CN106294032B - A kind of caching method and system of the recovery of disk array single-deck - Google Patents
A kind of caching method and system of the recovery of disk array single-deck Download PDFInfo
- Publication number
- CN106294032B CN106294032B CN201610637551.8A CN201610637551A CN106294032B CN 106294032 B CN106294032 B CN 106294032B CN 201610637551 A CN201610637551 A CN 201610637551A CN 106294032 B CN106294032 B CN 106294032B
- Authority
- CN
- China
- Prior art keywords
- data block
- priority
- caching
- restoration methods
- deck
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/07—Responding to the occurrence of a fault, e.g. fault tolerance
- G06F11/16—Error detection or correction of the data by redundancy in hardware
- G06F11/20—Error detection or correction of the data by redundancy in hardware using active fault-masking, e.g. by switching out faulty elements or by switching in spare elements
- G06F11/2017—Error detection or correction of the data by redundancy in hardware using active fault-masking, e.g. by switching out faulty elements or by switching in spare elements where memory access, memory control or I/O control functionality is redundant
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The caching method and system restored the invention discloses a kind of disk array single-deck is the different priority of definition data block this method comprises: reading in caching according to whether data block has also been employed that；Data block for each error enumerates all feasible restoration methods, is traversed in conjunction with the relationship between restoration methods by Di Jiesitela ergodic algorithm for all feasible restoration methods, selects that I/O is the smallest, satisfactory restoration methods；For selected restoration methods, the correcting and eleting codes restoration methods that the method for wherein loading relative equilibrium is restored as the disk array single-deck finally used are chosen；It is definition data block priority according to the reading times of correcting and eleting codes restoration methods, the buffer replacing method of the low data block of priority is preferentially replaced in use in the buffer, through the invention, it is larger to solve I/O in the prior art, the problem of speed is slower, and cache hit rate is low, influences service efficiency.
Description
Technical field
The present invention relates to computer field of storage, the caching method restored more particularly to a kind of disk array single-deck.
Background technique
Distributed memory system, exactly being stored in the multiple equipment in more areas data dispersion.Traditional storage system
All data are stored in the same storage server by system using the storage scheme of centralization, to become whole system
The bottleneck of functional promotion, the problem of being also easy to produce reliability and safety.Along with the promotion of storage demand, this problem is cured
Hair highlights.Distributed memory system utilizes expansible storage organization, stores data in different storage servers, improves
The reliability of data, safety, and it is easy extension.
For the reliability for realizing data in distributed memory system, system needs to recover the number of error from storage array
According to.Generally reach fault-tolerant and error correction target by the way that the method for redundancy check code is added in coding, this method is known as entangling
Delete code (Erasure Codes).There are two types of common correcting and eleting codes, and one is Reed-Solomon correcting and eleting codes, and one is XOR volumes
Code.Correcting and eleting codes need to generate the correcting and eleting codes comprising n block from k block, to allow to restore k block from the subset of n block.
The situation that disk array single-deck restores quickly is restored aiming at the situation of disk error.It is needed along with storage
How the increase asked quickly restores the data block of error, improves the hit rate that caches in recovery process and have become people's concern
Focus.
Summary of the invention
In order to overcome the deficiencies of the above existing technologies, purpose of the present invention is to provide a kind of recoveries of disk array single-deck
Caching method, use the disk array of correcting and eleting codes quickly to carry out single-deck recovery, with solve in existing restoration methods I/O compared with
Greatly, the problem of speed is slower, and cache hit rate is low, influences service efficiency.
In view of the above and other objects, the present invention proposes a kind of caching method that disk array single-deck restores, including as follows
Step:
Step 1 is read in caching according to whether data block has also been employed that, is the different priority of definition data block；
Step 2, the data block for each error enumerates all feasible restoration methods, for all feasible extensive
Compound method is traversed in conjunction with the relationship between restoration methods by Di Jiesitela ergodic algorithm, select I/O it is the smallest,
Satisfactory restoration methods；
Step 3 is chosen and wherein loads the method for relative equilibrium as finally using for selected restoration methods
The correcting and eleting codes restoration methods that disk array single-deck restores.
