CN104468701A - I/O service quality maintaining method for heterogeneous storage cluster system - Google Patents
I/O service quality maintaining method for heterogeneous storage cluster system Download PDFInfo
- Publication number
- CN104468701A CN104468701A CN201410551503.8A CN201410551503A CN104468701A CN 104468701 A CN104468701 A CN 104468701A CN 201410551503 A CN201410551503 A CN 201410551503A CN 104468701 A CN104468701 A CN 104468701A
- Authority
- CN
- China
- Prior art keywords
- storage resources
- storage
- grouping
- exclusive
- idle condition
- 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
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1097—Protocols in which an application is distributed across nodes in the network for distributed storage of data in networks, e.g. transport arrangements for network file system [NFS], storage area networks [SAN] or network attached storage [NAS]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/70—Admission control; Resource allocation
- H04L47/80—Actions related to the user profile or the type of traffic
- H04L47/805—QOS or priority aware
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/60—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources
- H04L67/61—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources taking into account QoS or priority requirements
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses an I/O service quality maintaining method for a heterogeneous storage cluster system. The method comprises the following steps that firstly, storage resources in a quick storage partition of the heterogeneous storage cluster system are divided into a plurality of storage resource packets, correlation does not exist in the I/O process among the storage resource packets, and the performance of the different storage resource packets cannot be affected during the respective I/O process; secondly, when the heterogeneous storage cluster system receives jobs, a current job is allocated to a storage resource packet selected from the storage resource packets according to the output performance requirement level of I/O of the current job and the states of the storage resource packets, and the states of the storage resource packets are updated. The I/O service quality maintaining method has the advantages that the I/O service performance is high, the I/O bandwidth requirement is low, the I/O service quality is good, and the method is especially suitable for I/O intensive job.
Description
Technical field
The present invention relates to isomery store group system in the I/O service quality maintenance technology field under the comparatively responsive scene of I/O performance, be specifically related to a kind of I/O service quality maintaining method for isomery storage cluster system.
Background technology
Storage system is the important component part in concurrent computational system.Along with concurrent computational system scale and the continuous expansion of applying deal with data scale, the performance of storage system becomes the key factor of impact and system for restricting practical application usefulness.For ensureing the execution efficiency of the I/O intensive applications such as mass data analysis, isomery storage architecture is usually adopted in group system, quick storage subregion and partition holding is at a slow speed divided into by storage system, quick storage subregion disposes the high-speed processing apparatus such as solid-state storage, and partition holding disposes ordinary magnetic disc or highly reliable memory device at a slow speed.For the more sensitive application of I/O Performance Ratio, adopt target that quick storage subregion exports as data to ensure that the demand of its I/O bandwidth is met, and for the insensitive application of I/O performance, then use the target that partition holding at a slow speed stores as data, thus avoid it to have an impact to the execution efficiency of I/O performance sensitive application.By said method, isomery storage architecture achieves isolates the stores service of dissimilar application, thus provides the quality of service guarantee of I/O process for I/O intensive applications.
Along with being on the increase of I/O intensive applications, also there is race problem in the storage resources of group system quick storage subregion.When a large amount of I/O intensive applications uses quick storage subregion, the increase of I/O request exacerbates the access conflict of storage resources on the one hand, operation I/O performance is caused to be difficult to be protected, on the other hand because the disk space of quick storage subregion is relatively limited, the quick generation of mass data causes the quick decline of free space, the availability of quick storage subregion also can be had a strong impact on, and these all make the I/O service quality of heterogeneous storage system to be guaranteed.In order to address these problems; heterogeneous storage system can construct the data synchronization mechanism of quick storage subregion and partition holding at a slow speed usually; namely data are after being saved in quick storage subregion; on backstage, data can be progressively synchronized to partition holding at a slow speed by machine in due course; after Complete Synchronization completes, remove the data of quick storage subregion, thus realize the effective release to quick storage partition space.In addition, for alleviating the hot issue of quick storage subregion local storage resource, load-balancing mechanism can be constructed between each I/O service node, be balanced the I/O operation number of each service node by scheduling mechanism, thus reach the object improving quick storage partitioned resources utilance.But, although said method improves the service condition of storage resources, but still fundamentally cannot solve the I/O service quality problem of I/O intensive applications, the I/O available bandwidth of key operation cannot be ensured, thus impact is brought on the quality of service guarantee of storage system.
