CN106330991A - Storage system with lightweight container nodes - Google Patents
Storage system with lightweight container nodes Download PDFInfo
- Publication number
- CN106330991A CN106330991A CN201510337050.3A CN201510337050A CN106330991A CN 106330991 A CN106330991 A CN 106330991A CN 201510337050 A CN201510337050 A CN 201510337050A CN 106330991 A CN106330991 A CN 106330991A
- Authority
- CN
- China
- Prior art keywords
- node
- service
- container
- host
- store
- 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]
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a storage system with lightweight container nodes. The storage system comprises: at least one cluster, wherein the cluster has a plurality of mutually connected nodes, each node comprises: a node host, wherein the node host is operated by a storage operating system and is provided with a plurality of service containers, and the plurality of service containers are used for providing specific client services and/or operating the nodes; at least one node storage device, wherein the node storage device is operated by the storage operating system and is used for providing access of service container data; and an information module, wherein the information module is connected to each node host and is used for bridging the node hosts and/or the node hosts in different clusters. A coarseness representational state transmission application programming interface is applied to the communication between any two service containers.
Description
Technical field
The present invention relates to a kind of stocking system, especially with regard to one, there is lightweight and hold
The stocking system of device framework.The capacity of this system possesses autgmentability, itself has intelligence
Characteristic and perceiving at the application program performed by stocking system.
Background technology
Daily life has the storage framework of many.For example, most widely used
It is to store LAN (Storage Area Network, SAN) to store up with network building-out
Deposit (Network Attached Storage, NAS).It is a kind of for storing LAN
Dedicated network, it is provided that depositing of block level (block level) Data Holding Equipment of integration
Taking, each server and storage facilities connection medium in this dedicated network are with light
Fibre is main.On the other hand, Network Attached Storage is a kind of file-level (file level)
Computer data storage server, be connected to computer network to provide heterogeneous client
Group data accesses.In Network Attached Storage, storage facilities is generally with network cable
It is connected.Thus, owing to connecting the narrower bandwidth of medium, the data volume that can transmit
Relatively storing LAN more limited, Network Attached Storage is often used as file server.
Compared to storing LAN, Network Attached Storage has the advantages such as low cost and easy operation.
But, store the better performances of LAN, thus be more suitable for that there is high capacity
Application, such as data base or mail server service.Need additionally, ought more access
When asking generation, Network Attached Storage can become unstable.Therefore, most enterprises are the most inclined
Good use stores LAN in business operation.
Additionally, for storing framework, a kind of technology become more and more popular is super fusion
Formula stores, and its collection stores, calculates, and network and virtualization are in a hardware cell.
Although these store framework maturation and are applied to specific area, but also have the sky improved
Between.For example, the storage LAN shown in Fig. 1 has calculating node (server)
1,2 and 3, and storage facilities 4,5 and 6, whole system can be used to
Client video.Storage facilities 4 is used as a customer data base, comprise user for
The personal information of registration, ID, and password.Storage facilities 5 sets number after being used as one
According to data base, set data after this and refer to the ground of a selected video physical holding of the stock
Side.Storage facilities 6 is used for accessing video.It is evident that in Video service different
Stage in, the plurality of calculating node 1,2 or 3 need for client send demand
Give different storage facilities.Because it is remote away with storage facilities, at them to calculate node
Between data transmit round cause waste of time, it means that system is inevitable
Cost allowance.
It is obvious that because software in need is all arranged in same unit, super
The stocking system of fusion type can solve above-mentioned problem.This super fusion type stocking system
Node will be calculated as close possible to storage facilities, and the storage facilities of redundancy will be provided, know
Road will need the information of how many resources (CPU, storage facilities, and network) to it future
For be extremely important.But, most super fusion type stocking systems cannot obtain
Such information.Additionally, either store LAN or super fusion type stocking system,
The common framework of storage facilities is unrelated with application surface, it means that storage facilities is very
Being optimized various application less, running also cannot be obtained by the data of storage with disposing
Coordinate.
If scrutinizing the running of each storage facilities, it appeared that they little frameworks prop up
Hold the whole life cycle of application surface.For Fig. 1, the data of three types may
Use under different frequencies in the different stages.Stage in early days, customer data base
(such as MySQL) is because user account to create and the most often to access.Afterwards,
Data database (such as MongoDB data base) is set because user browses and regards after video
Frequently more it is accessed.In the latter half, video data stores (such as Ceph storage
Device) the most often it is accessed because user watches video.In the different stages, distribution is not
Same stock number, the RAM such as used as cache or solid state hard disc, to having into
The system of this benefit is necessary.Distribute too much resource meaning waste, and distribute not
The resource of foot may cause time delay longer than require in service level agreement.
