.7 | 12.5 | 11.1 | 13.7 | LIRS(001) |
W2,2G | 12.5 | 30.2 | 27.3 | 29 | 25 | 35.1 | 32.9 | 30.5 | 27.7 | 33.8 | LIRS(001) |
W2,4G | 57.5 | 61.6 | 61.2 | 57.9 | 56.9 | 65.4 | 62.9 | 61.4 | 60.9 | 64.6 | LIRS(001) |
W3,1G | 0.4 | 2.9 | 15.7 | 5.8 | 5.8 | 15.2 | 14 | 12.7 | 11.2 | 14.5 | LIRS(0.01) |
W3,2G | 11.9 | 31.5 | 31.5 | 30.8 | 26.2 | 36.2 | 33.9 | 31.3 | 28.3 | 35.4 | LIRS(0.01) |
W3,4G | 58.7 | 62.8 | 64.6 | 60.3 | 58.8 | 67.5 | 65 | 63.3 | 62.7 | 67 | LIRS(0.01) |
Cello99 | |||||||||||
C1,4M | 34.8 | 36.5 | 36.9 | 38.3 | 38.2 | 25.5 | 36.3 | 36.3 | 36.2 | 382 | 2Q(0.2,0.3) |
C1,8M | 37.3 | 38.7 | 38.7 | 40.2 | 40.2 | 37.4 | 38.5 | 38.6 | 38.5 | 40.1 | 2Q(0.2,0.3) |
C1,16M | 40.4 | 41.3 | 41.3 | 42.7 | 42.7 | 40.4 | 40.9 | 40.9 | 41 | 42.5 | 2Q(0.3),LIRS(0.1),ARC |
C1,32M | 44.4 | 44.9 | 44.8 | 45.4 | 45.6 | 42.9 | 43.4 | 43.6 | 43.8 | 45.4 | 2Q(0.3),LIRS(0.2),ARC |
C1,64M | 48.8 | 49 | 49 | 49.4 | 49.8 | 45.9 | 46.6 | 47.1 | 47.4 | 49.2 | ARC,2Q(0.3) |
C1,128M | 53 | 53 | 53.1 | 53.7 | 54.1 | 49.2 | 50.4 | 51.1 | 51.7 | 53.3 | MQ,2Q(0.2) |
C1,256M | 56.2 | 56.2 | 56.2 | 57.2 | 57.6 | 52.4 | 54.2 | 55.3 | 55.7 | 56.2 | MQ,LIRS(0.3) |
C1,512M | 59 | 59.2 | 59 | 60.4 | 60.7 | 55.9 | 58.3 | 59 | 59.2 | 59.1 | LRU,LIRS(0.2) |
C1,1G | 61.6 | 61.8 | 61.6 | 63.5 | 63.7 | 59.3 | 61.8 | 62.1 | 62.2 | 61.7 | LRU,ARC |
TPC-C | |||||||||||
T1,128M | 5.2 | 12.5 | 5.8 | 7.4 | 7.4 | 19.6 | 18.3 | 16.8 | 15.4 | 19.5 | ARC,LIRS(0.01 |
T1,256M | 5.4 | 19.5 | 19.6 | 54 | 5.4 | 34.3 | 31.7 | 28.8 | 25.8 | 33.9 | ARC,LIRS(0.01) |
T1,512M | 9.3 | 13.4 | 35.6 | 43.6 | 31.5 | 63.6 | 58.4 | 52.8 | 47.4 | 63.3 | LIRS(0.01) |
T1,1G | 94 | 93.9 | 91.7 | 91.8 | 96 | 96 | 95.1 | 94.2 | 93.4 | 94.2 | LIRS(0.2) |
T1,2G | 99.6 | 99.6 | 99.6 | 99.6 | 99.6 | 99.6 | 99.6 | 99.6 | 99.6 | 99.6 | LRU |
Average response time (millisecond) in the table 2 Websearch load
LRU | ARC | MQ | This method | |
W3,1G | 9.5 | 9.4 | 8.5 | 8.3 |
W3,2G | 8.3 | 6.9 | 6.6 | 6.6 |
W3,3G | 6.1 | 5.4 | 5.4 | 5.0 |
Average response time (millisecond) in the table 3 TPCC load
LRU | ARC | MQ | This method | |
T1,256M | 6.5 | 6.3 | 6.5 | 6.1 |
T1,512M | 6.4 | 6.3 | 6.2 | 5.5 |
Claims (5)
1. the dynamic selection method of cache replacement strategy is characterized in that: described method is to realize according to the following steps successively on the digital integrated circuit in high-speed buffer storage system:
Step (1). initialization is provided with lower module:
The cache replacement strategy module, the cache service module, Visitor Logs collector module and Visitor Logs analyzer module, wherein:
1) cache replacement strategy module, set up with lower interface and describe logic when hitting and replacing:
A. initialization interface, for user's appointment or that determine when moving for the first time or switch after current cache policies module assignment internal memory, the initialization data structure;
B. discharge and use interface, discharge the cache replacement strategy module that is about to be replaced, comprise discharging data structure and internal memory, use during the cache replacement strategy module that Gong is replaced unloading;
C. upgrade and use interface, when cache data blocks is hit, upgrade the priority of this piece;
D. insert and use interface, when inserting new cache data blocks, the priority of this piece is set;
E. replace and use interface, get the minimum cache data blocks of priority and use when being replaced piece;
Total following five kinds of described cache replacement strategy module: be used deque, many formations recently, the low load of self-adaptation and the low nearest degree of correlation, a fixing entrance function register is respectively arranged, so that the interface of oneself is realized that registration is to described cache service module
In described each cache replacement strategy module, be in the current cache replacement strategy module that is of running status, all the other then are candidate's cache replacement strategy module;
2) cache service module, the metadata cache of the low speed storage device that comprises disk in internal memory, directly carry out internal storage access when user accesses data, this cache service module is provided with: cached data, high-speed cache metadata and Hash table, wherein
A. cached data is the good data in magnetic disk of buffer memory, is divided into the cacheline of a plurality of fixed sizes, and the size of piece is 4KB,
B. the high-speed cache metadata is used for the described cached data of maintenance and management, comprising:
1) physical address of described each cacheline;
2) described each cacheline sign bit is used to indicate whether this piece has been written into disk;
3) one is used formation recently, is used to write down the priority orders of described cacheline;
