CN1266604C - Management system for large scale scale heterogeneous cluster - Google Patents
Management system for large scale scale heterogeneous cluster Download PDFInfo
- Publication number
- CN1266604C CN1266604C CN 02123496 CN02123496A CN1266604C CN 1266604 C CN1266604 C CN 1266604C CN 02123496 CN02123496 CN 02123496 CN 02123496 A CN02123496 A CN 02123496A CN 1266604 C CN1266604 C CN 1266604C
- Authority
- CN
- China
- Prior art keywords
- layer
- node
- command
- management system
- service layer
- 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.)
- Expired - Lifetime
Links
Images
Landscapes
- Stored Programmes (AREA)
- Computer And Data Communications (AREA)
Abstract
The present invention relates to a management system for a large-scale heterogeneous cluster. The present invention at least comprises a GUI layer, a US layer and an NS layer, wherein the GUI layer is used for providing a graphic management interface, collecting a request from a user, transmitting the request to the US layer and processing and displaying a processed command execution result of the US layer; the US layer is used for transmitting data and distributing commands between the GUI layer and the NS layer; the NS layer is used for receiving and responding a parallel command socket request from the US layer and returning a command processing result. The present invention can visit and manage a cluster system through a browser and a local area network, the scale of the cluster managed by the present invention can be arbitrarily extended, the cluster can be heterogeneous, and different operating systems can be installed according to different nodes; in this way, the discrepancy of the operating systems is shielded, the work efficiency of a cluster system administrator is greatly improved, the dependency among programs is relatively reduced, software of each layer has relative independence, and the maintainability of the software is enhanced.
Description
Technical field:
The present invention relates to a kind of management system and Data transmission and distribution command method of large scale scale heterogeneous cluster, be meant a kind of management system and Data transmission and distribution command method especially, belong to computer networking technology by World Wide Web (WEB) and LAN (Local Area Network) flexible management large scale scale heterogeneous cluster.
Background technology:
Network of Workstation be one group separate, by the interconnected computing machine of express network, and managed with the pattern of triangular web; It can make full use of the resource of each computing machine in the group of planes, realizes the parallel processing of complex calculation.
Along with science and technology and networks development, people are more and more higher to the requirement of the operational speed of a computer and processing power, and the node scale in the group of planes is increasing; And each node may be installed different operating system and software according to user's requirement in the group of planes, be placed on different places simultaneously, these group of planes of maintenance and management how reduce Network of Workstation keeper's burden, improve the efficiency of management, become the problem that becomes more and more important.
At present, along with the development of group of planes technology, the management software of some Network of Workstation has been arranged.Though these system management softwares can be realized basic system management, they are relatively stricter to the requirement of the operating system of a group of planes, normally customize for certain node and some type; When configurations such as the node type of a group of planes and operating system changed, whole management software just need remodify.Simultaneously, these management softwares mainly are operation and management in LAN (Local Area Network), can not manage a group of planes easily by the WEB browser.In addition, a lot of cluster management softwares do not have the ability of extensibility and managing large scale cluster nodes, only are suitable for small-scale cluster management.
Summary of the invention:
Fundamental purpose of the present invention is to provide a kind of management system of large scale scale heterogeneous cluster, this system can be by browser and LAN (Local Area Network) visit, Guan Li group of planes scale can be expanded arbitrarily simultaneously, and a group of planes can be an isomery, and different nodes can be installed different operating system.
Another object of the present invention is to provide a kind of management system of large scale scale heterogeneous cluster, for the Network of Workstation keeper provides a Network of Workstation management tool with graphical group of planes subscriber administration interface, make the Network of Workstation keeper manage heterogeneous cluster easily by browser or LAN (Local Area Network), the otherness of masking operation system.