Accordingly, it would be desirable to the stocking system of a kind of innovation solves above-mentioned problem.This storage
Deposit system answers the request of intelligent corresponding each application, and can reach rapid configuration.It is also
High performance and cost benefit can be maintained.Most important, the extensibility of stocking system
Can be highly desirable.
Summary of the invention
In order to solve the problems referred to above, the invention provides one and there is Lightweight Container joint
The stocking system of point.This stocking system with Lightweight Container node includes at least one
Gathering together and an information module, wherein, this is gathered together and has multiple node being connected to each other,
Each node has a node host and at least one node store equipment, this node host
Stored operating system by one to be operated, and multiple service container is installed, the plurality of service
Container is used for providing client special services and/or node running;This node store equipment
Operated by this storage operating system, in order to provide the access of service container data;Should
Information module is connected to each node host, in order to bridge each node host and/or difference
Node host in gathering together.Coarseness representational state transfer (REpresentational
State Transfer, REST) application programming interface (Application
Programming Interface, API) leading between any two service container
News.
According to the present invention, this information module is separate server, is built in a host node master
To manage the software of all node host bridge joint or to be installed in each node host in machine
An application program.Communication can be carried out by a native protocol between two service containers,
Wherein this service container provides for operating the store-service of this node store equipment, use
Data, services in operational data storehouse or the application service for this special services.
This operating system can further include: a policy data base, defines in order to store
Each service container is deployed to a node host or by situation about removing in a node host
Rule;One distributed data processing engine, in order to coordinate each of this special services
Action and the running of the node by the service container proposition of the plurality of node host, and locate
Manage this special services and this running;One policy engine, in order to calculate in distributed data
Process the data in engine, and it is many to dispose this according to the rule in this policy data base
Individual service container;One common program, in order to provide for operating this node store equipment
Software logic and for dispose this service container service container image file;And one
Operating module, in order to configure customized input/output scheduling device and/or store-service cache
Mechanism, and operate this node store equipment.
It is also preferred that the left this store-service is a supervision service, the most every in order to monitor
The performance pointer of one service container.This store-service is a flow model service, in order to
Create a discharge model of at least one performance pointer in node, and produce this performance and refer to
The prediction of pin.This performance pointer is that the cpu load of this stocking system, input per second are defeated
Go out number of operations (Input/output Per Second, IOPS), circulation, delay
Time, write hit rate or apply the circulation of network of this stocking system soon.This storage
Service is an abnormity detecting service, in order to detect this performance obtained by being serviced by supervision
The unusual pattern of pointer.
One of this rule aforesaid is following rule: if a node host has a storage
Deposit service, cause the cpu load of this node host higher than a threshold value, the most more should
The service container of store-service is deployed in other node host;Have if one gathers together
A store-service cause this average cpu load of gathering together low compared with a threshold value, then move
Except the service container that this store-service is unnecessary;If a node host has storage clothes
Business, causes the storage loads of this working host or is higher than a threshold value time delay, then
The service container of these store-service more is deployed in other node host;If a joint
Point main frame has a store-service, causes offered load or the network flow of this working host
Flux is higher than a threshold value, and the service container of these store-service the most more is deployed to other
In node;If a node host has an estimated performance load higher than a threshold value,
The most unnecessary service container removes from this node;If it is different that a node host detects one
Often, the most unnecessary service container removes from this node;If a detecting is abnormal more than one
Threshold value, then remove all of service container in this node;The service of this store-service
The node store equipment of the data needing access closer to store-service container disposed by container
The best;If a node host has a data, services, cause the CPU of this node host
Load is higher than a threshold value, and the service container of these data, services the most more is deployed to other
In node host;If it is average that the data, services that gathers together has causes this to gather together
Cpu load is low compared with a threshold value, then remove the service container that this data, services is unnecessary;
If a node host has a data, services, cause the storage loads of this working host
Or time delay is higher than a threshold value, the service container deployment of these data, services the most more
In other node host;If a node host has a data, services, cause this
The offered load of working host or network throughput are higher than a threshold value, these numbers the most more
It is deployed in other node according to the service container of service;If a node host has one
Application service, causes the cpu load of this node host to be higher than a threshold value, the most more
The service container of this application service is deployed in other node host;If one gathers together tool
The application service having, causes this average cpu load of gathering together low compared with a threshold value,
Then remove the service container that this application service is unnecessary;Should if a node host has one
With service, the storage loads of this working host or time delay is caused to be higher than a threshold value,
The service container of these application services the most more is deployed in other node host;If one
Node host has an application service, causes offered load or the network of this working host
Circulation is higher than a threshold value, and the service container of these application services the most more is deployed to it
In its node.