As shown in Figure 1, isomery storage cluster system is made up of parts such as computing subsystem, ADMINISTRATION SUBSYSTEM, storage subsystem and monitoring and diagnosis subsystems usually, is interconnected between each several part by many cover communication networks.Quick storage subregion and partition holding two class partition holding is at a slow speed comprised in storage subsystem.Quick storage subregion adopts the memory device of high bandwidth, low delay usually, for I/O intensive applications provides higher I/O performance.The partition holding memory device that then provides Large Copacity highly reliable at a slow speed, the safe and reliable storage of the long term data for operation provides support.ADMINISTRATION SUBSYSTEM comprises management node and logs in node, during user's load operations, usually by logging in node submit job, and send operation resource requirement to management node, the operation of management node and resource management module calculate and storage resources according to user's request configuration is corresponding, thus are loaded into by user job on corresponding computing node and bring into operation.Monitoring and diagnosis subsystem is then responsible for monitoring the overall condition of system and safeguarding.At present, the requirement of real-time of computational tasks strengthens gradually, can operation complete in the given time and can bring serious impact to the validity of Job execution result, and data input/output procedure is the important step of computational tasks, if this process cannot guaranteed performance requirement, can bring the success or failure of computational tasks and have a strong impact on.Therefore, under the framework of existing isomery storage cluster system, how to ensure that the I/O performance of operation has become the key issue affecting existing high performance computing system availability.
Summary of the invention
The technical problem to be solved in the present invention is to provide that a kind of I/O service performance is high, I/O bandwidth demand is low, I/O service quality is good, is especially applicable to the I/O service quality maintaining method for isomery storage cluster system of the intensive operation of I/O.
In order to solve the problems of the technologies described above, the technical solution used in the present invention is:
For an I/O service quality maintaining method for isomery storage cluster system, implementation step is as follows:
1) storage resources of the quick storage subregion of isomery storage cluster system is divided into the grouping of multiple storage resources, and the I/O process between described storage resources grouping does not exist correlation, when making different storage resources divide into groups to carry out I/O respectively, its performance can not be affected mutually;
2) when isomery storage cluster system receives operation, the state of the performance requirement grade exported according to current work I/O and the grouping of each storage resources selects storage resources to distribute to current work from described multiple storage resources grouping, and upgrades the state of each storage resources grouping in described multiple storage resources grouping.
Preferably, the storage resources in described step 1) is grouped into a storage node or is responsible for one group of storage server of a set of disk array.