C. Hash table, as key word, corresponding high-speed cache metadata is data, constitutes a data look-up table, whether is used for definite cache data blocks at described cache memory with the address of described cacheline;
Described cache service module realizes cache service successively in the following order:
A. when user accesses data, this cache service module is at first according to the address search Hash table of the given cacheline of data wherein;
B. if described address in Hash table, then high several cache hit use interface by upgrading, and call the function of hitting of described current cache replacement strategy module, direct access cache, sense data returns to the user;
C. if the address is not present in the Hash table, be cache miss then, judge whether cacheline is all occupied:
If all not occupied, then take out a unappropriated cacheline, by described insertion interface, call the insertion function of described current cache replacement strategy module, data are called in described unappropriated cacheline,
If cacheline uses up, then use interface by replacing, call the replacement function of described current cache replacement strategy module, find one to be replaced piece at the described tabulation medium priority minimum cacheline that is used recently, data are read in the cacheline of replacing out, return to the user again;
3) Visitor Logs collector module, the visit information that record is imported from described cache service module, comprising: disk number, reference address and size, be stored in earlier in the core buffer, when being filled with, these visit informations are written in the disk file;
4) Visitor Logs analyzer module, receive the information of the n bar Visitor Logs that collects setting from described Visitor Logs collector module after, each Visitor Logs is imported described each candidate's cache replacement strategy module successively, write down the hit-count of described each candidate's cache replacement strategy module, get the maximum candidate's cache replacement strategy module of hit-count
Be optimum, the described cache service module of result notification;
Step (2). carry out the dynamic selection method of cache replacement strategy successively according to the following steps:
Step (2.1). initialization cache service module:
Distribute and all cache data blocks of initialization;
Distribute and all data cached block descriptors of universal high speed of initialization, described cache data blocks is used to write down the address and the sign bit of this cache data blocks, and " general " two words are meant what each cache replacement strategy module all need be used;
Initialization Hash table and be used formation recently;
Step (2.2). the initialization of current cache replacement strategy module
Step (2.2.1). determine a cache replacement strategy module be current cache replacement strategy module when specifying by the user, acquiescence is got the last time of used module during end of run;
When moving for the first time, get the described policy module that is used recently;
Step (2.2.2). this current cache replacement strategy module of the described register function registration of cache service module invokes, again by the described current cache replacement strategy module of initialization interface initialization;
Step (2.3). again during user accesses data, the cache service module set by step (1) described method to insert user accesses data be cacheline;
Step (2.4). the Visitor Logs collector is gathered the user capture record;
Step (2.5). the Visitor Logs analysis
After the Visitor Logs collector collects described n bar Visitor Logs, a virtual cache service module of being set by oneself by the Visitor Logs analyzer module writes down the record hit-count when Visitor Logs is visited each candidate's cache replacement strategy module successively, and selecting the highest candidate's cache replacement strategy module of hit-count is the optimal strategy module;
Step (2.6). switch the cache replacement strategy module,
At first the cache service module discharges policy metadata by the current release that is in the cache replacement strategy module of running status with interface, and described policy metadata is meant the peculiar metadata of each cache replacement strategy module;
Then, the cache service module is distributed new policy metadata space by the initialization interface of that the highest new cache replacement strategy module of Visitor Logs hit-count;
At last, the cache service module imports new cache replacement strategy module to existing cache data blocks in the high-speed cache to rebuild new policy metadata by the insertion of new cache replacement strategy module successively with interface.