The object of the present invention is achieved like this:
A kind of management system of large scale scale heterogeneous cluster, described system comprises: graphic user interface (Graphical User Interface, abbreviate GUI as) layer, user service (User Service abbreviates US as) layer and node service (Node Service abbreviates NS as) layer; Wherein,
The GUI layer is used to provide patterned administration interface, collects user's request and this request is passed to the US layer, shows after simultaneously the command execution results of US layer being handled;
The US layer is used for Data transmission and distribution command between GUI layer and NS layer;
The NS layer is used to receive and responds socket request from the parallel command of US layer, return command result.
Described NS layer comprises service processes (Service Deamon, abbreviate SD as) sublayer and service virtual level (Service Visual Layer, abbreviate the SVL sublayer as), the SD sublayer is used to receive socket (Socket) request from user service layer parallel command, this request of command response by calling the SVL sublayer simultaneously, last return command is service processes as a result;
The SVL sublayer is used for call operation system management order, and the result is analyzed and the SD sublayer is exported in processing.
Communicating by letter between described GUI layer and US layer and/or US layer and the NS layer all encrypted.
Described GUI layer is realized by Java language, and is operated on Local or Remote or the WEB.
All classes of GUI layer are carried out communication by a gui program of unified program exit and US layer by the remote method invocation (RMI) (Remote Method Invocation abbreviates RMI as) of Java language.
The GUI layer obtains the essential information of each node by database when initialization, set up the mapping table between operating system and the node name.
Described mapping table records the OS Type of each node at least.
The communication interface of US layer and GUI layer realizes that by Java language the communication interface of US layer and NS layer is realized by the C language.
The US layer is ordered and/or the parallel receive command execution results to parallel transmission of more than one node by Socket interface multi-thread concurrent mechanism.
Described SVL serves as reasons the sublayer organizes the command set that the perl scripts are formed more, and this perl order is left under the different catalogues, and this catalogue is named by operating system, and the perl script under the different directories calls the order of different operating system.Corresponding to same operating function, the perl script command under the different directories has identical title.
A kind of Data transmission of management system of the large scale scale heterogeneous cluster based on technique scheme and the method for distribution command, carry out following steps: the request of GUI layer is handled, be distributed to each node then, return to the GUI layer after the result treatment that each node is carried out at last.
The method that the US layer adopts the principal and subordinate to replace is by user service broker (abbreviating US_Agent as) and the communication of GUI layer.
The method that described principal and subordinate replaces is specially: move US_Agent simultaneously on the two or more node in a group of planes, one of them US_Agent is set at master server, other US_Agent is set at from server; The system of US layer surveys the running status of master server automatically; When master server operation just often, this master server is responsible for and the communication of GUI layer; When master server broke down, the US layer was made as master server selecting one of them to change from server; Repair the former master server that breaks down simultaneously, or select one to start operation US_Agent in the non-US_Agent node in a group of planes, as from server.
The present invention has following characteristics and beneficial effect:
1, adopts three-decker of the present invention, the interface of cluster management system is realized on WEB.Mobile office now more and more seems important, and the system manager often need manage a plurality of different group of planes, can bring waste on the plenty of time and travel to and fro between different machine rooms.Three-decker of the present invention can realize supporting the management of WEB mode, promptly can and manage a group of planes by the Internet visit, thereby improve Network of Workstation keeper's work efficiency greatly.
2, along with the development of user's demand and group of planes technology, heterogeneous cluster is more and more, and each user has different separately demands.All develop a cover management system for every kind of operating system user and then can be absorbed among a large amount of duplication of labour, cause huge waste; Therefore need shield the otherness of the operating system of each node to cluster management system, make that the nodal operation in the group of planes is transparent to the user.And cluster management person only need be by managing the group of planes with different node operating systems once the cover cluster management system easily.When the user selects new operating system, the structure of total system does not change, the SD layer of GUI layer, US layer and NS layer (background process of NS layer operation) does not need to change, only need increase a catalogue and corresponding script, just can realize management easily new operating system cluster nodes at the SVL layer.
3, the present invention is with good expansibility, and when the user had had new demand, GUI layer and SVL layer can optionally increase module and function, does not influence the realization of structure He other functions of total system simultaneously.Can make amendment according to user and actual the requirement in inside at all levels, do not influence other parts simultaneously.For example, when the node number is on a grand scale, the US layer can the employing process form of grouping improve concurrency and stability, when the node number after a little while, only need to adopt the form of sets of threads.The change of US layer inside can't change and influence the function of total system.This level makes the upgrading of system become more or less freely simultaneously, as long as guarantee that interface is constant, and at all levels can the upgrading according to actual conditions, the while does not need to revise the program of other layers, does not influence the realization of total system.
4, the invention enables the dependence of systemic software development to reduce.Hierarchy of the present invention makes the dependence between the system software reduce relatively, makes software have modular characteristics on stream.The exploitation of each layer does not need a lot of negotiations except interface, make the exploitation of each module have relative independence, has improved the efficient of exploitation, has improved the maintainability of software simultaneously yet.And do not influence the realization of structure He other functions of total system.
Description of drawings:
Fig. 1 is a three-decker synoptic diagram of the present invention.
Fig. 2 is a workflow diagram of the present invention.
Embodiment:
The present invention is described in further detail below in conjunction with accompanying drawing and specific embodiment:
Group of planes structure is a kind of loose coupling structure, each node in the group of planes all is an independently unit, the configuration of node and operating system can be according to user's demand flexible configuration, node can be personal computer or workstation, operating system can be any operating system, independently of one another between each node, simultaneously again can collaborative work under the management of group of planes software and configuration, a single system mapping is provided to external world.
The Network of Workstation management is exactly some the daily operations to cluster nodes, comprise: system maintenance work, network configuration, user management, system are checked etc., and to the network service configuration and the management of a whole group of planes, as network file system(NFS) (Network File System, abbreviation NFS), network information system (NetworkInformation System, abbreviation NIS), configuration and the management of domain name system (Domain Name System is called for short DNS), Samba (a kind of service software), LDAP (a kind of service software) etc.
Network of Workstation management for convenience, the graphic interface of cluster management system must have portability, and promptly can not need to revise and recompilate just can indifference strange land operation on various operating systems (for example Linux, Windows and various Unix operating system) platform.This patterned user interface also must be moved on browser simultaneously, supports the WEB system.In addition, this cluster management system is the node of management operating different operating system adaptively, therefore has the ability of masked nodes operating system otherness.
Referring to Fig. 1, in order to realize above-mentioned requirement, satisfy the target call of large scale scale heterogeneous cluster system management, management system of the present invention adopts three-decker: ground floor is the GUI layer, and the second layer is the US layer, and the 3rd layer is the NS layer; Wherein, the NS layer comprises SD sublayer and SVL sublayer two parts again.Simultaneously, for reliability and the security that guarantees network data transmission, the communication between above-mentioned each layer all adopts encryption technology to realize.
Referring to Fig. 1 and Fig. 2, the GUI layer is patterned user interface, and mainly being provides a patterned administration interface to the cluster management person, is responsible for collecting user's request, and corresponding request information passed to the US layer, and show after the command execution results of US layer handled.The GUI layer may operate on this locality, the long-range and WEB, may operate in simultaneously on the different operating system platforms.Therefore, consider that cross-platform and browser supports these factors, the GUI layer adopts Java language to realize.The gui program (usimport.class) of all classes of GUI layer by a unified program exit (GUI_exports.class) and US layer carries out communication by the RMI of java; Wherein, RMI is a kind of application programming interface (API) of Java language, it makes the Java language program of moving on the platform can be invoked at the program of moving on another platform, other programs of GUI layer do not need to know between US layer and the NS layer how to realize, the variation of US layer and NS layer and realization are transparent for the GUI layer.The inside of GUI layer realizes to US layer and NS layer it also being transparent simultaneously.
The GUI layer obtains the essential information of each node by the cluster management system database when initialization, set up the mapping table between operating system and the node name, when the GUI layer at every turn when the US layer sends order, the US layer is at first searched the OS Type that this table obtains node, thereby call the script command under the different directories of SVL layer of NS layer, realize management the heterogeneous cluster of forming by different node operating systems.
The US layer is a user service layer, and it serves as the media that information is transmitted between GUI layer and the NS layer, is responsible for Data transmission and distribution command between this is two-layer.Major function is the request with the GUI layer, through after the Treatment Analysis, is distributed to each node according to certain form, returns to the GUI layer after the result treatment that the SD layer of each node is carried out then.Therefore the US layer is the middleware communication layers.Consider the influence of speed and resources occupation rate, the US layer adopts C language and Java language to unite realization, adopt Java language to realize with the communication interface of GUI layer, adopt the C language to realize with the communication interface of NS layer, the Socket multi-thread concurrent mechanism by the C language is to concurrent transmission order of a plurality of nodes and concurrent reception command result.Simultaneously these interpretations of result are treated to the form that the GUI layer need show and send to the GUI layer.
The US layer asks to need to start the server program US_Agent of a Java for the user who responds the GUI layer.Stability for the US_Agent that guarantees the US layer, the method that US_Agent adopts high available principal and subordinate to replace realizes, two or more node operation US_Agent is promptly arranged in a group of planes simultaneously, and one is master server (master), other be from server (slave).Carry out communication by master server under the normal condition, when master server broke down, the system of US layer detected automatically, and found one to carry out communication from server, should be set to master server from server; Simultaneously, repair original master server or in a group of planes, look for another node to start operation US_Agent, as from server.
The switching of master and slave server and selection all are transparent to upper strata GUI layer and the NS of lower floor layer, can not influence the normal communication of Network of Workstation management software.For example: the GUI layer has an overtime setting (timeout) when the US layer sends connection request; After surpassing the time of timeout, think that then main US_Agent is dead, then from the configuration information table database, obtain address, and send connection request from US_Agent to this from US_Agent, carry out communication with it.Obtain the address of current master and slave US_Agent simultaneously from US_Agent, and revise configuration information table.Will LAN (Local Area Network), seek a lighter node of load from US_Agent and start the US_Agent process, will oneself rise to main US_Agent, return the address of master and slave US_Agent simultaneously.Realized the high availability of US layer in this way.The switching of this process will take a moment, in case but handover success, normal communication will be replied again by system.For the group of planes internal segment less situation of counting, can there be master and slave US_Agent, after main US_Agent crashed, GUI directly quoted mistake, and the prompting user repairs US_Agent.
The NS layer is a node service layer, and it is made up of SD layer and SVL layer.The SD layer is a background process that operates on the cluster nodes, and it receives the socket request from the parallel command of US layer, this request of command response by calling the SVL layer simultaneously, last return command result.The isomery management function of cluster management system realizes by the SVL layer, that is: SVL layer call operation system management order, and the result analyzed and handle, export to the SD layer then.The command set that the SVL layer is made up of many groups perl script, these perl orders are placed under the different catalogues, name these catalogues by the operating system name, and the perl script under the different directories calls the order of different operating system.For same function operations, the script command under these different directories has identical title.The GUI layer is distinguished these orders by different directory names, thus the order of calling different operating system, and correct operating system management order is called in final realization.
Referring to Fig. 2, in one embodiment of the invention, at length provided implementation concrete between above-mentioned each layer again.On the basis of setting up initialization nodal information database, for no matter being visit from LAN (Local Area Network), or the Internet visit, all need the authentication of carrying out through log-in interface; Then, read nodal information, that is: set up the mapping table of node/OS Type from above-mentioned database; The system manager sends request by gui management master interface, the mapping table of query node/operating system, judge the type of operating system of node operation after, call the perl script command under the SVL layer different directories, and order passed to the US layer; The US_Agent response request of US layer is called LPC, and this LPC will order and issue the NS layer of each node again; The SD layer responds the request of LPC, calls the perl script command of SVL layer then; The SVL layer calls the perl script command that the GUI layer transmits; The final call operation system command of perl script order.
Above embodiment is the unrestricted technical scheme involved in the present invention in order to explanation only, although the present invention is had been described in detail with reference to above preferred embodiment, those of ordinary skill in the art should be appreciated that technical scheme of the present invention can make amendment, is out of shape or is equal to replacement; And do not break away from the spirit and scope of technical solution of the present invention, all should be encompassed among the claim scope of the present invention.
Claims (13)
1, a kind of management system of large scale scale heterogeneous cluster is characterized in that: described system comprises: graphic user interface layer, user service layer and node service layer; Wherein,
The graphic user interface layer is used to provide patterned administration interface, collects user's request and this request is passed to user service layer, shows after simultaneously the command execution results of user service layer being handled;
User service layer is used for Data transmission and distribution command between graphic user interface layer and node service layer;
Node service layer is used to receive and responds socket request from the parallel command of user service layer, return command result;
Described node service layer comprises the service processes sublayer kimonos plan sublayer of debating on matters of principle, the service processes sublayer is used to receive the socket request from user service layer parallel command, by calling this request of command response of service virtual level sublayer, last return command is service processes as a result simultaneously;
Service virtual level sublayer is used for call operation system management order, and the result is analyzed and the service processes sublayer is exported in processing.
2, the management system of large scale scale heterogeneous cluster according to claim 1 is characterized in that: communicating by letter between described graphic user interface layer and user service layer and/or user service layer and the node service layer all encrypted.
3, the management system of large scale scale heterogeneous cluster according to claim 1 is characterized in that: described graphic user interface layer is realized by Java language, and is operated on Local or Remote or the WEB.
4, according to the management system of claim 1 or 3 described large scale scale heterogeneous clusters, it is characterized in that: all classes of graphic user interface layer are carried out communication by a unified program exit and a graphic user interface program of user service layer by the remote method invocation (RMI) of Java language.
5, according to the management system of claim 1 or 3 described large scale scale heterogeneous clusters, it is characterized in that: the graphic user interface layer obtains the essential information of each node by database when initialization, set up the mapping table between operating system and the node name.
6, the management system of large scale scale heterogeneous cluster according to claim 5 is characterized in that: described mapping table records the OS Type of each node at least.
7, the management system of large scale scale heterogeneous cluster according to claim 1, it is characterized in that: the communication interface of user service layer and graphic user interface layer realizes that by Java language the communication interface of user service layer and node service layer is realized by the C language.
8, the management system of large scale scale heterogeneous cluster according to claim 1 is characterized in that: user service layer orders and/or the parallel receive command execution results to parallel transmission of more than one node by sockets interface multi-thread concurrent mechanism.
9, the management system of large scale scale heterogeneous cluster according to claim 1, it is characterized in that: serve as reasons in described service virtual level sublayer organizes the command set that the perl scripts are formed more, this perl order is left under the different catalogues, this catalogue is named by operating system, and the perl script under the different directories calls the order of different operating system.
10, the management system of large scale scale heterogeneous cluster according to claim 9 is characterized in that: corresponding to same operating function, the perl script command under the different directories has identical title.
11, a kind of based on the Data transmission of the arbitrary described management system of claim 1-10 and the method for distribution command, it is characterized in that: the request of graphic user interface layer is handled, be distributed to each node then, return to the graphic user interface layer after the result treatment that each node is carried out at last.
12, method according to claim 11 is characterized in that: the method that user service layer adopts the principal and subordinate to replace, and by user service broker and the communication of graphic user interface layer.
13, method according to claim 12, it is characterized in that: the method that described principal and subordinate replaces is specially: while run user service broker on the two or more node in a group of planes, one of them user service broker is set at master server, and other user service broker is set at from server; The system of user service layer surveys the running status of master server automatically; When master server operation just often, this master server is responsible for and the communication of graphic user interface layer; When master server broke down, user service layer was made as master server selecting one of them to change from server; Repair the former master server that breaks down simultaneously, or select one to start the run user service broker in the non-user service broker node in a group of planes, as from server.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 02123496 CN1266604C (en) | 2002-07-02 | 2002-07-02 | Management system for large scale scale heterogeneous cluster |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 02123496 CN1266604C (en) | 2002-07-02 | 2002-07-02 | Management system for large scale scale heterogeneous cluster |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1466050A CN1466050A (en) | 2004-01-07 |
| CN1266604C true CN1266604C (en) | 2006-07-26 |
Family
ID=34142348
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 02123496 Expired - Lifetime CN1266604C (en) | 2002-07-02 | 2002-07-02 | Management system for large scale scale heterogeneous cluster |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN1266604C (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1315045C (en) * | 2004-03-16 | 2007-05-09 | 联想(北京)有限公司 | A method for implementing centralized concurrent management to cluster |
| CN103347088B (en) * | 2013-07-16 | 2016-03-23 | 诚迈科技（南京）有限公司 | A kind of server cluster trigger method |
| US10191887B2 (en) | 2013-07-18 | 2019-01-29 | Microsoft Technology Licensing, Llc | Context affinity in a remote scripting environment |
| CN104618486A (en) * | 2015-02-06 | 2015-05-13 | 浪潮电子信息产业股份有限公司 | Unified management method of multi-platform users of cluster storage system |
| CN106254103B (en) * | 2016-07-28 | 2019-08-16 | 北京国电通网络技术有限公司 | A kind of dynamically configurable method and device of RTMP group system |
| CN113608888A (en) * | 2021-07-09 | 2021-11-05 | 天津普林电路股份有限公司 | Remote operation system and operation method applying same |
-
2002
- 2002-07-02 CN CN 02123496 patent/CN1266604C/en not_active Expired - Lifetime
Also Published As
| Publication number | Publication date |
|---|---|
| CN1466050A (en) | 2004-01-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1108569C (en) | Scaleable and extensible system management architecture with dataless endpoints | |
| US5724512A (en) | Methods and apparatus for storage and retrieval of name space information in a distributed computing system | |
| US6976262B1 (en) | Web-based enterprise management with multiple repository capability | |
| US6466965B1 (en) | Centralized affinity maintenance in a workload managed client/server data processing system | |
| US5812130A (en) | Data management system and method for concurrent engineering | |
| US6658461B1 (en) | Method of, system for, and computer program product for providing a user interface for configuring connections between a local workstation file system and a remote host file system | |
| EP1107108A1 (en) | System and method for managing the configuration of hierarchically networked data processing devices | |
| CN103942281B (en) | The method and device that a kind of object to persistent storage is operated | |
| CN1858790A (en) | Method and system for managing automated resource provisioning with a workload shceduler | |
| Baker et al. | GridRM: an extensible resource monitoring system | |
| CN101741614B (en) | Equivalent type node manager and equivalent type node management method | |
| CN1494017A (en) | Holder selector used in global network service structure and its selection method | |
| CN101860564A (en) | Protocol-based service combination system and method | |
| Natrajan et al. | Grids: Harnessing geographically-separated resources in a multi-organisational context | |
| CN1647040A (en) | Method and system for managing a computer system | |
| CN1266604C (en) | Management system for large scale scale heterogeneous cluster | |
| CN1096638C (en) | Method for counting of custom/service unit and computer programme products | |
| CN1684430A (en) | Configurating data collecting system and its method in network management | |
| CN1256669C (en) | Heterogeneous machine group management system and information processing method | |
| CN1257463C (en) | System for automatic monitoring image management and communication | |
| Mittra | A road map for migrating legacy systems to client/server | |
| JP2002222227A (en) | Computer design aiding system and computer design aiding method | |
| Hura | Client-server computing architecture: an efficient paradigm for project management | |
| CN1556639A (en) | WWW service application conmon use integrated method | |
| CN1412691A (en) | Network integrating centralized calculation and distributed calculation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CX01 | Expiry of patent term |
Granted publication date: 20060726 |
|
| CX01 | Expiry of patent term |