Accompanying drawing explanation
Fig. 1 is a kind of existing storage LAN;
Fig. 2 is the embodiment of a kind of stocking system of the present invention;
Fig. 3 is record and the prediction of CPU use;
Fig. 4 is a kind of abnormal conditions that CPU uses in record;
Fig. 5 is the framework of a storage operating system;
Fig. 6 is the embodiment of the another kind of stocking system of the present invention;
Fig. 7 is the example of policy and rule;
Fig. 8 shows multiple detailed rule.
Description of reference numerals: 1-calculates node;2-calculates node;3-calculates node;4-
Storage facilities;5-storage facilities;6-storage facilities;10-primary nodal point;100-first
Node host;101-stores container;102-data capsule;103-application container;110-
First stores operating system;120-hard disk;130-solid state hard disc;20-secondary nodal point;
200-secondary nodal point main frame;201-stores container;202-data capsule;203-applies appearance
Device;210-second stores operating system;30-the 3rd node;300-the 3rd node host;
301a-first stores container;301b-second stores container;301c-the 3rd stores container;
301d-the 7th stores container;302-data capsule;303-application container;310-the 3rd
Store operating system;40-fourth node;400-fourth node main frame;401a-the 4th stores up
Deposit container;401b-the 5th stores container;401c-the 6th stores container;401d-the 8th
Store container;402-data capsule;403-application container;410-the 4th stores operation system
System.
Detailed description of the invention
The present invention is by more particularly described below by referring to following embodiment.
Disclose the embodiment of a kind of stocking system of the present invention as shown in Figure 2.This storage
System includes that an information module 500 has gathering together of multiple node, the plurality of joint with one
Point is primary nodal point 10, secondary nodal point 20, the 3rd node 30 and respectively
Fourth node 40.It should be noted that the present invention does not limit the quantity of node, aforementioned 4
Individual node is only used for the present invention is described.It is true that any amount of node is all suitable for.
In gathering together and they are connected to each other 4 nodes.The medium connected can be optical fiber
Or general networking ether cable.
Each node has a node host and at least one node store equipment.To not
For congenerous, the quantity of node store equipment can be different from pattern.Each node
Main frame is stored operating system by one and is operated, and installs multiple service container.Thus,
Node host can be provided special services to client by service container, and it is also provided that spy
Surely the running of node is served.Certainly, this special services can be come with the running of node
From identical node host.At least one node store equipment alternatively stores operating system
Being operated, its (or they) is provided that the access of service container data.It is installed on and appoints
The service container of what node host can provide a kind of store-service, is used for operating node storage
Deposit equipment;It also provides for a kind of data, services, for operational data storehouse.Or, should
Service container provides a kind of application service, and this application service is provided that aforesaid specific clothes
Business.
It should be noted that service container should be belonging to Lightweight Container.It is different from weight
Level container, the operating system under it is performed by program is managed, and Lightweight Container is fortune
The execution environment (runtime environment) of the row a kind of isolation in operating system
In.For example, aforementioned Lightweight Container can be that JAVA performs environment, and is somebody's turn to do
Heavyweight container can be a virtual machine (virtual machine).It can also be
One program containers, is created by cgroups, and cgroups is a Linux
Interior nuclear properties.For following description purpose, if a service container is for offer one
Kind of store-service and this store-service operates for node, this service container is defined as storage
Deposit container.In the same manner, if a service container is used for providing a kind of data, services,
This service container is defined as data capsule.If a service container is used for providing a kind of
Application service, this service container is defined as application container.
In the present embodiment, stocking system is used for providing a kind of video streaming service.The
One node 10 comprises 1 primary nodal point main frame 100 and 3 hard disks 120.This is first years old
Node host 100 is stored operating system 110 with 3 hard disks by one first and is driven.
Primary nodal point main frame 100 is provided with 101,2 data capsules 102 of 1 storage container
And 1 application container 103.The store-service that storing container 101 is provided is a kind of
Monitor service.This supervision service can monitor in primary nodal point 10, each service container
Performance pointer.It is described in this and performance pointer afterwards refers to stocking system first segment
The point cpu load of main frame 100, input-output operation number of times (Input/output per second
Per Second, IOPS), circulation, time delay, write hit rate or application soon
The circulation of the network of this stocking system.For example, store container 101 to be provided that
Service to primary nodal point main frame 100, to monitor and to record the use (%) of passing CPU.
As it is shown on figure 3, this record is by shown in real broken line.These multiple data capsules 102 are all
Being MySQL database, they are used for storing the user data of each user, such as
Address name, password, address, telephone number, e-mail address, sex etc..
Because there being too many user's logging request, live load is to a data capsule 102
Say that burden is the heaviest, thus configuration the second data capsule 102 is to maintain primary nodal point main frame
The steady running of 100.Application container 103 provides a kind of application logged in for user
Service, it processes from the interaction of client request and when logining successfully or being failed
Feedback information.First storage operating system 110 is storage based on a ZFS operation
System, it can be virtualization and backup offer rear end storage.
Secondary nodal point 20 comprises 1 secondary nodal point main frame 200 and 5 solid state hard discs
130.Secondary nodal point main frame 200 is stored behaviour with the plurality of solid state hard disc 130 by one second
Make system 210 to be driven.Secondary nodal point main frame 200 be mounted with 1 storage container 201,
2 data capsules 202 and 1 application container 203.Thered is provided by storing container 201
Store-service be also one monitor service.This supervision service function is as stored container
The supervision service of 101, here is omitted.The plurality of data capsule 102 is all
MongoDB data base, they are for storing the rear data that set of video, and this video
After set datagram containing the information of video storage position.When login user is by browser interface
After have selected a video, this request is sent to secondary nodal point 20.When this video location
When being found by one of data capsule 202, application container 203 will be re-directed to this client
Hold on the node of video storage.Similarly, because to set the request of data after accessing
Too many, and live load is the heaviest for a data capsule 202, thus configuration
Second data capsule 202.Application container 203 provides a kind of application service, and it is made
Function for relaying.Second store operating system 210 be also one based on Linux
Store operating system.Relatively primary nodal point 10 and secondary nodal point 20, it is known that joint
The quantity of some storage facilities is unnecessary with pattern identical.Meanwhile, the storage of each node
Operating system also can convert along with different tasks.
3rd node 30 has 1 the 3rd node host 300 and 9 hard disks 120,
Hard disk 120 can be a disk array.3rd node host 300 and hard disk 120
Stored operating system 310 by one the 3rd to be driven.3rd node host 300 is mounted with
1 first stores container 301a, 1 the second storage container 301b, 1 the 3rd storage
Deposit container 301c, 3 data capsules 302 and 1 application container 303.It will be apparent that
3rd node host 300 has than primary nodal point main frame 100 and secondary nodal point main frame 200
More operational function.The store-service that being stored container 301a by first provides is a kind of
Abnormity detecting services, and it can detect the unusual of the performance pointer that obtained by supervision service
Pattern.As shown in Figure 4.Detect the unusual pattern of cpu load by imaginary broken line institute
Show.Extremely generally mean that hardware, software fault or malice use.Store operation system
System will be taken action before possible situation occurs.Stored what container 301b provided by second
Store-service is discharge model service, and it can create at least one performance in the 3rd node 30
The discharge model of pointer, and and then produce the prediction of this performance pointer.As it is shown on figure 3,
Imaginary broken line illustrates the prediction case at following cpu load, this service of any offer
Suitably method or algorithm can be applied.The storage that container 301c is provided is stored by the 3rd
The service of depositing is that a kind of supervision services, and this supervision service function is as stored container 101
Monitoring service, here is omitted.
Data capsule 302 is all Ceph bin, and they are used for storing a part
Video archive and provide these archives to video streaming service use.Hold when carrying out self-application
The request of device 203 requires to access a video archive therein, one of data capsule 302
Start one section and connect one section of these archives of ground transmission.Because there have too many request to require to be different
Video, a data capsule 302 uses not, so 3 data capsules 302 of configuration.
Application container 303 provides by finite bandwidth transmission segmentation archives to each client
Application service.3rd storage operating system 310 is also a kind of storage based on Linux
Operating system.
Fourth node 40 has 400,2 solid state hard discs 130 of 1 fourth node main frame
With 7 hard disks 120, these 7 hard disks 120 can be a disk array.Solid-state
Hard disk 130 is used for storing some archives being often accessed, quickly to obtain archives.
Fourth node main frame 400, solid state hard disc 130 are stored behaviour with hard disk 120 by one the 4th
Making system 410 to be driven, this fourth node main frame 400 is mounted with 1 the 4th storage
Container 401a, 1 the 5th store container 401b, 1 the 6th store container 401c, 3
Individual data capsule 402 and 1 application container 403.Stored container 401a by the 4th to carry
The store-service of confession is abnormity detecting service, stores container 301a with first identical, different
No longer this repeats the function of often detecting service.The storage that container 401b provides is stored by the 5th
It is identical that the service of depositing stores container 301b supplier with second.Container 401c is stored by the 6th
There is provided store-service be snapshot service, this snapshot service carry out node store equipment (Gu
State hard disk 130 and hard disk 120) snapshot, in case certain logical block number (Logical
Unit Number, LUN) lost data recovery request.The plurality of data capsule 402
It it is all Ceph bin.Similarly, they are used for storing the video shelves of another part
Case and provide these archives to video streaming service use, their function mode with number
According to container 302.Application container 403 provides transmission segmentation archives to each client
Application service.4th storage operating system 410 is also a storage based on Linux
Operating system.
Although the storage operating system in the plurality of node can be identical or different, each
The framework of person should possess following assembly.As shown in Figure 5.The framework of this operating system
Comprise: a policy data base, a distributed data processing engine, a policy engine,
One common program and an operating module.Policy data base stores and defines each service appearance
Device is deployed to node host or removes the rule of situation from node host.According to the present invention,
If the live load of the service container in a node host can be by remaining same types
Service container performed by, any service container can be disposed according to demand, any
Service container can be removed.To different service containers, apply different rules.
These rules can be any one in following rule:
If a node host has a store-service, cause the CPU of this node host
Load is higher than a threshold value, and the service container of these store-service the most more is deployed to other
In node host;The store-service that if gathers together has, this is caused to gather together averagely
Cpu load low compared with a threshold value, then remove the service container that this store-service is unnecessary;
If a node host has a store-service, cause the storage loads of this working host
Or time delay is higher than a threshold value, the service container deployment of these store-service the most more
In other node host;If a node host has a store-service, cause this
The offered load of working host or network throughput are higher than a threshold value, these storages the most more
The service container depositing service is deployed in other node;If a node host has one
Estimated performance load is higher than a threshold value, and the most unnecessary service container removes from this node;
If a node host detects an exception, the most unnecessary service container moves from this node
Remove;If a detecting is abnormal more than a threshold value, remove all of service in this node
Container;The service container of this store-service is disposed to be needed to deposit closer to store-service container
The node store equipment of the data taken is the best;If a node host has data clothes
Business, causes the cpu load of this node host higher than a threshold value, these data the most more
The service container of service is deployed in other node host;If one gathers together have one
Data, services, causes this average cpu load of gathering together low compared with a threshold value, then removes
The service container that this data, services is unnecessary;If a node host has a data, services,
Cause the storage loads of this working host or be higher than a threshold value time delay, the most more
The service container of this data, services is deployed in other node host;If a node master
Facility have a data, services, cause offered load or the network throughput of this working host
Higher than a threshold value, the service container of these data, services the most more is deployed to other node
In;If a node host has an application service, cause the CPU of this node host
Load is higher than a threshold value, and the service container of these application services the most more is deployed to other
In node host;The application service that if gathers together has, this is caused to gather together averagely
Cpu load compared with a threshold value come low, then remove the service that this application service is unnecessary
Container;If a node host has an application service, cause the storage of this working host
Deposit load or time delay to hold higher than a threshold value, the service of these application services the most more
Device is deployed in other node host;If or a node host has an application service,
The offered load or the network throughput that cause this working host are higher than a threshold value, the most more
The service container of these application services many is deployed in other node.
It is specific that distributed data processing engine coordinates this by the service container of node host
Each action of service and the running of node, and process this special services and running.
For coordinating the action of special services, as shown in Figure 2.If video shelves
Case is too big and is stored in a data capsule 302 and a data capsule 402, distributed
Data processing engine, during video streaming, assists the linking of archives.For coordinating
For running action, further understand in order to distributed data processing engine is had,
Another embodiment being shown in Fig. 6 in this for illustrating.Use in Fig. 2 identical
Element, and first store container 301a with the 4th storage container 401a respectively by one
7th stores container 301d and the 8th storage container 401d is replaced.7th stores
Container 301d and the 8th stores container 401d and provides data de-duplication service.Work as storage
When deposit system starts data de-duplication, the service container of same types, the 7th stores
Container 301d and the 8th store container 401d can together with process, co-ordination by
Distributed data processing engine is responsible for.7th stores container 301d and the 8th stores container
The data de-duplication service of 401d can be from hithermost storage facilities (in same node point
In) obtain the data repeated, and it is sent to origin node storage facilities or other appointment
Storage facilities.
Policy engine, according to the rule in this policy data base, calculates at distributed number
According to the data processed in engine, with deployment services container.Policy is to service container
Deployment requirements, it generally comprises many rules.When applying a policy, stocking system
Service container will be turned on and off.For purposes of illustration, as shown in Figure 7.Store container
Following storage policy, data policy followed by data capsule, and application container follows application side
Pin.For each service container, there is the policy of many, each party's needle set has bag
Rule set containing multiple rules.These rules are probably different, such as, apply policy
Strictly all rules.It is identical that these rules are probably part, such as rule set A and rule
The then rule 1 in collection B and rule 3.These rules are probably and are integrated into a certain rule
Concentrate and be dispersed in again Else Rule to concentrate, such as store the rule in policy.In order to
Being better understood from, Fig. 8 illustrates an example.Rule 1 in rule set A is wanted
Cpu load is asked at least to remain 50% in following 60 minutes, the rule in rule set A
Then 2 require that IOPS at least remained 50%, in rule set A in following 60 minutes
Rule 3 requires that network at least remained 40% in following 60 minutes.Rule 1 is to rule
3 should meet simultaneously.If rule set A maintain, increase application container to node host,
Until all these rules all cannot meet.Similarly, the rule 4 in rule set B
Require that cpu load at least remained 15%, in rule set B in following 60 minutes
Rule 5 requires that IOPS at least remained 15%, in rule set B in following 60 minutes
Rule 6 require that network at least remained 25% in following 60 minutes.Rule 4 is to rule
Then 6 should meet simultaneously.If rule set B maintains, the application removing node host is held
Device, until all these rules all cannot meet.If rule set A and rule set B
It is unsatisfactory for, then existing application container is the most constant simultaneously.For each service container,
All of rule is enumerated as above, and explanation is not repeated.
Common program is provided that software logic, in order to operate node store equipment, it
Also provide for service container image file and carry out deployment services container.If a rule definition one
The demand of service container, by by needing the storage in the node host of service container to grasp
The image file of service container is initiateed, to dispose this service container as system.Running
Module can configure customized input/output scheduling device and/or store-service cache mechanism, it
Also can operative attachment node store equipment on node host.
Information module 500 is a platform independent server, and it is connected with each node host.
The function of information module 500 is the node host during bridge joint is gathered together.In the present embodiment,
Only use one to gather together.In other embodiment, two or more can be had to have
Multiple nodes is had to gather together.Thus, the function of information module 500 can be with bridging nodes
Main frame and/or difference gather together in node host.It is true that information module is not limited to clothes
Business device, it can be a set of software being built in host node main frame, all for managing
The bridge joint of node host;Or a set of application journey being installed in each node host
Sequence.
Different storage operating system may use different communications protocol, even if one
In individual node host, different agreement possible applications is in different services.In service
If there being too much communication during communication, the efficiency of stocking system can be reduced.The present invention makes
With a kind of improvement communication modes for stocking system.Coarseness REST
(REpresentational State Transfer) application programming interface
(Application Programming Interface, API) is used for any two service
Communication between container.This coarseness REST application programming interface is that HTTP please
Seek method, such as GET, PUT, POST, DELETE etc..From URL's
Resource is by an operable service container, and non-usage specific protocol operates.Thus,
These coarseness REST application programming interfaces are that this service eliminates without fixed duties leading to
News.But, if two storage operating systems in two nodes are identical, and
There is a set of native protocol serviced for these, after i.e. corresponding video, set data clothes
Business stores operation system with the second storage operating system 210 and the 3rd of video streaming service
System 310, two service containers can also via this native protocol communication with one another.
From the description above, clearly know that stocking system has extensibility, it
Can be to extend up and outward expansion.If the live load of a service container is very
Heavy, during another service container can be added to identical host node or gather together, other is main
Machine node, to reach the purpose extended up.If multiple main frames or gather together need to use
Come locally-supported (such as the main frame in different company position or data center),
More main frame is connected by information module and reaches outward expansion.Additionally, initial clothes
Business container position, perhaps without being fixed in a certain node host, stores the defeated of container
Enter/output profiles (I/O Profile) service can be deployed in each node.As
Really this input/output profile service detects secondary nodal point 20 and suffers from different recently
Often, the 3rd node 30 normal operation a period of time, input/output configuration file
Instruction the 3rd is stored operating system 310 by service, and the exception removing the 3rd node 30 is detectd
Survey and store container, and notify that the second storage operating system running 210 deployment one is detectd extremely
Survey and store container.It is emphasized, however, that be each store-service should initiate near it
Data (in same node point main frame) place.That is relevant application, data,
Should be in identical node host with storing container.For example, for video streaming
The data capsule 402 of application container 403 and Ceph bin at identical node
On main frame 40.
Although the present invention is disclosed above with embodiment, so it is not limited to this
Invention, has usually intellectual, without departing from the present invention in any art
Spirit and scope in, when making a little change and retouching, the therefore guarantor of the present invention
Scope of protecting ought be defined depending on this case right to be as the criterion.