2. the dynamic selection method of cache replacement strategy according to claim 1 is characterized in that, described Visitor Logs times of collection n=1,000,000.
3. the dynamic selection method of cache replacement strategy according to claim 1 is characterized in that, after execution of step (2.6), if epicycle is the first round, then returns step (2.4).
4. the dynamic selection method of cache replacement strategy according to claim 1 is characterized in that, after execution of step (2.6), if the epicycle policy selection draw with last round of identical, then finish, otherwise, change step (2.4).
5. the dynamic selection method of cache replacement strategy according to claim 1 is characterized in that, after execution of step (2.6), carries out the cache service step of unloading.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2008100571727A CN101236530B (en) | 2008-01-30 | 2008-01-30 | High speed cache replacement policy dynamic selection method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2008100571727A CN101236530B (en) | 2008-01-30 | 2008-01-30 | High speed cache replacement policy dynamic selection method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101236530A true CN101236530A (en) | 2008-08-06 |
CN101236530B CN101236530B (en) | 2010-09-01 |
Family
ID=39920160
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2008100571727A Expired - Fee Related CN101236530B (en) | 2008-01-30 | 2008-01-30 | High speed cache replacement policy dynamic selection method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101236530B (en) |
Cited By (40)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101866321A (en) * | 2010-06-13 | 2010-10-20 | 北京北大众志微系统科技有限责任公司 | Adjustment method and system for cache management strategy |
CN101916302A (en) * | 2010-09-01 | 2010-12-15 | 中国地质大学(武汉) | Three-dimensional spatial data adaptive cache management method and system based on Hash table |
CN101556557B (en) * | 2009-05-14 | 2011-03-23 | 浙江大学 | Object file organization method based on object storage device |
CN102137139A (en) * | 2010-09-26 | 2011-07-27 | 华为技术有限公司 | Method and device for selecting cache replacement strategy, proxy server and system |
CN101478472B (en) * | 2008-10-21 | 2011-09-07 | 北京闪联讯通数码科技有限公司 | Socket data transmission processing method and apparatus |
CN102395957A (en) * | 2009-04-17 | 2012-03-28 | 韩商英得联股份有限公司 | Cache and disk management method, and a controller using the method |
CN102411541A (en) * | 2010-10-13 | 2012-04-11 | 微软公司 | Dynamic cache configuration using separate read and write caches |
CN103207838A (en) * | 2012-01-17 | 2013-07-17 | 展讯通信(上海)有限公司 | Method for improving property of chip |
WO2014094306A1 (en) * | 2012-12-21 | 2014-06-26 | 华为技术有限公司 | Method and device for setting working mode of cache |
US8839275B1 (en) | 2011-06-06 | 2014-09-16 | Proximal Data, Inc. | Method for intercepting input/output requests and responses |
CN104104710A (en) * | 2013-04-15 | 2014-10-15 | 同济大学 | Low energy consumption-based data cache method in mobile cloud computing environment |
CN104572502A (en) * | 2015-01-12 | 2015-04-29 | 浪潮电子信息产业股份有限公司 | Self-adaptive method for cache strategy of storage system |
CN104598394A (en) * | 2013-10-31 | 2015-05-06 | 中国石油天然气集团公司 | Data caching method and system capable of conducting dynamic distribution |
CN104731980A (en) * | 2015-04-17 | 2015-06-24 | 吉林大学 | Method and device for management of cache pages |
US9104552B1 (en) | 2012-06-23 | 2015-08-11 | Samsung Electronics Co., Ltd. | Method for the use of shadow ghost lists to prevent excessive wear on FLASH based cache devices |
CN105027211A (en) * | 2013-01-31 | 2015-11-04 | 惠普发展公司,有限责任合伙企业 | Voltage regulator |