Step 4, is definition data block priority according to the reading times of correcting and eleting codes restoration methods, and use is excellent in the buffer
The first buffer replacing method of the low data block of replacement priority.
Further, in step 1, higher priority is assigned for data block in the buffer, for not in the buffer
Data block assigns lower priority.
Further, step 2 includes:
All feasible restoration methods are traversed by Di Jiesite pulling method, node at the beginning of establishing one indicates traversal
Initial conditions；
The data block number for the reading that the data block that cyclic query restores an error may need connects as between node
The weight of line；
If necessary to reading data block in the buffer, the reading of this data block will not participate in the calculating of weight, if need
Not in the buffer, the reading of this data block will participate in the calculating of weight to the data block to be read；
The method that the weight of traversal can be preferably minimized by the data block of all errors again can effectively be restored by selecting.
Further, in step 2, in Dijkstra's algorithm ergodic process, in conjunction with the effective ways chosen
Feature deletes ineffective methods in time.
Further, step 3 includes:
Calculate separately the quantity of effective I/O on each piece of disk；
I/O in array on each piece of disk is compared, the wherein maximum disk of I/O is chosen；
The smallest method of I/O number in all methods on the maximum disk of I/O is selected as being the best practice restored.
Further, in step 3, restore under situation for the disk array single-deck of the load relative equilibrium of selection
Priority in the buffer is arranged according to the number that data block is accessed in best practice.
Further, step 4 further comprises:
Data block is put into queue different in caching according to priority, the identical number of priority in the same queue
It is arranged according to root tuber according to lru algorithm, not visited data block will be preferentially replaced at most recently；
Data block in the high queue of priority is such as accessed, then is added in the data block of the low level-one of priority, with
This analogizes；
When space is inadequate in caching, in lowest priority queue not visited data block will be replaced at most recently
It changes, the data block being newly added is added in corresponding queue according to its priority.
In order to achieve the above objectives, the present invention also provides the caching systems that a kind of disk array single-deck restores, comprising:
Priority definition module is read in caching according to whether data block has also been employed that, is that definition data block is different
Priority；
Spider module enumerates all feasible restoration methods for the data block of each error, for all feasible extensive
Compound method is traversed in conjunction with the relationship between restoration methods by Di Jiesitela ergodic algorithm, select I/O it is the smallest,
Satisfactory restoration methods；
Restoration methods choose module, for selected restoration methods, choose the method conduct for wherein loading relative equilibrium
The correcting and eleting codes restoration methods that the disk array single-deck finally used restores；
Replacement module is cached, is definition data block priority according to the reading times of correcting and eleting codes restoration methods, using slow
The buffer replacing method of the low data block of priority is preferentially replaced in depositing.
Further, which restores under situation for the disk array single-deck for the load relative equilibrium chosen
Best practice, the number setting priority in the buffer being accessed according to data block.
Further, which is put into data block in queue different in caching according to priority, together
The identical data block of priority is arranged according to lru algorithm in one queue, and not visited data block will be excellent at most recently
First it is replaced, the data block in the high queue of priority is such as accessed, then it is added in the data block of the low level-one of priority, with
This analogizes；When space is inadequate in caching, in lowest priority queue not visited data block will be replaced at most recently,
The data block being newly added is added in corresponding queue according to its priority.
Compared with prior art, the caching method and system that a kind of disk array single-deck of the present invention restores propose a kind of new
The situation restored of error correcting code single-deck error, effectively connected applications can read the data block in caching and formulate corresponding restore
Strategy, the speed due to reading the data block in caching are significantly improved in speed than the data block in reading disk, this
Invention can effectively reduce the resume speed under error correcting code single-deck recovery situation, while guarantee being normally carried out for application access.?
Service rate and efficiency can be effectively promoted under the storage server environment that access request is more, task is intensive, are promoted distribution and are deposited
The reliability and safety of storage system.