Preferably, described step 2) detailed step as follows:
2.1) performance requirement grade classification operation exported according to I/O is in advance exclusive demand operation and non-exclusive demand operation two class, described exclusive demand operation refers to that the output bandwidth of such operation to I/O has clear and definite QoS requirement, must exclusivity be ensured to distributed I/O resource, before the end of job, not allow other operations to use the storage resources shared by it to divide into groups, described non-exclusive demand operation refers to that the output bandwidth of such operation to I/O has performance requirement, but does not need to ensure exclusivity to distributed I/O resource, allows other operations to use the storage resources shared by it to divide into groups when system memory resources anxiety, the situation whether situation about according to whether occupied described grouping by operation and data have been synchronized to memory device is at a slow speed divided into the free time, exclusive, accurate idle, accurate exclusive four kinds of states, described idle condition represent this grouping current do not occupied by operation and there are not data in dividing into groups be not synchronized to memory device at a slow speed, described exclusive state represents that this grouping is current and to be occupied by exclusive operation and operation is not exited, described accurate idle condition represent this grouping current do not occupied by operation but there are data be not synchronized to memory device at a slow speed, described accurate exclusive state represents that this grouping is current and to be occupied by non-exclusive operation and operation is not exited,
2.2) when isomery storage cluster system receives operation, the type of performance requirement grade to current work that the current work I/O specified according to user exports judges, if current work is exclusive demand operation, redirect performs step 2.3), if current work is non-exclusive demand operation, redirect performs step 2.4);
2.3) the storage resources grouping being in idle condition is searched, if the storage resources number of packet being in idle condition can meet the demand of current work, the idle storage resources then returning respective amount divides into groups to current work, and will the state of storage resources grouping be selected to be set to exclusive state, assigning process terminates, and redirect performs step 2.5); If the storage resources number of packet of idle condition cannot meet the demand of current work, then report the operation I/O exception that demand cannot meet, assigning process terminates and exits;
2.4) the storage resources grouping being in idle condition, accurate idle condition and accurate exclusive state is searched, if the storage resources total number packets amount of above-mentioned three kinds of states can meet the demand of current work, then divide into groups according to the storage resources of priority allocation idle condition, the grouping of the storage resources of its sub-distribution accurate idle condition, the storage resources of way selection respective numbers that again divides the storage resources of registration exclusive state divide into groups divide into groups to return to current work, by being selected the state of storage resources grouping, the exclusive state that is as the criterion is set, assigning process terminates, and redirect performs step 2.5); If the storage resources total number packets amount of idle condition, accurate idle condition and accurate exclusive state cannot meet the demand of current work, then report the operation I/O exception that demand cannot meet, assigning process terminates and exits;
2.5) at the end of current work performs, if still there are data in the storage resources grouping occupied by current work not to be synchronized to memory device at a slow speed, then the state of being divided into groups by described storage resources arranges the idle condition that is as the criterion; If there are not data in described storage resources grouping not to be synchronized to memory device at a slow speed, then described storage resources Packet State is set to idle condition; Meanwhile, after all data syn-chronization storage resources divided into groups in the packet synchronization program of isomery storage cluster system to memory device at a slow speed, if storage resources grouping is not occupied by operation at present, then this storage resources is divided into groups to be set to idle condition.
The I/O service quality maintaining method that the present invention is used for isomery storage cluster system has following advantage: towards the isomery memory module in group system, when quick storage partitioned storage resource contention fierceness, the problem that the I/O QoS requirement of I/O intensive applications cannot be met, the storage resources of the quick storage subregion of isomery storage cluster system is divided into the grouping of multiple storage resources by the present invention, and there is not correlation in the I/O process between storage resources grouping, when making different storage resources divide into groups to carry out I/O respectively, its performance can not be affected mutually, when isomery storage cluster system receives operation, the state of the performance requirement grade exported according to current work I/O and the grouping of each storage resources selects storage resources to distribute to current work from multiple storage resources grouping, and upgrade the state of each storage resources grouping in the grouping of multiple storage resources, by setting allocation rule according to the I/O real-time demand of operation and the state of each storage resources grouping, thus the operation and storage resources grouping with different I/O output demand is effectively managed, the I/O bandwidth demand of the intensive operation of I/O can be ensured in mechanism, realize the I/O quality of service guarantee of job class, there is I/O service performance high, I/O bandwidth demand is low, I/O service quality is good, especially the advantage of the intensive operation of I/O is applicable to.
Accompanying drawing explanation
Fig. 1 is the frame structure schematic diagram of prior art isomery storage cluster system.
Fig. 2 is the basic implementing procedure schematic diagram of the embodiment of the present invention.
Fig. 3 is the workflow schematic diagram of the isomery storage cluster system of the application embodiment of the present invention.