WO2015169245A1 (en) * | 2014-05-09 | 2015-11-12 | 华为技术有限公司 | Data caching method, cache and computer system |
CN105095354A (en) * | 2015-06-19 | 2015-11-25 | 北京奇虎科技有限公司 | Data updating method and data updating device |
CN105512051A (en) * | 2015-12-16 | 2016-04-20 | 鸿秦(北京)科技有限公司 | Self-learning type intelligent solid-state hard disk cache management method and device |
CN105740167A (en) * | 2016-01-28 | 2016-07-06 | 浪潮(北京)电子信息产业有限公司 | File system cache deletion method and system |
CN106164875A (en) * | 2014-04-04 | 2016-11-23 | 高通股份有限公司 | Carry out adaptivity cache prefetch to reduce cache pollution based on the special strategy that prefetches of the competitiveness in private cache group |
CN106649041A (en) * | 2016-12-27 | 2017-05-10 | 郑州云海信息技术有限公司 | Device and method for automatically testing storage caching mode |
CN106873917A (en) * | 2017-02-24 | 2017-06-20 | 深圳市中博睿存信息技术有限公司 | A kind of object storage system based on disk |
US9690496B2 (en) | 2004-10-21 | 2017-06-27 | Microsoft Technology Licensing, Llc | Using external memory devices to improve system performance |
CN107220188A (en) * | 2017-05-31 | 2017-09-29 | 莫倩 | A kind of automatic adaptation cushion block replacement method |
CN104077241B (en) * | 2014-07-14 | 2017-10-17 | 华为技术有限公司 | Cache life cycle algorithm switching handling method and device |
CN108073527A (en) * | 2016-11-07 | 2018-05-25 | 华为技术有限公司 | It is a kind of to cache the method and apparatus replaced |
CN108133267A (en) * | 2016-12-01 | 2018-06-08 | 上海兆芯集成电路有限公司 | With the processor that can be used as most rear class cache tile or the memory array of neural network cell memory operation |
CN109298972A (en) * | 2018-09-06 | 2019-02-01 | 高盈量化云科技(深圳)有限公司 | A kind of High Availabitity policy engine |
WO2019048969A1 (en) * | 2017-09-05 | 2019-03-14 | International Business Machines Corporation | Asynchronous update of metadata tracks in response to a cache hit generated via an i/o operation over a bus interface |
CN109656832A (en) * | 2017-10-11 | 2019-04-19 | 深圳市中兴微电子技术有限公司 | A kind of look-up method, computer readable storage medium |
CN109669882A (en) * | 2018-12-28 | 2019-04-23 | 贵州华芯通半导体技术有限公司 | Dynamic caching replacement method, device, system and the medium of bandwidth aware |
CN110032528A (en) * | 2019-04-19 | 2019-07-19 | 苏州浪潮智能科技有限公司 | Internal storage data lookup method, device, equipment and the storage medium of storage system |
US10387313B2 (en) | 2008-09-15 | 2019-08-20 | Microsoft Technology Licensing, Llc | Method and system for ensuring reliability of cache data and metadata subsequent to a reboot |
CN110249318A (en) * | 2017-02-07 | 2019-09-17 | 华为技术有限公司 | Cache replacement system and method |
US10509730B2 (en) | 2008-09-19 | 2019-12-17 | Microsoft Technology Licensing, Llc | Aggregation of write traffic to a data store |
CN110992671A (en) * | 2019-12-16 | 2020-04-10 | 宁波三星医疗电气股份有限公司 | Ammeter address forwarding method, device and system and electronic equipment |
CN113342706A (en) * | 2021-05-13 | 2021-09-03 | 武汉大学 | Write-optimized extensible hash index structure based on nonvolatile memory and inserting, refreshing and deleting methods |
CN114072774A (en) * | 2019-07-01 | 2022-02-18 | 国际商业机器公司 | Block mode switching in a data storage system |
WO2023098277A1 (en) * | 2021-12-01 | 2023-06-08 | International Business Machines Corporation | Augmenting cache replacement operations |
-
2008
- 2008-01-30 CN CN2008100571727A patent/CN101236530B/en not_active Expired - Fee Related
Cited By (63)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9690496B2 (en) | 2004-10-21 | 2017-06-27 | Microsoft Technology Licensing, Llc | Using external memory devices to improve system performance |