Detailed description of the invention
Fig. 1 is the step flow chart for the caching method that a kind of disk array single-deck of the present invention restores；
Fig. 2 illustrates to generate all methods for restoring specified data block according to generator matrix in the specific embodiment of the invention；
Fig. 3 shows that Di Jiesitela in the specific embodiment of the invention traverses the process of all possible restoration methods；
Fig. 4 is shown in the specific embodiment of the invention in Di Jiesitela ergodic process, deletes ineffective methods, is improved and is calculated
The process of speed；
Fig. 5 illustrates the mistake of the restoration methods selection best practice obtained in the specific embodiment of the invention according to ergodic algorithm
Journey；
Fig. 6 illustrates the method for adding new data block in the specific embodiment of the invention in caching；
Fig. 7 illustrates the process that data block degrades after being accessed in the specific embodiment of the invention；
Fig. 8 illustrates the method that data block replaces selection in caching in the specific embodiment of the invention；
Fig. 9 is the system architecture diagram for the caching system that a kind of disk array single-deck of the present invention restores.
Specific embodiment
Below by way of specific specific example and embodiments of the present invention are described with reference to the drawings, those skilled in the art can
Understand further advantage and effect of the invention easily by content disclosed in the present specification.The present invention can also pass through other differences
Specific example implemented or applied, details in this specification can also be based on different perspectives and applications, without departing substantially from
Various modifications and change are carried out under spirit of the invention.
Fig. 1 is the step flow chart for the caching method that a kind of disk array single-deck of the present invention restores.As shown in Figure 1, this hair
The caching method that a kind of bright disk array single-deck restores, includes the following steps:
Step 101, it is read in caching according to whether data block has also been employed that, is different preferential of definition data block
Grade.Specifically, higher priority is assigned for data block in the buffer, is assigned for data block not in the buffer lower excellent
First grade.
That is, when system detection to storage server when the error occurs, read data block coding in the buffer first
Number, traversal hour hands to data block whether in the buffer, preferentially select to have read the data in caching, secondary strategic point selection
The data being not yet read into caching, carry out the comprehensive consideration of restoration methods
Step 102, pass through Di Jiesitela in conjunction with the relationship between restoration methods for all possible restoration methods
Ergodic algorithm is traversed, and selects that I/O is the smallest, satisfactory restoration methods.Preferably, in Dijkstra's algorithm time
During going through, in conjunction with the effective ways feature chosen, ineffective methods are deleted in time, calculating speed is improved, avoids invalid time
It goes through.
Specifically, step 102 further comprises:
All feasible restoration methods are traversed by Di Jiesite pulling method, node at the beginning of establishing one indicates traversal
Initial conditions, i.e., all data blocks are still unselected.Cyclic query restores the reading that the data block of an error may need
Weight of the data block number taken as line between node；
If necessary to reading data block in the buffer, the reading of this data block will not participate in the calculating of weight；
On the contrary, not in the buffer if necessary to the data block that reads, the reading of this data block will participate in the calculating of weight.
The target of Di Jiesite pulling method traversal is to select one kind can effectively restore the data block and energy of all errors
The method that the weight of traversal is preferably minimized.Here weight is directly proportional to the I/O in realization.
Further, the method that the present invention traverses Di Jiesite pulling method is improved.For having stepped through to obtain
Data, such as minimal weight, the part that residue is not traversed needs to be considered according to what weight was traversed.For not having
The ergodic part for being likely to be obtained smaller weight is timely given up, comprising:
It can restore the method for all error data blocks, if the minimal weight exceeded, it is invalid that this method is considered
Method；
The method of all error data blocks can not yet be restored, if its weight has had exceeded known minimal weight, that
This method and the method that can restore all error data blocks according to derived from this method, are all considered invalid side
Method
Step 103, it for selected restoration methods, chooses and wherein loads the method for relative equilibrium as finally using
The correcting and eleting codes restoration methods that disk array single-deck restores.
Specifically, step 103 further comprises:
Calculate separately the quantity of effective I/O on each piece of disk, that is, the number for the data block not in the buffer for needing to read
Amount；
I/O in array on each piece of disk is compared, the wherein maximum disk of I/O is chosen；
The smallest method of I/O number in all methods on the maximum disk of I/O is selected as being the best practice restored.
Step 4, is definition data block priority according to the reading times of correcting and eleting codes restoration methods, and use is excellent in the buffer
The first buffer replacing method of the low data block of replacement priority.
