CN101251789A - Cheap magnetic disc redundant array RAID5 roll rapid capacitance enlarging method - Google Patents
Cheap magnetic disc redundant array RAID5 roll rapid capacitance enlarging method Download PDFInfo
- Publication number
- CN101251789A CN101251789A CNA2008101028935A CN200810102893A CN101251789A CN 101251789 A CN101251789 A CN 101251789A CN A2008101028935 A CNA2008101028935 A CN A2008101028935A CN 200810102893 A CN200810102893 A CN 200810102893A CN 101251789 A CN101251789 A CN 101251789A
- Authority
- CN
- China
- Prior art keywords
- data
- metadata
- data block
- window
- raid5
- 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.)
- Pending
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention discloses a redundant array of inexpensive disks RAID5 volume quick capacity-expanding method, belonging to the information storing technical field, and characterized in that: changing a logic sequence of data redistribution I/O into a physical sequence; accessing a plurality of physically continuous data blocks via a single I/O, so that the size of the data redistribution I/O is increased, and the number of the data redistribution I/O is reduced. The metadata is idly updated and the metadata is updated until the menace of the data reliability appears. The update times of the metadata are reduced prominently, while the reliability of the data is ensured strictly. All loads of the RAID5 volume are monitored; the data migration speed is adjusted and controlled by adaptively using a simple logic valve. The two requirements of data redistribution speed and applying I/O performance can be comprehensively and adaptively satisfied. Experiments show that in the condition of a plurality of real loads, compared with the traditional methods, the method of the present invention is prominently improved in two performance indexes of user response time and data redistribution time.
Description
Technical field
Cheap magnetic disc redundant array RAID 5 roll rapid capacitance enlarging method belongs to the computer information storage technology field, relates in particular to the high-performance of peripheral storage device, highly reliable, the extendible optimization of height organizational technology field.
Background technology
The dilatation of RAID5 volume is to point in the existing RAID5 volume to increase disk, so both can improve its I/O performance, also can enlarge its memory capacity.The dilatation of RAID5 volume makes the RAID5 volume can dynamically satisfy application demand better.The machine cost is high because delay, and the RAID5 in the dilatation rolls up necessary heavy distributed data on one side, on one side processing application I/O request.Data heavily distribute and use that I/O shares even the I/O resource of contention system, make that carrying out the dilatation of RAID5 volume is a technological challenge.
Have only a few RAID5 volume expansion method at present, there is following several common defects in they: the first, they initiate very little heavily distribution I/O, cause data heavily to distribute and the phase mutual interference of using between the I/O very big; The second, be the assurance data consistency, their frequent updating metadata, and then cause frequent tracking expense; The 3rd, they use the rate controlled of non-self-adapting, or greatly reduce application performance, or seriously contain the speed that data heavily distribute.These defectives cause existing RAID5 volume expansion method efficient not high, directly and have seriously influenced performance, extendability and the reliability of RAID5 volume.
Based on the heavy characteristic distributions of data in the analyzing RAID 5 volume dilation process, a whole set of new RAID5 volume rapid volume-expanding method has been proposed, solved the problems referred to above effectively.
Summary of the invention
The object of the present invention is to provide a high efficiency RAID5 volume expansion method that can satisfy the heavy distribution speed target of data comprehensively and use the I/O performance requirement, realize the online quick dilatation of RAID5 volume.Emphasis of the present invention is: data heavily the distribute flowcollector aggregation scheme FlowCollector of I/O, the lazy renewal technology of metadata and the method for rate control of logic-based valve.
The invention is characterized in: heavily distribute I/O polymerization visit of data, do not change the migration data amount, but significantly reduce the I/O number that heavily distributes; Metadata is lazy to be upgraded, and guarantees data reliability, significantly reduces the metadata updates number of times simultaneously; Monitor application load and, guarantee to use the I/O performance on the statistics with the heavy distribution speed of logical valve control, and the heavy distribution speed of dynamic maximum data.
Heavily the distribute general practice of I/O of data is to carry out according to succession in logic, that is: data block 0, data block 1, and data block 2 ..., and no matter how these data blocks distribute on disk.Consider the distribution characteristics of data block on disk, this method is changed into the order of physical distribution on disk with the data I/O that heavily distributes by order in logic, by a plurality of continuous data block physically of single I/O visit, the size of I/O reduces its number simultaneously thereby the increase data heavily distribute.