US10387313B2 (en) | 2008-09-15 | 2019-08-20 | Microsoft Technology Licensing, Llc | Method and system for ensuring reliability of cache data and metadata subsequent to a reboot |
US10509730B2 (en) | 2008-09-19 | 2019-12-17 | Microsoft Technology Licensing, Llc | Aggregation of write traffic to a data store |
CN101478472B (en) * | 2008-10-21 | 2011-09-07 | 北京闪联讯通数码科技有限公司 | Socket data transmission processing method and apparatus |
CN102395957A (en) * | 2009-04-17 | 2012-03-28 | 韩商英得联股份有限公司 | Cache and disk management method, and a controller using the method |
CN101556557B (en) * | 2009-05-14 | 2011-03-23 | 浙江大学 | Object file organization method based on object storage device |
CN101866321A (en) * | 2010-06-13 | 2010-10-20 | 北京北大众志微系统科技有限责任公司 | Adjustment method and system for cache management strategy |
CN101916302B (en) * | 2010-09-01 | 2012-11-21 | 中国地质大学(武汉) | Three-dimensional spatial data adaptive cache management method and system based on Hash table |
CN101916302A (en) * | 2010-09-01 | 2010-12-15 | 中国地质大学(武汉) | Three-dimensional spatial data adaptive cache management method and system based on Hash table |
CN102137139A (en) * | 2010-09-26 | 2011-07-27 | 华为技术有限公司 | Method and device for selecting cache replacement strategy, proxy server and system |
US8607001B2 (en) | 2010-09-26 | 2013-12-10 | Huawei Technologies Co., Ltd. | Method, apparatus, proxy server, and system for selecting cache replacement policies |
CN102411541A (en) * | 2010-10-13 | 2012-04-11 | 微软公司 | Dynamic cache configuration using separate read and write caches |
CN102411541B (en) * | 2010-10-13 | 2014-12-24 | 微软公司 | Dynamic cache configuration using separate read and write caches |
US8839275B1 (en) | 2011-06-06 | 2014-09-16 | Proximal Data, Inc. | Method for intercepting input/output requests and responses |
CN103207838A (en) * | 2012-01-17 | 2013-07-17 | 展讯通信(上海)有限公司 | Method for improving property of chip |
CN103207838B (en) * | 2012-01-17 | 2016-03-30 | 展讯通信(上海)有限公司 | Improve the method for chip performance |
US9104552B1 (en) | 2012-06-23 | 2015-08-11 | Samsung Electronics Co., Ltd. | Method for the use of shadow ghost lists to prevent excessive wear on FLASH based cache devices |
CN104321754A (en) * | 2012-12-21 | 2015-01-28 | 华为技术有限公司 | Method and device for setting working mode of Cache |
WO2014094306A1 (en) * | 2012-12-21 | 2014-06-26 | 华为技术有限公司 | Method and device for setting working mode of cache |
CN105027211B (en) * | 2013-01-31 | 2018-09-21 | 慧与发展有限责任合伙企业 | Adaptive granularity line buffer cache |
CN105027211A (en) * | 2013-01-31 | 2015-11-04 | 惠普发展公司,有限责任合伙企业 | Voltage regulator |
CN104104710A (en) * | 2013-04-15 | 2014-10-15 | 同济大学 | Low energy consumption-based data cache method in mobile cloud computing environment |
CN104104710B (en) * | 2013-04-15 | 2017-05-24 | 同济大学 | Low energy consumption-based data cache method in mobile cloud computing environment |
CN104598394A (en) * | 2013-10-31 | 2015-05-06 | 中国石油天然气集团公司 | Data caching method and system capable of conducting dynamic distribution |
CN106164875A (en) * | 2014-04-04 | 2016-11-23 | 高通股份有限公司 | Carry out adaptivity cache prefetch to reduce cache pollution based on the special strategy that prefetches of the competitiveness in private cache group |
CN105094686A (en) * | 2014-05-09 | 2015-11-25 | 华为技术有限公司 | Data caching method, cache and computer system |
US10241919B2 (en) | 2014-05-09 | 2019-03-26 | Huawei Technologies Co., Ltd. | Data caching method and computer system |
WO2015169245A1 (en) * | 2014-05-09 | 2015-11-12 | 华为技术有限公司 | Data caching method, cache and computer system |
CN105094686B (en) * | 2014-05-09 | 2018-04-10 | 华为技术有限公司 | Data cache method, caching and computer system |