Further, restore the best practice under situation, root for the disk array single-deck of the load relative equilibrium of selection
Priority in the buffer is set according to the number that data block is accessed, by taking the fault-tolerant disk array of three disks as an example, access times are
Three data block has highest priority, and it is the number that access times are one again that followed by access times, which are two data block,
According to block.It is specific:
Data block is put into queue different in caching according to priority, the identical number of priority in the same queue
It is arranged according to root tuber according to lru algorithm, not visited data block will be preferentially replaced at most recently；
Data block in the high queue of priority is such as accessed, then is added in the data block of the low level-one of priority, with
This analogizes；
When space is inadequate in caching, in lowest priority queue not visited data block will be replaced at most recently
It changes, the data block being newly added is added in corresponding queue according to its priority.
Below in conjunction with Fig. 2-Fig. 8 by taking the recovery process under the single-deck error situation of an error correcting code as an example, detailed description
Process of the invention:
Step S1: caching data block is obtained to be defined with priority；For the different priority of the definition data block before restoring；Root
It is read in caching according to whether data block has also been employed that, is the different priority of definition data block, number in the buffer
Assign higher priority according to block, be preferred, the data block not read in caching assigns lower priority, not by
Preferential selection.Defaulting the data block to have malfunctioned will not be pre-read into caching, if had issued using to the data block of error
Request, then need to make requests processing again after restoring requested error data block.
Step S2: data block restoration methods are enumerated.It is enumerated by the data block that generator matrix is each error all
Recovery equation list；The example that Fig. 2 is enumerated as data block restoration methods utilizes generator matrix progress piece to illustrate
The process of act.Assuming that there is mistake in the D0 and D1 in disk array, to restore D0, first row corresponding to longitudinal searching D0,
It can be found that R4 and R6 take part in the calculating of D0, therefore it can be used for restoring data block D0.Similarly R5, R7 can be used for restoring data
Block D1.
Step S3: restoration methods traversal.Each of disk array data block Boolean 0,1 is indicated.0 indicates extensive
It does not need to use the data block during multiple, needs to use the data block in 1 expression recovery process.If by data number of blocks
Boolean forms a node in traversing graph, and start node is complete zero, and expression is used without data block at this time, is not also counted
It can be resumed according to block.One node layer of every increase increases by one piece of data block that can be resumed.Start node is full 0 in Fig. 3,
4 kinds of feasible restoration methods are enumerated to restore data block D0.The shortest node in path is therefrom selected, D1 is further restored.
When all data blocks can effectively be restored, choosing wherein is shortest path apart from shortest ID of trace route path.It is connecing
In the traversal got off, all nodes beyond this path length are considered as invalid node, are not involved in subsequent traversal and search.Fig. 3
The node that the middle path length restoring D0 and generating is 4, because its total length necessarily has exceeded known shortest path, therefore is regarded
For invalid node, all nodes thereafter are not involved in traversal operation.Confirm path length the step of include:
It is zero that original path length, which is arranged,；
Increased data block between two nodes of label；If increased data block is in the buffer, path length is kept
It is constant；If not in the buffer, path length adds one to increased data block；
Defaulting increased data block is the readable data block for not occurring mistake；If there is mistake in increased data block
Accidentally, newly-increased node is considered as invalid node.
Step S4: restoration methods screening.From by Dijkstra's algorithm traverse with identical shortest path
The method of load relative equilibrium is chosen in restoration methods.Screening criteria is the load minimum for making to load on maximum disk.In Fig. 5
One group and third group only have read a data block in addition to the disk that malfunctions on each disk, load relative equilibrium.Second group in Fig. 5
With the 4th group in addition to the disk that malfunctions, two disks have read two pieces of data blocks, and a disk does not have read block, relative to first,
For three groups, load imbalance.Therefore final method should between first group and third group optional one.First group and
Three groups of restoration methods have symmetry, without point of superiority and inferiority.