Existing RAID5 volume expansion method is in order to guarantee data reliability frequent updating metadata, or whenever moves a data block and upgrade a metadata, or whenever moves a small data window and upgrade a metadata.This method weighs distributed data and non-update metadata constantly, ability update metadata when the threat to the data reliability occurs.On the one hand, the laziness of metadata is upgraded the update times that has significantly reduced metadata, on the other hand, and data reliability that the laziness of metadata has been upgraded strict guarantee.
Existing expansion method adopts the rate controlled of non-self-adapting, is difficult to satisfy the heavy distribution speed of data comprehensively and uses two demands of I/O performance.The load that this method monitoring RAID5 volume is born is regulated and control data migtation speed with a simple logic valve adaptively.When load is light, carry out data and heavily distribute, and when application load is heavier, strangle data heavily to distribute.Can satisfy the heavy distribution speed of data so adaptively comprehensively and use two demands of I/O performance.
Described method is to realize according to the following steps successively on main frame that the Redundant Array of Inexpensive Disc function is arranged or array system:
Step (1): the polymerization visit of data block
Step (1.1): with the asynchronous I/O mode of operation simultaneously by single I/O operation read a plurality of on m piece disk physical distribution continuous data block, and deposit internal memory in;
Step (1.2): n piece disk add a RAID5 who forms by m piece disk roll up in the time, adjust the data block pointer and realize the redistribution of data block: before the dilatation, a data block x is distributed on the x/m piece physical data block of x mod m disk, is distributed in after the dilatation on x/ (m+n) the piece physical data block of x mod (m+n) number disk;
Step (1.3): the checking data piece of the data organization that COMPUTER CALCULATION is new;
Step (1.4): be written in continuous data block on the described physical distribution of step (1.2) by single I/O operation simultaneously with the asynchronous I/O mode of operation;
Step (2): the laziness of metadata is upgraded
Step (2.1): in case all data blocks in the polymerization window have all moved, the polymerization window is to polymerization window size of front slide, described polymerization window is meant the data block set of reading into memory by one group of asynchronous I/O operation, and moving window is used for describing the map information of RAID5 volume the preceding paragraph continuous space;
Step (2.2): metadata is not updated, when one of following two states occur, and update metadata:
When the polymerization window arrives current moving window border, new moving window of update metadata and initialization,
When address area of user's write request arrival, when metadata is not upgraded as yet but the data of described address area have moved, update metadata, and new moving window of initialization, this moment is the service-user write request;
Step (2.3): all data blocks in first polymerization window all are moved, and the heavy distribution of data switches to original state;
Step (3): the rate controlled of logic-based valve
Step (3.1): monitor the application I/O load that described RAID5 volume is born;
Step (3.2): in the time cycle of a setting, when system takes place congestedly, the application I/O that perhaps surpasses critical value service time accounts in this cycle all ratios of using I/O and judges then that greater than the number percent of setting present load is high capacity, otherwise, be low load;
Step (3.3): when the current period internal burden was light, following one-period carried out data and heavily distributes, and when the current period internal burden was attached most importance to, following one-period did not carry out data and heavily distributes.
Idiographic flow is seen Fig. 8.
Advantage of the present invention is as follows:
(1) heavily distribute I/O polymerization visit of data does not change the migration data amount, but significantly reduces the I/O number that heavily distributes, and has effectively increased the data throughput of the I/O that heavily distributes;
(2) metadata is lazy upgrades, and the assurance data reliability significantly reduces the metadata updates number of times simultaneously, reduces the influence to application I/O, and reduces the cost that data heavily distribute;
(3) monitoring application load and with the heavy distribution speed of logical valve control guarantees to use the I/O performance statistics on, and the dynamic heavy distribution speed of maximum data, can satisfy heavy distribution speed target of data and application I/O performance requirement comprehensively;
(4) except the heavy distribution efficiency of higher data, this method also provides the better application performance, and reason is that it has shortened disk queue length and significantly alleviated data and heavily distributes to using the interference of I/O.