CN104077241B (en) * | 2014-07-14 | 2017-10-17 | 华为技术有限公司 | Cache life cycle algorithm switching handling method and device |
CN104572502B (en) * | 2015-01-12 | 2018-06-19 | 浪潮电子信息产业股份有限公司 | Self-adaptive method for cache strategy of storage system |
CN104572502A (en) * | 2015-01-12 | 2015-04-29 | 浪潮电子信息产业股份有限公司 | Self-adaptive method for cache strategy of storage system |
CN104731980A (en) * | 2015-04-17 | 2015-06-24 | 吉林大学 | Method and device for management of cache pages |
CN105095354A (en) * | 2015-06-19 | 2015-11-25 | 北京奇虎科技有限公司 | Data updating method and data updating device |
CN105512051A (en) * | 2015-12-16 | 2016-04-20 | 鸿秦(北京)科技有限公司 | Self-learning type intelligent solid-state hard disk cache management method and device |
CN105512051B (en) * | 2015-12-16 | 2019-03-12 | 鸿秦(北京)科技有限公司 | A kind of self learning type intelligent solid-state hard disk cache management method and device |
CN105740167B (en) * | 2016-01-28 | 2018-12-07 | 浪潮(北京)电子信息产业有限公司 | A kind of method and system that file system cache is deleted |
CN105740167A (en) * | 2016-01-28 | 2016-07-06 | 浪潮(北京)电子信息产业有限公司 | File system cache deletion method and system |
CN108073527B (en) * | 2016-11-07 | 2020-02-14 | 华为技术有限公司 | Cache replacement method and equipment |
CN108073527A (en) * | 2016-11-07 | 2018-05-25 | 华为技术有限公司 | It is a kind of to cache the method and apparatus replaced |
CN108133267A (en) * | 2016-12-01 | 2018-06-08 | 上海兆芯集成电路有限公司 | With the processor that can be used as most rear class cache tile or the memory array of neural network cell memory operation |
CN108133267B (en) * | 2016-12-01 | 2020-08-21 | 上海兆芯集成电路有限公司 | Processor having memory array operable as last level cache slice or neural network unit memory |
CN106649041A (en) * | 2016-12-27 | 2017-05-10 | 郑州云海信息技术有限公司 | Device and method for automatically testing storage caching mode |
CN110249318A (en) * | 2017-02-07 | 2019-09-17 | 华为技术有限公司 | Cache replacement system and method |
CN106873917A (en) * | 2017-02-24 | 2017-06-20 | 深圳市中博睿存信息技术有限公司 | A kind of object storage system based on disk |
CN107220188A (en) * | 2017-05-31 | 2017-09-29 | 莫倩 | A kind of automatic adaptation cushion block replacement method |
CN107220188B (en) * | 2017-05-31 | 2020-10-27 | 中山大学 | Self-adaptive buffer block replacement method |
US11010295B2 (en) | 2017-09-05 | 2021-05-18 | International Business Machines Corporation | Asynchronous update of metadata tracks in response to a cache hit generated via an i/o operation over a bus interface |
WO2019048969A1 (en) * | 2017-09-05 | 2019-03-14 | International Business Machines Corporation | Asynchronous update of metadata tracks in response to a cache hit generated via an i/o operation over a bus interface |
GB2579754B (en) * | 2017-09-05 | 2020-12-02 | Ibm | Asynchronous update of metadata tracks in response to a cache hit generated via an I/O operation over a bus interface |
US10565109B2 (en) | 2017-09-05 | 2020-02-18 | International Business Machines Corporation | Asynchronous update of metadata tracks in response to a cache hit generated via an I/O operation over a bus interface |
GB2579754A (en) * | 2017-09-05 | 2020-07-01 | Ibm | Asynchronous update of metadata tracks in response to a cache hit generated via an I/O operation over a bus interface |
CN109656832A (en) * | 2017-10-11 | 2019-04-19 | 深圳市中兴微电子技术有限公司 | A kind of look-up method, computer readable storage medium |
CN109298972A (en) * | 2018-09-06 | 2019-02-01 | 高盈量化云科技(深圳)有限公司 | A kind of High Availabitity policy engine |
CN109669882B (en) * | 2018-12-28 | 2021-03-09 | 贵州华芯通半导体技术有限公司 | Bandwidth-aware dynamic cache replacement method, apparatus, system, and medium |