Embodiment
As shown in Figure 2, the present embodiment is as follows for the implementation step of the I/O service quality maintaining method of isomery storage cluster system:
1) storage resources of the quick storage subregion of isomery storage cluster system is divided into the grouping of multiple storage resources, and the I/O process between storage resources grouping does not exist correlation, when making different storage resources divide into groups to carry out I/O respectively, its performance can not be affected mutually;
2) when isomery storage cluster system receives operation, the state of the performance requirement grade exported according to current work I/O and the grouping of each storage resources selects storage resources to distribute to current work from multiple storage resources grouping, and upgrades the state of each storage resources grouping in the grouping of multiple storage resources.
In the present embodiment, the storage resources in step 1) is grouped into a storage node or is responsible for one group of storage server of a set of disk array.After according to the principle of step 1) the storage resources of the quick storage subregion of isomery storage cluster system being divided into the grouping of multiple storage resources, in the present embodiment, structural plane is to the unified SRM module of job class I/O demand, by SRM module perform step 2) be responsible for above-mentioned each grouping use and distribution manages.Traditional SRM or dispatching method are usually based on different data files or storage object, None-identified is from multiple data file of same operation or storage object, therefore the repeatedly I/O request from same operation may be assigned on different storage resources, stable I/O performance cannot be obtained, thus the I/O access efficiency of key operation cannot be ensured; SRM module faces in the present embodiment is to job class I/O demand, the state of the performance requirement grade exported according to current work I/O and the grouping of each storage resources selects storage resources to distribute to current work from multiple storage resources grouping, therefore carry out resource distribution for different work, thus effectively can meet the overall I/O output bandwidth demand of the intensive operation of I/O.SRM module both can run on the management node of isomery storage cluster system together with the job handling routine of isomery storage cluster system, with job management system collaborative work; In addition also can the working alone of isomery storage cluster system, states its I/O demand, and sends it to SRM module when job request I/O resource.SRM module can be distributed according to the running status of the I/O demand of operation and current each grouping, thus ensure the performance of key operation I/O process, described for the I/O demand of operation, the running status of grouping and the distribution principle of operation respectively below.
In the present embodiment, step 2) detailed step as follows:
2.1) performance requirement grade classification operation exported according to I/O is in advance exclusive demand operation and non-exclusive demand operation two class, exclusive demand operation refers to that the output bandwidth of such operation to I/O has clear and definite QoS requirement, must exclusivity be ensured to distributed I/O resource, before the end of job, not allow other operations to use the storage resources shared by it to divide into groups, non-exclusive demand operation refers to that the output bandwidth of such operation to I/O has performance requirement, but does not need to ensure exclusivity to distributed I/O resource, allows other operations to use the storage resources shared by it to divide into groups when system memory resources anxiety, the situation whether situation about according to whether occupied grouping by operation and data have been synchronized to memory device is at a slow speed divided into the free time, exclusive, accurate idle, accurate exclusive four kinds of states, idle condition represent this grouping current do not occupied by operation and there are not data in dividing into groups be not synchronized to memory device at a slow speed, exclusive state represents that this grouping is current and to be occupied by exclusive operation and operation is not exited, accurate idle condition represent this grouping current do not occupied by operation but there are data be not synchronized to memory device at a slow speed, accurate exclusive state represents that this grouping is current and to be occupied by non-exclusive operation and operation is not exited,
2.2) when isomery storage cluster system receives operation, the type of performance requirement grade to current work that the current work I/O specified according to user exports judges, if current work is exclusive demand operation, redirect performs step 2.3), if current work is non-exclusive demand operation, redirect performs step 2.4);
2.3) the storage resources grouping being in idle condition is searched, if the storage resources number of packet being in idle condition can meet the demand of current work, the idle storage resources then returning respective amount divides into groups to current work, and will the state of storage resources grouping be selected to be set to exclusive state, assigning process terminates, and redirect performs step 2.5); If the storage resources number of packet of idle condition cannot meet the demand of current work, then report the operation I/O exception that demand cannot meet, assigning process terminates and exits;
2.4) the storage resources grouping being in idle condition, accurate idle condition and accurate exclusive state is searched, if the storage resources total number packets amount of above-mentioned three kinds of states can meet the demand of current work, then divide into groups according to the storage resources of priority allocation idle condition, the grouping of the storage resources of its sub-distribution accurate idle condition, the storage resources of way selection respective numbers that again divides the storage resources of registration exclusive state divide into groups divide into groups to return to current work, by being selected the state of storage resources grouping, the exclusive state that is as the criterion is set, assigning process terminates, and redirect performs step 2.5); If the storage resources total number packets amount of idle condition, accurate idle condition and accurate exclusive state cannot meet the demand of current work, then report the operation I/O exception that demand cannot meet, assigning process terminates and exits;
2.5) at the end of current work performs, if still there are data in the storage resources grouping occupied by current work not to be synchronized to memory device at a slow speed, then the state that storage resources divides into groups is arranged the idle condition that is as the criterion; If there are not data in storage resources grouping not to be synchronized to memory device at a slow speed, then storage resources Packet State is set to idle condition; Meanwhile, after all data syn-chronization storage resources divided into groups in the packet synchronization program of isomery storage cluster system to memory device at a slow speed, if storage resources grouping is not occupied by operation at present, then this storage resources is divided into groups to be set to idle condition.