The present invention realizes and tests in department of computer science, Tsinghua university high-performance calculation technical institute.We adopt RAID5 volume rapid volume-expanding method to realize a prototype system ALV.Performance evaluation is to finish by the RAID5 dilatation instrument MD-RESHAPE that comparison the present invention and linux kernel 2.6.18 carry.MD-RESHAPE is present most effective RAID5 expansion method.Performance is relatively carried out from subscriber response time and two indexs of the heavy distribution time of data.The result shows that under multiple real load, the present invention is having very significant raising on two performance index than classic method on subscriber response time and the heavy distribution time of data.
Test is as follows with server configures: Intel Xeon 2.4GHz processor; The 1GB internal memory; Emulex LP982 optical channel card; SEAGATE ST3146807FC hard disk, capacity are 146GB.(SuSE) Linux OS, kernel version 2 .6.18 are adopted in experiment.Experiment has used the trace player of department of computer science, Tsinghua university high-performance calculation technical institute exploitation as testing tool, uses disclosed three true trace:TPC-C, Cello-99 and SPC-web.Test result is seen Fig. 5, Fig. 6, Fig. 7 respectively.
From test result as can be seen: as Fig. 5, under the TPC-C load, this method improves 73.91% than classic method on subscriber response time, improves 25.43% on the heavy distribution time of data; As Fig. 6, under the Cello-99 load, this method improves 59.9% than classic method on subscriber response time, improves 29.27% on the heavy distribution time of data; As Fig. 7, under the SPC-web load, this method improves 53.31% than classic method on subscriber response time, improves 24.07% on the heavy distribution time of data.
Description of drawings
Fig. 1. the polymerization read operation of data block.
Fig. 2. the polymerization write operation of data block.
Fig. 3. the state transition graph that the laziness of metadata is upgraded.
The polymerization window
Fig. 4. the state transition graph of the rate controlled of logic-based valve is described.
Performance under Fig. 5 .TPC-C load between MD-RESHAPE and the ALV relatively.
Zero: traditional expansion method
●: RAID5 rolls up rapid volume-expanding method
Performance under Fig. 6 .Cello-99 load between MD-RESHAPE and the ALV relatively.
Zero: traditional expansion method
●: RAID5 rolls up rapid volume-expanding method
Performance under Fig. 7 .SPC-web load between MD-RESHAPE and the ALV relatively.
Zero: traditional expansion method
●: RAID5 rolls up rapid volume-expanding method
The FB(flow block) of Fig. 8 .RAID5 volume rapid volume-expanding method ALV.
Embodiment
RAID5 volume rapid volume-expanding method can be implemented in any system of RAID function is arranged, and includes but not limited to host computer system and array system.Described RAID5 volume rapid volume-expanding method contains following steps:
Step 1: the polymerization visit of data block
Step 1.1: change the read-write order of data block, by single I/O operation read a plurality of on the disk physical distribution continuous data block.With the heavy distribution of data shown in Figure 1 is example.Here be that a disk is added a RAID5 volume of being made up of three disks.This method is sent first I/O and is asked read block 18,20, and 24 and 26; Send second I/O and ask read block 19,22 and 25; Send the 3rd I/O and ask read block 21 and 23.These three requests are sent simultaneously by the asynchronous I/O operation.Like this,, only need three I/O requests for read block 18-26, rather than nine I/O requests.
Step 1.2: after these nine data blocks are read into internal memory, adjust the data block pointer and realize the redistribution of data block, such as, n piece disk is added during a RAID5 who is made up of m piece disk rolls up, a data block x is distributed on (x/m) piece physical data block of (x mod m) number disk before dilatation, and is being distributed in after the dilatation on (x/ (m+n)) piece physical data block of (x mod (m+n)) number disk;
Step 1.3:CPU calculates the checking data piece of new data organization.This moment, data block and the checking data block organization form in the internal memory was the disk topological structure that is fit to after the dilatation.
Step 1.4: by single I/O write a plurality of in disk physical organization continuous data block.The data example that heavily distributes above continuing, it is a disk to be joined a RAID5 who is made up of three disks roll up.As shown in Figure 2, this method is sent first I/O and is asked writing data blocks 18, checking data piece P7, data block 26; Send second I/O and ask to write checking data piece P6, data block 21 and 25; Send the 3rd I/O and ask writing data blocks 19,22 and 26; Send the 4th I/O and ask writing data blocks 20 and 23, checking data piece P8.These four requests are sent simultaneously by asynchronous I/O.Like this,, only need four I/O requests for writing data blocks 18-26 and three checking data pieces, rather than 12 I/O requests.
Step 2: the laziness of metadata is upgraded
This method is more newly arrived with the laziness of metadata and is minimized metadata write operation number of times, guarantees data reliability simultaneously.Data block is read and write, and metadata is not updated; When the incident that threatens data reliability took place, metadata just was updated.Fig. 3 has described the state transition graph of the laziness renewal of metadata.Wherein, the polymerization window is meant the data block set of reading into memory by one group of asynchronous I/O operation, and moving window is used for describing the map information of RAID5 volume the preceding paragraph continuous space.
Step 2.1: in case all data blocks in the polymerization window have all moved, the polymerization window is to window size of front slide (state 0).
Step 2.2: metadata is not updated; When one of following two states occur: when (1) arrives current moving window border when the polymerization window (state 1), the moving window that metadata is updated and is new be initialised (state 2); (2) arrive an address area when user's write request, the there data have moved but metadata when also not upgrading (state 3), and metadata is updated, and a new moving window is initialised, and this moment, user's write request was just serviced.
Step 2.3: at state 2 or state 4, in case all data blocks in first polymerization window are moved, the heavy distribution of data switches to state 0.
Step 3: the rate controlled of logic-based valve
Step 3.1: the application I/O load that monitoring RAID5 volume is born.
Step 3.2: judge that current application I/O load is that height is low.In a time cycle, if there is system congestion to take place, the application I/O ratio that perhaps surpasses certain critical value service time judges then that greater than default number percent present load is high capacity.Otherwise be low load.
Step 3.3: regulate and control the heavy distribution speed of data with a logical valve.According to cycle current time internal burden weight, determine whether in next time cycle, carrying out data and heavily distribute.As shown in Figure 4, when the current period internal burden was light, following one-period carried out data and heavily distributes; When otherwise the current period internal burden was attached most importance to, following one-period did not carry out data and heavily distributes.This means that we come the approximate estimation loading condition of following one-period with current period internal burden weight, so that the time cycle can not be provided with is oversize.
RAID5 volume rapid volume-expanding method can be implemented in any system of RAID function is arranged, and includes but not limited to host computer system and array system.Be example with the Linux host computer system below, introduce and how to implement RAID5 volume rapid volume-expanding method:
When a RAID5 rolled up dilatation operation beginning, such as being to increase n piece disk in a RAID5 who is made up of m piece disk rolls up, establishment data mover thread was carried out data and is heavily distributed.Come mark dilatation progress how and the mobile status of RAID5 volume target part with three variablees.
Want heavily to distribute data in the new polymerization window must mail in this polymerization window all by the time and use I/O request and all be done.For the data in the polymerization window that heavily distributes, carry out following three steps successively:
Step 1: send m I/O simultaneously and ask to read all interior data blocks of polymerization window;
Step 2: finish in case all these read I/O, adjust the pointer data block of resequencing, make it to meet the physical organization's structure after the dilatation, and the calculation check data;
Step 3: send m+n I/O simultaneously and ask to write polymerization window interior all data blocks and corresponding check data block;
When the data block in the polymerization window that heavily distributes, any I/O request of attempting to enter this polymerization window all will get clogged, and all data blocks in this polymerization window are all heavily distributed to finish.In case all interior data blocks of polymerization window are all heavily distributed and are finished, and wake all I/O requests of waiting in line at this polymerization window up.
When using the I/O request comes for one, whether the buffer resource of detection system is used up; When application I/O request is finished, check whether this I/O response time surpasses maximum response time.If judge that the load pressure that the RAID5 volume is born in cycle current time is bigger, the data mover thread was slept in the next time cycle; Otherwise the data mover thread is carried out data in the next time cycle and is heavily distributed.In our realization, a time cycle was set to one second.
Claims (1)
1. Cheap magnetic disc redundant array RAID 5 roll rapid capacitance enlarging method is characterized in that: described method is to realize according to the following steps successively on main frame that the Redundant Array of Inexpensive Disc function is arranged or array system:
Step (1): the polymerization visit of data block
Step (1.1): with the asynchronous I/O mode of operation simultaneously by single I/O operation read a plurality of on m piece disk physical distribution continuous data block, and deposit internal memory in;
Step (1.2): n piece disk add a RAID5 who forms by m piece disk roll up in the time, adjust the data block pointer and realize the redistribution of data block: before the dilatation, a data block x is distributed on the x/m piece physical data block of x mod m disk, is distributed in after the dilatation on x/ (m+n) the piece physical data block of x mod (m+n) number disk;
Step (1.3): the checking data piece of the data organization that COMPUTER CALCULATION is new;
Step (1.4): be written in continuous data block on the described physical distribution of step (1.2) by single I/O operation simultaneously with the asynchronous I/O mode of operation;
Step (2): the laziness of metadata is upgraded
Step (2.1): in case all data blocks in the polymerization window have all moved, the polymerization window is to polymerization window size of front slide, described polymerization window is meant the data block set of reading into memory by one group of asynchronous I/O operation, and moving window is used for describing the map information of RAID5 volume the preceding paragraph continuous space;
Step (2.2): metadata is not updated, when one of following two states occur, and update metadata:
When the polymerization window arrives current moving window border, new moving window of update metadata and initialization,
When address area of user's write request arrival, when metadata is not upgraded as yet but the data of described address area have moved, update metadata, and new moving window of initialization, this moment is the service-user write request;
Step (2.3): all data blocks in first polymerization window all are moved, and the heavy distribution of data switches to original state;
Step (3): the rate controlled of logic-based valve
Step (3.1): monitor the application I/O load that described RAID5 volume is born;
Step (3.2): in the time cycle of a setting, when system takes place congestedly, the application I/O that perhaps surpasses critical value service time accounts in this cycle all ratios of using I/O and judges then that greater than the number percent of setting present load is high capacity, otherwise, be low load;
Step (3.3): when the current period internal burden was light, following one-period carried out data and heavily distributes, and when the current period internal burden was attached most importance to, following one-period did not carry out data and heavily distributes.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNA2008101028935A CN101251789A (en) | 2008-03-28 | 2008-03-28 | Cheap magnetic disc redundant array RAID5 roll rapid capacitance enlarging method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNA2008101028935A CN101251789A (en) | 2008-03-28 | 2008-03-28 | Cheap magnetic disc redundant array RAID5 roll rapid capacitance enlarging method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN101251789A true CN101251789A (en) | 2008-08-27 |
Family
ID=39955195
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CNA2008101028935A Pending CN101251789A (en) | 2008-03-28 | 2008-03-28 | Cheap magnetic disc redundant array RAID5 roll rapid capacitance enlarging method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101251789A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102200892A (en) * | 2011-04-29 | 2011-09-28 | 华中科技大学 | Capacity expansion method based on dynamic redundant array of independent disks (RAID) system |
WO2012164419A1 (en) * | 2011-06-01 | 2012-12-06 | International Business Machines Corporation | Facilitating routing by selectively aggregating contiguous data units |
US9225586B2 (en) | 2009-09-08 | 2015-12-29 | Huawei Technologies Co., Ltd. | Automatic expansion method, management device, management system |
CN108804338A (en) * | 2017-04-28 | 2018-11-13 | 爱思开海力士有限公司 | Data storage device and its operating method |
CN111208933A (en) * | 2018-11-21 | 2020-05-29 | 北京百度网讯科技有限公司 | Data access method, device, equipment and storage medium |
CN115599313A (en) * | 2022-12-05 | 2023-01-13 | 苏州浪潮智能科技有限公司(Cn) | Disk array capacity expansion method, system, storage medium and equipment |
-
2008
- 2008-03-28 CN CNA2008101028935A patent/CN101251789A/en active Pending
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9225586B2 (en) | 2009-09-08 | 2015-12-29 | Huawei Technologies Co., Ltd. | Automatic expansion method, management device, management system |
CN102200892A (en) * | 2011-04-29 | 2011-09-28 | 华中科技大学 | Capacity expansion method based on dynamic redundant array of independent disks (RAID) system |
CN102200892B (en) * | 2011-04-29 | 2013-09-18 | 华中科技大学 | Capacity expansion method based on dynamic redundant array of independent disks (RAID) system |
US9747233B2 (en) | 2011-06-01 | 2017-08-29 | International Business Machines Corporation | Facilitating routing by selectively aggregating contiguous data units |
US9117032B2 (en) | 2011-06-01 | 2015-08-25 | International Business Machines Corporation | Facilitating routing by selectively aggregating contiguous data units |
GB2503634A (en) * | 2011-06-01 | 2014-01-01 | Ibm | Facilitating routing by selectively aggregating contiguous data units |
WO2012164419A1 (en) * | 2011-06-01 | 2012-12-06 | International Business Machines Corporation | Facilitating routing by selectively aggregating contiguous data units |
CN108804338A (en) * | 2017-04-28 | 2018-11-13 | 爱思开海力士有限公司 | Data storage device and its operating method |
US11249917B2 (en) | 2017-04-28 | 2022-02-15 | SK Hynix Inc. | Data storage device and operating method thereof |
CN111208933A (en) * | 2018-11-21 | 2020-05-29 | 北京百度网讯科技有限公司 | Data access method, device, equipment and storage medium |
US11650754B2 (en) | 2018-11-21 | 2023-05-16 | Kunlunxin Technology (Beijing) Company Limited | Data accessing method, device, and storage medium |
CN111208933B (en) * | 2018-11-21 | 2023-06-30 | 昆仑芯(北京)科技有限公司 | Method, device, equipment and storage medium for data access |
CN115599313A (en) * | 2022-12-05 | 2023-01-13 | 苏州浪潮智能科技有限公司(Cn) | Disk array capacity expansion method, system, storage medium and equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Dunn et al. | A new I/O scheduler for solid state devices | |
CN103049397B (en) | A kind of solid state hard disc inner buffer management method based on phase transition storage and system | |
US8380928B1 (en) | Applying data access activity measurements | |
US9092141B2 (en) | Method and apparatus to manage data location | |
Gill et al. | WOW: Wise Ordering for Writes-Combining Spatial and Temporal Locality in Non-Volatile Caches. | |
US7383290B2 (en) | Transaction processing systems and methods utilizing non-disk persistent memory | |
US8719495B2 (en) | Concatenating a first raid with a second raid | |
US8364858B1 (en) | Normalizing capacity utilization within virtual storage pools | |
US9703717B2 (en) | Computer system and control method | |
US8312121B2 (en) | Storage system comprising function for alleviating performance bottleneck | |
CN101251789A (en) | Cheap magnetic disc redundant array RAID5 roll rapid capacitance enlarging method | |
US20110153972A1 (en) | Free space defragmention in extent based file system | |
CN103631537A (en) | Method and device for managing virtual disk | |
US20240036756A1 (en) | Systems, methods, and devices for partition management of storage resources | |
CN109086009A (en) | A kind of method for managing and monitoring and device, computer readable storage medium | |
US8972645B2 (en) | Request sent to storage device based on moving average | |
US7363453B1 (en) | Method and apparatus for improving storage device performance by selective volume swapping based on hot spot analysis | |
CN107301270A (en) | The Analytic modeling method of DDR storage system Memory accessing delays | |
CN101957729A (en) | Logical block transformation method and method and device compatible with reading and writing of user based on same | |
US10572464B2 (en) | Predictable allocation latency in fragmented log structured file systems | |
CN109144908A (en) | A kind of data-storage system and method based on cascade Expander | |
US11055001B2 (en) | Localized data block destaging | |
US10235053B1 (en) | Method and system for using host driver for flexible allocation fast-sideways data movements | |
CN103399783A (en) | Storage method and device of mirror image documents of virtual machines | |
US9542326B1 (en) | Managing tiering in cache-based systems |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C12 | Rejection of a patent application after its publication | ||
RJ01 | Rejection of invention patent application after publication |
Open date: 20080827 |