CN109669882A (en) * | 2018-12-28 | 2019-04-23 | 贵州华芯通半导体技术有限公司 | Dynamic caching replacement method, device, system and the medium of bandwidth aware |
CN110032528A (en) * | 2019-04-19 | 2019-07-19 | 苏州浪潮智能科技有限公司 | Internal storage data lookup method, device, equipment and the storage medium of storage system |
CN114072774A (en) * | 2019-07-01 | 2022-02-18 | 国际商业机器公司 | Block mode switching in a data storage system |
CN110992671A (en) * | 2019-12-16 | 2020-04-10 | 宁波三星医疗电气股份有限公司 | Ammeter address forwarding method, device and system and electronic equipment |
CN110992671B (en) * | 2019-12-16 | 2020-11-20 | 宁波三星医疗电气股份有限公司 | Ammeter address forwarding method, device and system and electronic equipment |
CN113342706A (en) * | 2021-05-13 | 2021-09-03 | 武汉大学 | Write-optimized extensible hash index structure based on nonvolatile memory and inserting, refreshing and deleting methods |
WO2023098277A1 (en) * | 2021-12-01 | 2023-06-08 | International Business Machines Corporation | Augmenting cache replacement operations |
US11886342B2 (en) | 2021-12-01 | 2024-01-30 | International Business Machines Corporation | Augmenting cache replacement operations |
Also Published As
Publication number | Publication date |
---|---|
CN101236530B (en) | 2010-09-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101236530B (en) | High speed cache replacement policy dynamic selection method | |
Chen et al. | Hystor: Making the best use of solid state drives in high performance storage systems | |
US8595463B2 (en) | Memory architecture with policy based data storage | |
US6625695B2 (en) | Cache line replacement policy enhancement to avoid memory page thrashing | |
CN1834950B (en) | Multicore processor having active and inactive execution cores | |
US8200902B2 (en) | Cache device for coupling to a memory device and a method of operation of such a cache device | |
CN101221534B (en) | Method and system for controlling computer readable memory | |
CN102012791B (en) | Flash based PCIE (peripheral component interface express) board for data storage | |
CN102306503B (en) | Method and system for detecting false capacity memory | |
US20130145095A1 (en) | Melthod and system for integrating the functions of a cache system with a storage tiering system | |
CN102637147A (en) | Storage system using solid state disk as computer write cache and corresponding management scheduling method | |
KR20140098199A (en) | A dram cache with tags and data jointly stored in physical rows | |
CN102023931A (en) | Self-adaption cache pre-fetching method | |
CN110297787B (en) | Method, device and equipment for accessing memory by I/O equipment | |
US8583873B2 (en) | Multiport data cache apparatus and method of controlling the same | |
CN110532200B (en) | Memory system based on hybrid memory architecture | |
CN110795363B (en) | Hot page prediction method and page scheduling method of storage medium | |
CN100377117C (en) | Method and device for converting virtual address, reading and writing high-speed buffer memory | |
CN109213693A (en) | Memory management method, storage system and computer program product | |
Kim et al. | Charge-aware DRAM refresh reduction with value transformation | |
On et al. | FD-buffer: a buffer manager for databases on flash disks | |
Jung et al. | Fass: A flash-aware swap system | |
Zhang et al. | Design and optimization of large size and low overhead off-chip caches | |
CN115735198A (en) | Configuring cache policies of a cache based on combined cache policy tests | |
Tan et al. | APMigration: Improving performance of hybrid memory performance via an adaptive page migration method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20100901 Termination date: 20170130 |