As shown in Figure 3, the SRM module of isomery storage cluster system runs on management node, the SRM module that can provide with the present embodiment with the use of, also can be used alone.When user's load operations, can be sent by SRM module and calculate and storage resource demands, by the job and resource management system of isomery storage cluster system, corresponding storage resource demands is transmitted to SRM module, also can submits storage resources user demand to directly to SRM module.After SRM module receives job requirements, can allocate according to the service condition of current storage resources, and resource allocation result be fed back to user, the storage resources assigning process of the level that fulfils assignment.The Job execution overall procedure of the application embodiment of the present invention is as follows: 1. the walk, and the storage resources request of operation is sent to SRM module by job and resource management system; 2. the walk, and the state of the performance requirement grade that SRM module exports according to current work I/O and the grouping of each storage resources selects storage resources to divide into groups from multiple storage resources grouping; 3. walk, and resource allocation result (storage resources grouping) is returned to operation by SRM module; 4. the walk, if the storage resource demands of operation is met, then start to load corresponding operation, otherwise report resource cannot meet, and operation is blocked or exited.
The above is only the preferred embodiment of the present invention, protection scope of the present invention be not only confined to above-described embodiment, and all technical schemes belonged under thinking of the present invention all belong to protection scope of the present invention.It should be pointed out that for those skilled in the art, some improvements and modifications without departing from the principles of the present invention, these improvements and modifications also should be considered as protection scope of the present invention.
Claims (3)
1., for an I/O service quality maintaining method for isomery storage cluster system, it is characterized in that implementation step is as follows:
1) storage resources of the quick storage subregion of isomery storage cluster system is divided into the grouping of multiple storage resources, and the I/O process between described storage resources grouping does not exist correlation, when making different storage resources divide into groups to carry out I/O respectively, its performance can not be affected mutually;
2) when isomery storage cluster system receives operation, the state of the performance requirement grade exported according to current work I/O and the grouping of each storage resources selects storage resources to distribute to current work from described multiple storage resources grouping, and upgrades the state of each storage resources grouping in described multiple storage resources grouping.
2. the I/O service quality maintaining method for isomery storage cluster system according to claim 1, is characterized in that: the storage resources in described step 1) is grouped into a storage node or is responsible for one group of storage server of a set of disk array.