Step S5: cache replacement policy.The best practice traversed according to Dijkstra's algorithm, according to recovery process
The accessed number of middle data block sets up different priority；For using the disk array of the fault-tolerant correcting and eleting codes of three disks,
The data block that access times are three has highest priority, is placed in the queue P1 of highest priority in the buffer；Access times
Take second place for two data block priority, is placed in the queue P2 that priority is only second to P1 in the buffer；The data that access times are one
Third block priority, is placed in the minimum queue P3 of priority in the buffer；The data block of inner queue is arranged according to LRU
Column, are connected with each other by pointer, are updated convenient for the arrangement of database relationship.
The replacement policy of caching is divided into two kinds of situations, specifically: when the disk for needing to read is when caching internal, such as Fig. 7 institute
Show the disk for finding and needing to read, the first place of next priority query is downgraded to after reading, the disk such as read is minimum
In the queue of priority, then queue first place is placed it according to LRU rule after reading, does not do downgrade processing；When needing to read
The disk taken when caching internal, does not need to read the data block in disk in caching, and be inserted into corresponding queue, has
Body are as follows:
If there is vacant data block space in caching, reading disk data block to corresponding caching data block, according to
The priority of data block is inserted into corresponding queue；
If not having vacant data block space in caching, the end data block of lowest priority queue, which is replaced with, to be needed
The data block to be read, and be inserted into corresponding queue.
Fig. 9 is the system architecture diagram for the caching system that a kind of disk array single-deck of the present invention restores.As shown in figure 9, this hair
The caching system that a kind of bright disk array single-deck restores, comprising: priority definition module 901, spider module 902, restoration methods
Choose module 903, caching replacement module 904.
Priority definition module 901 is read in caching according to whether data block has also been employed that, not for definition data block
Same priority.Specifically, higher priority is assigned for data block in the buffer, is assigned for data block not in the buffer
Lower priority.That is, when system detection to storage server when the error occurs, read data in the buffer first
Block number, traversal hour hands to data block whether in the buffer, preferentially select to have read the data in caching, secondary strategic point
Selection is not yet read into the data in caching, carries out the comprehensive consideration of restoration methods
Spider module 902 enumerates all possible restoration methods for the data block of each error, for all possible
Restoration methods are traversed in conjunction with the relationship between restoration methods by Di Jiesitela ergodic algorithm, and I/O minimum is selected
, satisfactory restoration methods.Preferably, having efficacious prescriptions in conjunction with what is chosen in Dijkstra's algorithm ergodic process
Method feature deletes ineffective methods in time, improves calculating speed, avoids traversing in vain.Specifically, spider module 902 passes through enlightening
Jie Site pulling method traverses all feasible restoration methods, and node at the beginning of establishing one indicates the initial conditions of traversal, i.e.,
All data blocks are still unselected.The data block number for the reading that the data block that cyclic query restores an error may need
As the weight of line between node, in the buffer, the reading of this data block will not participate in power to the data block read if necessary
The calculating of weight；On the contrary, not in the buffer if necessary to the data block that reads, the reading of this data block will participate in the meter of weight
It calculates.The target of Di Jiesite pulling method traversal be select a kind of data block that can effectively restore all errors again can will be all over
The method that the weight gone through is preferably minimized.Here weight is directly proportional to the I/O in realization.Preferably, the traversal mould in the present invention
The method that block traverses Di Jiesite pulling method is improved.For the data having stepped through, such as minimal weight, remain
The remaining part not traversed needs to be considered according to what weight was traversed, for not being likely to be obtained time of smaller weight
Part is gone through timely to be given up, comprising:
It can restore the method for all error data blocks, if the minimal weight exceeded, it is invalid that this method is considered
Method；
The method of all error data blocks can not yet be restored, if its weight has had exceeded known minimal weight, that
This method and the method that can restore all error data blocks according to derived from this method, are all considered invalid side
Method
Restoration methods choose module 903, for selected restoration methods, choose the method for wherein loading relative equilibrium and make
The correcting and eleting codes restoration methods restored for the disk array single-deck finally used.
Specifically, restoration methods selection module 903 further comprises:
Calculate separately the quantity of effective I/O on each piece of disk, that is, the number for the data block not in the buffer for needing to read
Amount；
I/O in array on each piece of disk is compared, the wherein maximum disk of I/O is chosen；
The smallest method of I/O number in all methods on the maximum disk of I/O is selected as being the best practice restored.