Claims (13)
1. a stocking system with Lightweight Container node, it is characterised in that bag
Gather together and an information module containing at least one, wherein:
This is gathered together and has multiple node being connected to each other, and each node has a node
Main frame and at least one node store equipment, wherein:
This node host is stored operating system by one and is operated, and installs multiple service appearance
Device, the plurality of service container is used for providing client special services and/or node running;
This node store equipment is operated by this storage operating system, in order to provide service
The access of container data;
This information module is connected to each node host, in order to bridge each node host and/
Or difference gather together in node host,
Wherein coarseness representational state transfer application programming interface is used for any two
Communication between individual service container.
The storage system with Lightweight Container node the most according to claim 1
System, it is characterised in that this information module is separate server, is built in a host node master
To manage the software of all node host bridge joint or to be installed in each node host in machine
An application program.
The storage system with Lightweight Container node the most according to claim 1
System, it is characterised in that carry out communication by a native protocol between two service containers.
The storage system with Lightweight Container node the most according to claim 1
System, it is characterised in that this service container provides for operating this node store equipment
Store-service, data, services for operational data storehouse or for this special services should
With service.
The storage system with Lightweight Container node the most according to claim 4
System, it is characterised in that this operating system comprises:
One policy data base, defines each service container be deployed to a node in order to store
Main frame or the rule by situation about removing in a node host;
One distributed data processing engine, in order to coordinate each action of this special services
With the running of node proposed by the service container of the plurality of node host, and process should
Special services and this running;And
One policy engine, in order to calculate the data in distributed data processing engine,
And dispose the plurality of service container according to the rule in this policy data base.
The storage system with Lightweight Container node the most according to claim 5
System, it is characterised in that this operating system comprises further:
One common program, patrols for the software operating this node store equipment in order to provide
Collect and for disposing the service container image file of this service container;And
One operating module, in order to configure customized input/output scheduling device and/or to store clothes
Business cache mechanism, and operate this node store equipment.
The storage system with Lightweight Container node the most according to claim 4
System, it is characterised in that this store-service is a supervision service, in order to monitor at node
In the performance pointer of each service container.
The storage system with Lightweight Container node the most according to claim 4
System, it is characterised in that this store-service is a flow model service, in order to create
A discharge model of at least one performance pointer in node, and produce the pre-of this performance pointer
Survey.
9. according to the storage with Lightweight Container node described in claim 7 or 8
System, it is characterised in that this performance pointer is the cpu load of this stocking system, every
Second input-output operation number of times, circulation, time delay, write hit rate or application soon
The circulation of the network of this stocking system.
The storage system with Lightweight Container node the most according to claim 4
System, it is characterised in that this store-service is an abnormity detecting service, in order to detect by
The unusual pattern of this performance pointer obtained by supervision service.
The 11. storage systems with Lightweight Container node according to claim 5
System, it is characterised in that this rule is: if a node host has a store-service,
Cause the cpu load of this node host higher than a threshold value, these store-service the most more
Service container be deployed in other node host;The storage that if gathers together has
Service causes this average cpu load of gathering together low compared with a threshold value, then remove this storage
Service unnecessary service container;If a node host has a store-service, cause
The storage loads of this working host or time delay are higher than a threshold value, these storages the most more
The service container depositing service is deployed in other node host;If a node host tool
There is a store-service, cause the offered load of this working host or network throughput to be higher than
One threshold value, the service container of these store-service the most more is deployed in other node;
If a node host has an estimated performance load higher than a threshold value, the most unnecessary
Service container removes from this node;If a node host detects an exception, the most
Remaining service container removes from this node;If a detecting is abnormal more than a threshold value,
Then remove all of service container in this node;Or the service container portion of this store-service
Administration is the best closer to the node store equipment of the data of store-service container needs access.
The 12. storage systems with Lightweight Container node according to claim 5
System, it is characterised in that this rule is: if a node host has a data, services,
Cause the cpu load of this node host higher than a threshold value, these data, services the most more
Service container be deployed in other node host;The data that if gathers together has
Service causes this average cpu load of gathering together low compared with a threshold value, then remove these data
Service unnecessary service container;If a node host has a data, services, cause
The storage loads of this working host or time delay are higher than a threshold value, these numbers the most more
It is deployed in other node host according to the service container of service;If an or node host
There is a data, services, cause the offered load of this working host or network throughput high
In a threshold value, the service container of these data, services the most more is deployed in other node.
The 13. storage systems with Lightweight Container node according to claim 5
System, it is characterised in that this rule is: if a node host has an application service,
Cause the cpu load of this node host higher than a threshold value, these application services the most more
Service container be deployed in other node host;The application that if gathers together has
Service causes this average cpu load of gathering together low compared with a threshold value, then remove this application
Service unnecessary service container;If a node host has an application service, cause
The storage loads of this working host or time delay, the most more being somebody's turn to do should higher than a threshold value
It is deployed in other node host with the service container of service;If an or node host
There is an application service, cause the offered load of this working host or network throughput high
In a threshold value, the service container of these application services the most more is deployed in other node.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510337050.3A CN106330991B (en) | 2015-06-17 | 2015-06-17 | Stocking system with Lightweight Container node |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510337050.3A CN106330991B (en) | 2015-06-17 | 2015-06-17 | Stocking system with Lightweight Container node |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106330991A true CN106330991A (en) | 2017-01-11 |
CN106330991B CN106330991B (en) | 2019-06-21 |
Family
ID=57732674
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510337050.3A Active CN106330991B (en) | 2015-06-17 | 2015-06-17 | Stocking system with Lightweight Container node |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106330991B (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060085785A1 (en) * | 2004-10-15 | 2006-04-20 | Emc Corporation | Method and apparatus for configuring, monitoring and/or managing resource groups including a virtual machine |
US20070011485A1 (en) * | 2004-12-17 | 2007-01-11 | Cassatt Corporation | Application-based specialization for computing nodes within a distributed processing system |
CN1996843A (en) * | 2005-12-26 | 2007-07-11 | 北大方正集团有限公司 | Light distributed file storage system and file upload method |
CN103036699A (en) * | 2011-10-05 | 2013-04-10 | 广达电脑股份有限公司 | Server cluster and control mechanism thereof |
CN103036700A (en) * | 2011-09-29 | 2013-04-10 | 广达电脑股份有限公司 | Server cluster and control mechanism thereof |
-
2015
- 2015-06-17 CN CN201510337050.3A patent/CN106330991B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20060085785A1 (en) * | 2004-10-15 | 2006-04-20 | Emc Corporation | Method and apparatus for configuring, monitoring and/or managing resource groups including a virtual machine |
US20070011485A1 (en) * | 2004-12-17 | 2007-01-11 | Cassatt Corporation | Application-based specialization for computing nodes within a distributed processing system |
CN1996843A (en) * | 2005-12-26 | 2007-07-11 | 北大方正集团有限公司 | Light distributed file storage system and file upload method |
CN103036700A (en) * | 2011-09-29 | 2013-04-10 | 广达电脑股份有限公司 | Server cluster and control mechanism thereof |
CN103036699A (en) * | 2011-10-05 | 2013-04-10 | 广达电脑股份有限公司 | Server cluster and control mechanism thereof |
Also Published As
Publication number | Publication date |
---|---|
CN106330991B (en) | 2019-06-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP4264001B2 (en) | Quality of service execution in the storage network | |
JP4462969B2 (en) | Failover cluster system and failover method | |
US8296475B2 (en) | Systems and methods for performing multi-path storage operations | |
US9817584B2 (en) | Storage system having node with light weight container | |
CN101960427B (en) | Balanced consistent hashing for distributed resource management | |
US9785691B2 (en) | Method and apparatus for sequencing transactions globally in a distributed database cluster | |
CN100357916C (en) | Multi-protocol storage appliance that provides integrated support for file and block access protocols | |
US9449019B2 (en) | Peer-to-peer redundant file server system and methods | |
US6389432B1 (en) | Intelligent virtual volume access | |
CN103763383B (en) | Integrated cloud storage system and its storage method | |
US8261282B1 (en) | System and method for virtual machine host load protection | |
US8312454B2 (en) | System administration method and apparatus | |
US8856091B2 (en) | Method and apparatus for sequencing transactions globally in distributed database cluster | |
US9800459B1 (en) | Dynamic creation, deletion, and management of SCSI target virtual endpoints | |
CN101808139A (en) | Data storage system in cloud environment | |
JP2005505814A (en) | Load balancing in storage networks | |
US9747180B1 (en) | Controlling virtual endpoint failover during administrative SCSI target port disable/enable | |
WO2003027856A1 (en) | Pooling and provisionig storage resources in a storage network | |
JP6040612B2 (en) | Storage device, information processing device, information processing system, access control method, and access control program | |
US9817732B1 (en) | Method for controlling failover and failback of virtual endpoints in a SCSI network | |
EP1999655A2 (en) | Methods for dynamic partitioning of a redundant data fabric | |
CN111522499B (en) | Operation data reading device and reading method thereof | |
US20240248914A1 (en) | Methods for hierarchical propagation in tree structures and devices thereof | |
US20100057989A1 (en) | Method of moving data in logical volume, storage system, and administrative computer | |
CN106330991A (en) | Storage system with lightweight container nodes |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 1233079 Country of ref document: HK |
|
GR01 | Patent grant | ||
GR01 | Patent grant |