3. the I/O service quality maintaining method for isomery storage cluster system according to claim 1 and 2, is characterized in that, described step 2) detailed step as follows:
2.1) performance requirement grade classification operation exported according to I/O is in advance exclusive demand operation and non-exclusive demand operation two class, described exclusive demand operation refers to that the output bandwidth of such operation to I/O has clear and definite QoS requirement, must exclusivity be ensured to distributed I/O resource, before the end of job, not allow other operations to use the storage resources shared by it to divide into groups, described non-exclusive demand operation refers to that the output bandwidth of such operation to I/O has performance requirement, but does not need to ensure exclusivity to distributed I/O resource, allows other operations to use the storage resources shared by it to divide into groups when system memory resources anxiety, the situation whether situation about according to whether occupied described grouping by operation and data have been synchronized to memory device is at a slow speed divided into the free time, exclusive, accurate idle, accurate exclusive four kinds of states, described idle condition represent this grouping current do not occupied by operation and there are not data in dividing into groups be not synchronized to memory device at a slow speed, described exclusive state represents that this grouping is current and to be occupied by exclusive operation and operation is not exited, described accurate idle condition represent this grouping current do not occupied by operation but there are data be not synchronized to memory device at a slow speed, described accurate exclusive state represents that this grouping is current and to be occupied by non-exclusive operation and operation is not exited,
2.2) when isomery storage cluster system receives operation, the type of performance requirement grade to current work that the current work I/O specified according to user exports judges, if current work is exclusive demand operation, redirect performs step 2.3), if current work is non-exclusive demand operation, redirect performs step 2.4);
2.3) the storage resources grouping being in idle condition is searched, if the storage resources number of packet being in idle condition can meet the demand of current work, the idle storage resources then returning respective amount divides into groups to current work, and will the state of storage resources grouping be selected to be set to exclusive state, assigning process terminates, and redirect performs step 2.5); If the storage resources number of packet of idle condition cannot meet the demand of current work, then report the operation I/O exception that demand cannot meet, assigning process terminates and exits;
2.4) the storage resources grouping being in idle condition, accurate idle condition and accurate exclusive state is searched, if the storage resources total number packets amount of above-mentioned three kinds of states can meet the demand of current work, then divide into groups according to the storage resources of priority allocation idle condition, the grouping of the storage resources of its sub-distribution accurate idle condition, the storage resources of way selection respective numbers that again divides the storage resources of registration exclusive state divide into groups divide into groups to return to current work, by being selected the state of storage resources grouping, the exclusive state that is as the criterion is set, assigning process terminates, and redirect performs step 2.5); If the storage resources total number packets amount of idle condition, accurate idle condition and accurate exclusive state cannot meet the demand of current work, then report the operation I/O exception that demand cannot meet, assigning process terminates and exits;
2.5) at the end of current work performs, if still there are data in the storage resources grouping occupied by current work not to be synchronized to memory device at a slow speed, then the state of being divided into groups by described storage resources arranges the idle condition that is as the criterion; If there are not data in described storage resources grouping not to be synchronized to memory device at a slow speed, then described storage resources Packet State is set to idle condition; Meanwhile, after all data syn-chronization storage resources divided into groups in the packet synchronization program of isomery storage cluster system to memory device at a slow speed, if storage resources grouping is not occupied by operation at present, then this storage resources is divided into groups to be set to idle condition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410551503.8A CN104468701B (en) | 2014-10-17 | 2014-10-17 | A kind of I/O service quality maintaining methods for isomery storage cluster system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410551503.8A CN104468701B (en) | 2014-10-17 | 2014-10-17 | A kind of I/O service quality maintaining methods for isomery storage cluster system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104468701A true CN104468701A (en) | 2015-03-25 |