Cache replacement module 904, according to the reading times of correcting and eleting codes restoration methods be definition data block priority, using
The buffer replacing method of the low data block of priority is preferentially replaced in caching.
Replacement module 904 is cached for the optimal side under the disk array single-deck recovery situation for the load relative equilibrium chosen
Priority in the buffer is arranged according to the number that data block is accessed in method.By taking the fault-tolerant disk array of three disks as an example, access time
Number for three data block have highest priority, followed by access times be two data block, be again access times be one
Data block.The replacement policy for caching replacement module is specific as follows:
Data block is put into queue different in caching according to priority, the identical number of priority in the same queue
It is arranged according to root tuber according to lru algorithm, not visited data block will be preferentially replaced at most recently；
Data block in the high queue of priority is such as accessed, then is added in the data block of the low level-one of priority, with
This analogizes；
When space is inadequate in caching, in lowest priority queue not visited data block will be replaced at most recently
It changes, the data block being newly added is added in corresponding queue according to its priority.
In conclusion caching method and system that a kind of disk array single-deck of the present invention restores propose a kind of new error correction
The situation that code single-deck error restores effectively connected applications can read the data block in caching and formulate and corresponding restore plan
Slightly, since the speed for reading the data block in caching is significantly improved in speed than the data block in reading disk, this hair
The bright resume speed that can be effectively reduced under error correcting code single-deck recovery situation, while guaranteeing being normally carried out for application access.It is visiting
It asks and requests effectively promote service rate and efficiency under the storage server environment more, task is intensive, promote distributed storage
The reliability and safety of system.
The above-described embodiments merely illustrate the principles and effects of the present invention, and is not intended to limit the present invention.Any
Without departing from the spirit and scope of the present invention, modifications and changes are made to the above embodiments by field technical staff.Therefore,
The scope of the present invention, should be as listed in the claims.
Claims (9)
1. the caching method that a kind of disk array single-deck restores, includes the following steps:
Step 1 is read in caching according to whether data block has also been employed that, is that the different access characteristics of definition data block are excellent
First grade assigns higher access characteristics priority for data block in the buffer, assigns for data block not in the buffer lower
Access characteristics priority；
Step 2, the data block for each error enumerates all feasible restoration methods, for all feasible recovery sides
Method is traversed in conjunction with the relationship between restoration methods by Di Jiesitela ergodic algorithm, selects that I/O is the smallest, meets
It is required that restoration methods；
Step 3 is chosen for selected restoration methods and wherein loads the method for relative equilibrium as the disk finally used
The correcting and eleting codes restoration methods that array single-deck restores；
Step 4 is definition data block recovery algorithms priority according to the reading times of correcting and eleting codes restoration methods, using caching
In preferentially replace the buffer replacing method of the low data block of recovery algorithms priority.
2. the caching method that a kind of disk array single-deck as described in claim 1 restores, which is characterized in that step 2 includes:
All feasible restoration methods are traversed by Di Jiesite pulling method, node at the beginning of establishing one indicates traversal most
Initial equilibrium state；
The data block number for the reading that the data block that cyclic query restores an error may need is as line between node
Weight；
If necessary to reading data block in the buffer, the reading of this data block will not participate in the calculating of weight, if necessary to read
Not in the buffer, the reading of this data block will participate in the calculating of weight to the data block taken；
The method that the weight of traversal can be preferably minimized by the data block of all errors again can effectively be restored by selecting.
3. the caching method that a kind of disk array single-deck as claimed in claim 2 restores, it is characterised in that: in step 2,
In Dijkstra's algorithm ergodic process, in conjunction with the effective ways feature chosen, ineffective methods are deleted in time.
4. the caching method that a kind of disk array single-deck as claimed in claim 3 restores, it is characterised in that: step 3 includes:
Calculate separately the quantity of effective I/O on each piece of disk；
I/O in array on each piece of disk is compared, the wherein maximum disk of I/O is chosen；
The smallest method of I/O number in all methods on the maximum disk of I/O is selected as being the best practice restored.
5. the caching method that a kind of disk array single-deck as claimed in claim 4 restores, it is characterised in that: in step 3,
Restore the best practice under situation for the disk array single-deck of the load relative equilibrium of selection, time being accessed according to data block
The access characteristics priority of number setting in the buffer.
6. the caching method that a kind of disk array single-deck as claimed in claim 5 restores, which is characterized in that step 4 is further
It include: that data block is put into queue different in caching according to recovery algorithms priority, recovery algorithms in the same queue
The identical data block of priority is arranged according to lru algorithm, and not visited data block will be preferentially replaced at most recently；
Data block in the high queue of recovery algorithms priority is such as accessed, then is added to the number of the low level-one of recovery algorithms priority
In block, and so on；
When space is inadequate in caching, in minimum recovery algorithms priority query not visited data block will be by most recently
Replacement, the data block being newly added are added in corresponding queue according to its recovery algorithms priority.
7. the caching system that a kind of disk array single-deck restores, comprising:
Access characteristics priority definition module is read in caching according to whether data block has also been employed that, is definition data block
Different access characteristics priority；
Spider module enumerates all feasible restoration methods for the data block of each error, for all feasible recovery sides
Method is traversed in conjunction with the relationship between restoration methods by Di Jiesitela ergodic algorithm, selects that I/O is the smallest, meets
It is required that restoration methods；
Restoration methods choose module, for selected restoration methods, choose the method for wherein loading relative equilibrium as final
The correcting and eleting codes restoration methods that the disk array single-deck of use restores；
Replacement module is cached, is definition data block recovery algorithms priority according to the reading times of correcting and eleting codes restoration methods, uses
The buffer replacing method of the low data block of recovery algorithms priority is preferentially replaced in the buffer.
8. the caching system that a kind of disk array single-deck as claimed in claim 7 restores, it is characterised in that: the caching replaces mould
Block restores the best practice under situation for the disk array single-deck for the load relative equilibrium chosen, and is accessed according to data block
The access characteristics priority of number setting in the buffer.
9. the caching system that a kind of disk array single-deck as claimed in claim 8 restores, it is characterised in that: the caching replaces mould
Data block is put into queue different in caching by root tuber according to recovery algorithms priority, and recovery algorithms are preferential in the same queue
The identical data block of grade is arranged according to lru algorithm, and not visited data block will preferentially be replaced at most recently, restores calculation
Data block in the high queue of method priority is such as accessed, then is added in the data block of the low level-one of recovery algorithms priority,
And so on；Not visited data at most recently when space is inadequate in caching, in minimum recovery algorithms priority query
Block will be replaced, and the data block being newly added is added in corresponding queue according to its recovery algorithms priority.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610637551.8A CN106294032B (en) | 2016-08-05 | 2016-08-05 | A kind of caching method and system of the recovery of disk array single-deck |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610637551.8A CN106294032B (en) | 2016-08-05 | 2016-08-05 | A kind of caching method and system of the recovery of disk array single-deck |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106294032A CN106294032A (en) | 2017-01-04 |
| CN106294032B true CN106294032B (en) | 2019-06-28 |
Family
ID=57665495
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610637551.8A Active CN106294032B (en) | 2016-08-05 | 2016-08-05 | A kind of caching method and system of the recovery of disk array single-deck |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106294032B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101135955A (en) * | 2007-10-19 | 2008-03-05 | 杭州华三通信技术有限公司 | Method for realizing magnetic disc redundant array rebuilding and magnetic disc redundant array |
| CN103744750A (en) * | 2014-01-13 | 2014-04-23 | 杭州华为数字技术有限公司 | Method and device for recovering data |
| CN104035830A (en) * | 2014-06-24 | 2014-09-10 | 浙江宇视科技有限公司 | Method and device for recovering data |
| CN104536698A (en) * | 2014-12-10 | 2015-04-22 | 华为技术有限公司 | Disk reconfiguration method based on RAID and related apparatus |
| CN104866433A (en) * | 2015-05-31 | 2015-08-26 | 上海交通大学 | Multi-level caching method based on historical information |
| CN105045894A (en) * | 2015-07-31 | 2015-11-11 | 中国科学院计算技术研究所 | Cache method and system oriented to distributed sequence list |
| WO2016048314A1 (en) * | 2014-09-24 | 2016-03-31 | Hewlett Packard Enterprise Development Lp | Block priority information |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8726070B2 (en) * | 2010-09-27 | 2014-05-13 | Dell Products L.P. | System and method for information handling system redundant storage rebuild |
-
2016
- 2016-08-05 CN CN201610637551.8A patent/CN106294032B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101135955A (en) * | 2007-10-19 | 2008-03-05 | 杭州华三通信技术有限公司 | Method for realizing magnetic disc redundant array rebuilding and magnetic disc redundant array |
| CN103744750A (en) * | 2014-01-13 | 2014-04-23 | 杭州华为数字技术有限公司 | Method and device for recovering data |
| CN104035830A (en) * | 2014-06-24 | 2014-09-10 | 浙江宇视科技有限公司 | Method and device for recovering data |
| WO2016048314A1 (en) * | 2014-09-24 | 2016-03-31 | Hewlett Packard Enterprise Development Lp | Block priority information |
| CN104536698A (en) * | 2014-12-10 | 2015-04-22 | 华为技术有限公司 | Disk reconfiguration method based on RAID and related apparatus |
| CN104866433A (en) * | 2015-05-31 | 2015-08-26 | 上海交通大学 | Multi-level caching method based on historical information |
| CN105045894A (en) * | 2015-07-31 | 2015-11-11 | 中国科学院计算技术研究所 | Cache method and system oriented to distributed sequence list |
Non-Patent Citations (2)
| Title |
|---|
| Balancing I/O Response Time and Disk Rebuild Time in a RAID5 Disk Array;Robert Y.Hou;《Proceedings of the Twenty-sixth Hawaii International Conference on System Sciences》;19930108;全文 |
| 一种提高磁盘阵列重建效率的缓存替换方法;毛波 等;《华中科技大学学报（自然科学版）》;20110630;全文 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106294032A (en) | 2017-01-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7374939B2 (en) | distributed storage system | |
| CN108733518B (en) | Method, apparatus, and computer-readable medium for managing a storage system | |
| CN103942157B (en) | Method and system for calculating the data processing in storage environment | |
| CN104937574B (en) | The method and system of data processing in storage environment is calculated | |
| US10901619B2 (en) | Selecting pages implementing leaf nodes and internal nodes of a data set index for reuse | |
| CN105630423B (en) | A kind of correcting and eleting codes cluster-based storage expansion method based on data buffer storage | |
| US9690657B2 (en) | Writing data across storage devices in an erasure-coded system | |
| US9424132B2 (en) | Adjusting dispersed storage network traffic due to rebuilding | |
| US10642690B1 (en) | Policy-driven raid rebuild | |
| Neglia et al. | Access-time-aware cache algorithms | |
| CN110262922A (en) | Correcting and eleting codes update method and system based on copy data log | |
| CN108021717B (en) | Method for implementing lightweight embedded file system | |
| CN103345472A (en) | Redundancy removal file system based on limited binary tree bloom filter and construction method of redundancy removal file system | |
| CN103907100A (en) | Dynamically adjusted threshold for population of secondary cache | |
| US10592365B2 (en) | Method and apparatus for managing storage system | |
| US10061517B2 (en) | Apparatus and method for data arrangement | |
| CN105404595B (en) | Buffer memory management method and device | |
| CN103688249A (en) | Adaptive record caching for solid state disks | |
| US20100106808A1 (en) | Replica placement in a distributed storage system | |
| CN107133228A (en) | A kind of method and device of fast resampling | |
| US11687272B2 (en) | Method and system for dynamic topology-aware space allocation in a distributed system | |
| CN104503923A (en) | Asymmetrical disk array caching dispatching method | |
| US20170168956A1 (en) | Block cache staging in content delivery network caching system | |
| CN106294032B (en) | A kind of caching method and system of the recovery of disk array single-deck | |
| CN106547484A (en) | It is a kind of that internal storage data reliability method and system realized based on RAID5 |
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 |