| CN104468701B CN104468701B (en) | 2017-10-13 |
Family
ID=52914054
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410551503.8A Active CN104468701B (en) | 2014-10-17 | 2014-10-17 | A kind of I/O service quality maintaining methods for isomery storage cluster system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104468701B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107800574A (en) * | 2017-11-03 | 2018-03-13 | 郑州云海信息技术有限公司 | Store QOS adjusting methods, system, equipment and computer-readable memory |
| CN113467700A (en) * | 2020-03-31 | 2021-10-01 | 阿里巴巴集团控股有限公司 | Data distribution method and device based on heterogeneous storage |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060095686A1 (en) * | 2004-10-29 | 2006-05-04 | Miller Wayne E | Management of I/O operations in data storage systems |
| CN102197395A (en) * | 2008-09-19 | 2011-09-21 | 甲骨文国际公司 | Storage-side storage request management |
| CN102541460A (en) * | 2010-12-20 | 2012-07-04 | 中国移动通信集团公司 | Multiple disc management method and equipment |
-
2014
- 2014-10-17 CN CN201410551503.8A patent/CN104468701B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060095686A1 (en) * | 2004-10-29 | 2006-05-04 | Miller Wayne E | Management of I/O operations in data storage systems |
| CN102197395A (en) * | 2008-09-19 | 2011-09-21 | 甲骨文国际公司 | Storage-side storage request management |
| CN102541460A (en) * | 2010-12-20 | 2012-07-04 | 中国移动通信集团公司 | Multiple disc management method and equipment |
Non-Patent Citations (3)
| Title |
|---|
| WEI ZHANG ET AL.: "An I/O Fault-Tolerant Method Based on Application-Level Agent in Object-Based Storage System", 《ADVANCED MATERIALS RESEARCH》 * |
| XU WEIXIA ET AL.: "Hybrid hierarchy storage system in MilkyWay-2 supercomputer", 《FRONTIERS OF COMPUTER SCIENCE》 * |
| 郭御风 等: "一种新型高效共享的并行I/O系统", 《计算机工程》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107800574A (en) * | 2017-11-03 | 2018-03-13 | 郑州云海信息技术有限公司 | Store QOS adjusting methods, system, equipment and computer-readable memory |
| CN113467700A (en) * | 2020-03-31 | 2021-10-01 | 阿里巴巴集团控股有限公司 | Data distribution method and device based on heterogeneous storage |
| CN113467700B (en) * | 2020-03-31 | 2024-04-23 | 阿里巴巴集团控股有限公司 | Heterogeneous storage-based data distribution method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104468701B (en) | 2017-10-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107066319B (en) | Multi-dimensional scheduling system for heterogeneous resources | |
| CN107025205B (en) | Method and equipment for training model in distributed system | |
| Krishnamurthy et al. | Pratyaastha: an efficient elastic distributed sdn control plane | |
| KR101781063B1 (en) | Two-level resource management method and appratus for dynamic resource management | |
| CN102724277B (en) | The method of live migration of virtual machine and deployment, server and group system | |
| US9104492B2 (en) | Cloud-based middlebox management system | |
| EP3281359B1 (en) | Application driven and adaptive unified resource management for data centers with multi-resource schedulable unit (mrsu) | |
| Hermenier et al. | Bin repacking scheduling in virtualized datacenters | |
| Deshpande et al. | Gang migration of virtual machines using cluster-wide deduplication | |
| EP3001644A1 (en) | Distributed storage system, cluster node and interval management method thereof | |
| US20180018197A1 (en) | Virtual Machine Resource Allocation Method and Apparatus | |
| US10394606B2 (en) | Dynamic weight accumulation for fair allocation of resources in a scheduler hierarchy | |
| CN108462656B (en) | Container-based resource scheduling method and device for service convergence deployment | |
| US9092272B2 (en) | Preparing parallel tasks to use a synchronization register | |
| KR20160087706A (en) | Apparatus and method for resource allocation of a distributed data processing system considering virtualization platform | |
| CN111459641B (en) | Method and device for task scheduling and task processing across machine room | |
| Rupprecht et al. | SquirrelJoin: Network-aware distributed join processing with lazy partitioning | |
| KR101585160B1 (en) | Distributed Computing System providing stand-alone environment and controll method therefor | |
| JP2017037492A (en) | Distributed processing program, distributed processing method and distributed processor | |
| US11438271B2 (en) | Method, electronic device and computer program product of load balancing | |
| US9124587B2 (en) | Information processing system and control method thereof | |
| US20200272526A1 (en) | Methods and systems for automated scaling of computing clusters | |
| US20150365474A1 (en) | Computer-readable recording medium, task assignment method, and task assignment apparatus | |
| CN104468701A (en) | I/O service quality maintaining method for heterogeneous storage cluster system | |
| US11182189B2 (en) | Resource optimization for virtualization environments |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |