CN1173281C - A method and apparatus for implementing a workgroup server array - Google Patents

Abstract

A method and apparatus for implementing a workgroup server array ideal for web-based Intranet, Extranet and Internet applications. The inventive server array comprises a plurality of team/workgroup computers (408) equipped with work-group-based direct-access servers and modular controlling devices (1), creating workgroup-based fault-tolerant and fail-over capabilities, providing console-based monitoring and management support, and accommodating highly available and scalable web-based applications with optimal performance. These workgroup server arrays can be used as the basic building blocks to construct large-scale server clusters, so that more users can be served concurrently. Furthermore, workgroup-server-array-based architecture is created for building various highly available, scalable and mission critical server clusters, which enable distributed computing services for enerris-based Intranet, Extranet and Internet mission critical applications.

Description

Be used to realize the method and apparatus of workgroup server array

The application requires in the U.S. Provisional Application No.60/135 of submission on May 20th, 1999,318 right of priority, this application is that the United States serial of submitting on November 1st, 1996 is No.08/742, the further part of the patented claim of " using a kind of multiserver IVOD television system of the Workgroup Servers of directly visiting " by name of 741 by request, and No.08/742, the patent No. of 741 U. S. applications is No.6,049, on April 11st, 823,2000 authorized.

Technical field

A kind of method and apparatus that is used to realize the workgroup server array that the in-house network based on web, extranets and the Internet are used.

Background technology

The present invention relates generally to a kind of cluster of servers, be specifically related to a kind of method and apparatus that is used to realize workgroup server array, and be used to construct various cluster of servers to adapt to upgradeable in-house network, the framework that the vital task of extranets and the Internet is used based on web.

Server array of the present invention comprises and being equipped with based on the direct access type server of working group and group/working group's computing machine of opertaing device, this point is No.5 at applicant's U.S. Patent number, in 802,391 by name " do not use network operating system by the shared direct access type group/Workgroup Servers of group/working group computing machine " description is not arranged.And server array of the present invention has also been created a kind of framework based on workgroup server array, and it is available to utilize it to construct various height, and is upgradeable and handle the cluster of servers of vital task.

Have innovative significance technique of internet explosive development pole earth effect the exploitation of using and the mode of use.Once be that ten hundreds of internet website of static " handbook type article " is just developing rapidly and becomes the height switched Internet with processing power and use.Inside in large enterprises, the developer is just using the Web technology that enterprise is used and is being merged into large-scale in-house network.Between enterprise, business partners are just being set up safe extranets and are being communicated by letter with improving with the supply system of simplifying them.

Since based on the application of web just on the Internet, swollen rising rapidly on intranet and extranets, their performed function also becomes more and more important.And because enterprise can continue the technology based on web is applied to vital task, they need more ripe method to make its application highly available and scalable.

In order to reach the requirement of height upgradability and high degree of availability, present trend forward relate to system that a large amount of servers work together just the direction of cluster of servers develop, with the application that provides the terminal user to be asked.And, extensive based on a kind of framework of cluster of servers of setting up of the service request of web so that can adapt to availability, upgradability, reliability, serviceability, the requirement of management and secure context.

But, existing by using a plurality of independently servers to set up the height upgradability, high degree of availability and the technology of finishing the cluster of servers of using based on web of vital task tend to produce 3 layer architectures based on individual server, just the SS-3 framework of back indication.This SS-3 framework usually need be as the ground floor assembly of load equalizer, as the second layer assembly of application server, as the 3rd layer assembly of database and file server.Each independent server can be based on PC, based on the superminiature machine or based on the server of small-size computer, comprise a plurality ofly having parallel processing capability and use WinNT, Solaris, the CPU of Linux and Unix operating system.

On the basis of SS-3 framework, can set up based on the height practicality of web application and the cluster of servers of upgradability.But this framework also can produce following shortcoming.

1. about every layer assembly:

A) all incoming traffic of load equalizer-analysis and each independent queries/requests based on web is redirected to one of available second layer application server that is connected.Load equalizer according to the character of request and be performed the availability of web application server of load balancing and ability to the application server request for allocation based on web of the specific second layer.The load equalizer that three kinds of fundamental types are arranged: switch, software-type balanced device and applied balanced device.But if any pressure in the described load equalizer is excessive, the Internet connection all may be stopped up.

B) application server-from the task that the reception of ground floor load equalizer distributes, execution connects the 3rd layer data storehouse and file server to carry out application oriented data retrieval based on the application of web.But each application server may have nothing in common with each other owing to different hardware and software configurations, therefore can produce the managerial complicacy of load equalizer.In addition, each application server should handle based on fidelity also to handle inquiry based on non-fidelity, therefore will produce noncoherent package with various different stage safety yardsticks.And, each application server does not have the remote boot ability, unless comprise the auxiliary processor of a visit Network Based, if when making the primary processor of server break down, can trigger auxiliary processor by other based on network management server visit to guide primary processor again.

C) database/file server-be based on the server of client server is used to handle the database/file polling that is considered to client's second layer application server from all.Because based on not having difference between the fidelity and the traffic based on non-fidelity, therefore be used for all classifying, if therefore too many inquiry takes place to produce potential database/document retrieval bottleneck simultaneously at a hub file server and a database server based on fidelity with based on the application oriented data of non-fidelity.And, if these files and database server are realized as the part of data center, the complicacy of managing such data center will be very high, and this above data center comprises a plurality of databases that are assigned with and the file server that is linked on the memory device that a plurality of SAN (storage area networks) enable.Make server customer-centric and be that the server at center all needs complicated database software program with the server just because of such fact.But, in a data center, be application oriented data and business sensitive data centralization unfavorable, because must release special safety practice such as firewall filtering in case the potential destruction of causing based on the behavior of browsing of web.

D) all need such interlayer communication switch between interlayer communication switch-ground floor load equalizer and the second layer application server and between this application server and the 3rd layer of file and the database server.Because each assembly all is based on network, so all communications between these servers all handle by these two switches, so will cause the bottleneck and the managerial system overhead of unnecessary interlayer communication amount.

E) more layer promptly means more assembly, this can cause more Single Point of Faliure-on the basis of SS-3 framework, all load equalizers, application server, file and database server, router and switch all should have an anti-fault (fail-over) scheme, will guarantee like this that vital task is used not break down.Even develop comprehensively anti-fault scheme, other can be not effectively and cost efficient, and this is because of relating to too many hardware configuration and software program.

2. cluster of servers management:

A) because the complicacy of each assembly relevant, to the very complexity that also can become based on the monitoring of the cluster of servers of individual server and management with interlayer communication.The upgrading of single software tends to produce the incompatible of software, and this is owing to such fact, because there is too many involved software program all need be upgraded by each dealer.

B) whole serviceability and be not easy to optimize.In case cluster of servers is set up on the basis of SS-3 framework, it just must satisfy such standard: can handle steady state operation reposefully at least and adapt to peak value constantly operation and glitch does not take place.But, therefore will when the control steady state operation, increase uncertain factor and the necessary measurement of restriction when handling the operation in the peak value moment owing to the small-scale optimum point that does not have distribution that can be measured.

C) if causing too many database owing to high speed web visit is very popular is that the request at center will be served simultaneously, high degree of availability and cost-efficient linear upgradability will be difficult to keep so.At present, being based on the narrow band transmission speed of 56kbps and relevant service based on the inquiry of web all concentrates on webpage and sends.But, if popular message transmission rate concentrates on individual database and sends as the webpage at center owing to using cable modem or ADSL to rise to 1Mbps or higher and popular service, the SS-3 framework will be met difficulty aspect the high availability keeping so.This is because produced 20 times the traffic of the original traffic in cluster of servers, capacity to trouble-proof load equalizer has caused very big pressure, and produces bottleneck and seriously reduced repayment based on the upgradability of SS-3 between interlayer communication.

Above-mentioned cluster of servers is based on the Single-Server framework, therefore can not be effectively and cost the solution of high degree of availability and upgradability is provided for the application of large-scale vital task based on web effectively.

Summary of the invention

The objective of the invention is by solve above-mentioned deficiency and set up based on the server array of working group with and framework aspect be achieved after seeking technological breakthrough.Through such technological breakthrough, just can be more effective and more cost realize being used on a large scale high degree of availability and highly upgradeable solution effectively based on the vital task application of web.

The present invention has used a plurality of groups/working group computing machine, hereinafter be referred to as to organize processor (TeamProcessor), they and a plurality of direct access type server based on working group, hereinafter being referred to as to organize server (TeamServer) is positioned on working group's computer rack together, hereinafter be referred to as to organize frame (TeamChassis), this point is stated in applicant's U.S. Patent No. 5,802,391.On the basis of these building blocks, can realize various workgroup server array configuration.

The present invention further comprises the modular control and the supervising device based on working group of a uniqueness, hereinafter be referred to as to organize control panel (TeamPanel), it provides local and remote monitoring and guide management again, task switching, and load balancing and fault prevent control function.In addition, the workgroup server array of any specific configuration can be regulated by independent one or more cascades group control panel together.

The present invention further comprises a plurality of above-mentioned group building blocks, can make up like this be used for various configurations preferred workgroup server array so that the basic function of a plurality of uniquenesses to be provided.On the basis of preferred data structure and data stream, these basic functions include but is not limited to: the controlled task switching of inner/outer, share load balancing, anti-fault, monitoring and management, safety and Performance Detection based on the equipment of working group.

The present invention and related framework thereof the assembly and replacing by removing unnecessary visit Network Based has solved traditional based on the intrinsic deficiency in the framework of individual server with the direct access type assembly based on working group, therefore also just reduce unnecessary Internet traffic, reduced the quantity of Single Point of Faliure.

And, can form Workgroup Servers based on a plurality of workgroup server arrays of application-specific and troop, so just can adapt to based on the available and upgradeable vital task web service of the height on this specific application.In addition, can make up a plurality of Workgroup Servers with serial and parallel mode and troop morely, use when being used to adapt to thousands of user and keep broadband quality of services (QOS) constant so that the large-scale solutions of using based on web to be provided based on different application.

Description of drawings

Above-mentioned aspect of the present invention and advantage and its others and advantage are by can comprehensively being understood a following detailed description of preferred embodiment with reference to following accompanying drawing.

Figure 1A is a theory diagram, has described the preferred working group processor as one of device that is used to set up preferred workgroup server array, just organizes processor.

Figure 1B is a theory diagram, has described the preferred working group computer rack as one of device that is used to set up preferred workgroup server array, just organizes frame, and this group frame is used to settle a plurality of groups of processors.

Fig. 1 C is a theory diagram, has described a preferred integrated morphology as one embodiment of the present of invention, and it comprises by a plurality of links is networked and form 8 of working group preferred group processors and 4 preferred group servers.

Fig. 1 D is a theory diagram, preferred modular monitoring and management based on working group have been described, just organize control panel, it comprise 4 basic control units with 1 have a dual processor be used to be connected the nearly main control unit of 4 group processors, and can be encapsulated in the group frame with built-in front control plate.

Fig. 1 E is a theory diagram, has described the modularization cascade of main group of control panel and assisted group control panel, can hold the structure of 8 group processors.

Fig. 2 A is a theory diagram, the preferred workgroup server array of an embodiment of system according to the invention has been described, it comprises 8 group processors, 4 based on the group server of scsi disk and the group control panel of 2 cascades, and all above assemblies all evenly are encapsulated in 2 group frames.

Fig. 2 B is a theory diagram, the preferred workgroup server array of an embodiment of system according to the invention has been described, it comprises 4 group processors, 2 group server and 1 group control panel based on scsi disk, and all above assemblies all are encapsulated in 1 group frame.

Fig. 2 C is a theory diagram, the preferred workgroup server array of an embodiment of system according to the invention has been described, it comprises 12 group processors, 6 utilize two SCSI passages links based on the group server of scsi disk and the group control panel of 3 cascades, all above assemblies all evenly are encapsulated in 3 and are organized in the frames.

Fig. 3 A is a theory diagram, described that the systematization of preferred data structure and data stream realizes on preferred 8 group processor-server arrays, in this server array, can realize a plurality of and built-in function, anti-fault, load balancing, security, management and optimal performance detect the basic function that uses together.

Fig. 3 B is a theory diagram, having described the Workgroup Servers that comprises a plurality of single application workgroup server arrays troops, wherein above-mentioned each workgroup server array provides the database storing section of a mutual exclusion on the basis that optimal performance detects, so just can realize based on the basic function between working group, for example Gao Du availability and upgradability.

Fig. 4 is a theory diagram, described an application oriented Workgroup Servers that preferably utilizes FC-AL hub or FC switch to integrate and trooped and back end database server, so just in a distributed computing environment, createed a preferred data center/warehouse structure for application based on the vital task of web based on various places of safety.

Embodiment

Below will be on the preferred embodiments of the present invention described in Fig. 1-4 be described as the basis of group/working group's computing machine of the building block of preferred workgroup server array.

Group/working group's computing machine is a sets of computer, and they are combined by working group's peer link, and can link to each other with a large amount of directly access type Workgroup Servers by the Workgroup Servers link.These details are No.5 at applicant's U.S. Patent number, 530,892 " having the single chassis multicomputer system that the intersection that is used for working group's telegon interconnects independent display of having of conversion and keyboard " by name and applicant's U.S. Patent number No.5, all have a detailed description in 802,391 by name " the direct access type group/Workgroup Servers that does not use group/working group's computing machine of network operating system to be shared ".In addition, working group's peer link and Workgroup Servers link (if using identical Physical layer cable) can link together, can move based on the storage with based on the data link protocol of communicate by letter, for example improved SCSI, this point in two above-mentioned patents all to some extent the description.These carry out the computing machine that working group connects, the back is referred to as to organize processor, all based on identical or different CPU/OS platforms, and these direct access type Workgroup Servers, the back is referred to as to organize server, can be designed as the identical file system of being supported by different operating system.The group server can be by based on disk, based on tape and realize based on the driver of CD, also can realize by fault-tolerant disk array.

It is the Workgroup Servers LI(link interface) card at center that each group processor based on specific OS all is equipped with this feature OS, just organizes server card, being all group server identification direct access type local drive.But each group server has only a main group processor to have absolute privilege to read, and writes and create file.And a physical hard drive and an error-tolerance type disk array can and be formatted as a plurality of logical drives by subregion, and each logical drive is controlled by each different group processor as primary processor.Even all these group processors all are connected on the internal network link and network operating system is installed, but but these group servers are not mapped as the driver of access to netwoks in the group processor.

And, highly integrated group/working group's computing machine, the back is referred to as the TeamPro computing machine, comprise a plurality of groups of processors, they all are encapsulated in working group's group frame, this point is that 5,577,205 " frame that is used for multicomputer system " by name is described at applicant's U.S. Patent number.The TeamPro computing machine further is equipped with monitoring and management devices again, just organizes control panel, be used for by a control desk VGA monitor with one based on the device control of RAP (remote access port) be connected each and organize processor.It comprises 2 serial ports again, 1 keyboard, 1 system led, 1 hummer and 1 reset button, this point is No.5 at U.S. Patent number, 530,892 " having the single chassis multicomputer system that the intersection that is used for working group's telegon interconnects independent display of having of conversion and keyboard " by name is described.

Shown in Figure 1A, preferred group processor based on group/working group's computing machine based on the PC computing platform generally includes: the one tunnel, two road or four road Intel Pentium CPU, mainboard with 128MBRAM based on WINNT PCI, the floppy disk interface module, the ide interface module, the VGA card module, the sound card module, USB module, parallel interface module, the RAP module, use the network link LAN module of Ethernet, use working group's peer-to-peer network link module of Ethernet, use the working group's peer link module of SCSI and the Workgroup Servers link module of use SCSI.The group processor further be equipped with again based on the peripheral drive of module and device as: floppy disk, IDE dish and CD-ROM drive, VGA monitor, based on the digital camera of USB, mouse is based on the hub and the switch of Ethernet, SCSI dish and tape drive, printer and a cover loudspeaker.

Shown in Figure 1B, preferred working group computer rack is just organized frame, comprises 4 based on the group processors of CPU card with in a large number based on the driver and the equipment of module, for example based on disk and CD drive, SCSI driver and the group control panel of IDE.Same group frame can also comprise 2 group processors based on mainboard, has various driver and equipment based on module.The group frame further can also be equipped with the built-in redundancy power supply, intelligent electric power management, but heat interchange disk and fan and outside UPS.

The maximum quantity that can carry out the independent group processor of the formation workgroup server array that working group connects is subjected to the restriction of internal work group server-link.If the Workgroup Servers link uses SCSI-II, the effective length that can guarantee correct data transmission so is 6 meters, and the node number that can connect is 16.The group frame why Here it is can encapsulate at least 2 group processors is used to support better the scsi cable scheme based on working group's peer link, and second group processor extends cable and be used for outside the connection because cable is extended from the outside first group processor.Same group frame can also be settled 4 group processors based on the CPU card, makes that so just scsi cable is shorter.At present, 4 different SCSI standards are arranged, be respectively: FAST SCSI, Ultra SCSI, Ultra2, LVD SCSI and Ultra3, LVD SCSI.Each standard all has narrow (8) and wide (16) configuration.Therefore, preferred SCSI realizes using Ultra-wide LVD SCSI, and the maximum data transfer rate that it has is 160MB/ second, and cable length can reach 12 meters.

Fig. 1 C has shown that a preferred working group link is integrated, and wherein 8 preferred group processors are by the working group's peer link link that utilizes SCSI, and 4 group servers based on the SCSI hard disk are linked by the Workgroup Servers link that utilizes SCSI.These group processors link together with the identical scsi cable of group server by utilizing.Like this, each group processor just can directly be visited each group server and do not related to other group processor, particularly has the main group processor of absolute privilege.Shown in Fig. 1 C, each group server based on scsi disk has 2 logical drives, and each group processor is assigned with a logical drive and is endowed absolute privilege.The group server can only be in read-only mode by other non-main group processor access.

Fig. 1 C has also described working group's peer link of using Ethernet by the group link cards with ethernet concentrator, if use working group's peer link of SCSI that fault has taken place like this, use working group's peer link of Ethernet just to can be used as another alternative communication link, vice versa.The major advantage of implementing working group's peer link of use Ethernet is: the communication between the group processor in the working group can influence Internet traffic sharply, and the communication between the group processor of other working group.Use working group's peer link of Ethernet can hold the various group of communication between processor, for example based on the reflection driver, based on socket, based on safety encipher and secure decryption.Except SCSI, also can adopt other suitable peripheral bus working group's peer link and Workgroup Servers link to be linked at together as the link of reality, if their data link layer can carry out based on storage with based on standard of communicating by letter or proprietary protocol.But, depend on configuration, as long as it is set up Workgroup Servers link and the working group's peer link of using Ethernet, all dispensable based on working group's peer link of any peripheral bus that is suitable for.

Fig. 1 D has described the pattern of preferred group of control panel, and it comprises 4 basic control units and 1 main control unit and can pass through RAP that VGA, USB are connected nearly 4 group processors with audio port.Basic control unit comprises a microprocessor and 3 switches, and to allow the VGA signal, sound signal flow into the public VGA that links to each other with main control unit with other basic control unit, audio frequency and usb bus with usb signal to switch by microprocessor control.In addition, also have one to use I ²The communication link based on the group control panel of C, it links to each other with main control unit with other basic control unit.In addition, also have one group of 10 interface signal, they link to each other with front control plate.

Preferred main control unit can comprise and be used for fault-tolerant dual micro processor, and they provide physical layer interface to connect keyboard, and based on the equipment and the printer of serial ports, they can classify as the working group's shareable device between the group processor that connects with working group.Main control unit also keeps various state tables to follow the tracks of each signal of interest with the group processor of working group's connection, cpu load and activity, and the counting rate meter that utilizes that is used to manage common bus and peripherals, so that after conflict is not used in having detected this table, the request that the processor of execution group is sequentially sent.

Preferred front control plate comprises two Interactive Buttons: one is used to select selected group of processor to be used for the demonstration of outside VGA monitor and the control of external keyboard and mouse, and another is used to reset selected group of processor.Also have three groups of LED, they represent electric power on/off respectively, activity that principal series is overall and the selection that is allowed.Group control panel feature board and front control plate are encapsulated in the group frame so that can more easily arrange cable configurations.

The default group processor that is used for control group control panel is referred to as group manager (TeamManager).For the working group communication that is dealt into group manager, any group processor at first is sent to coupled control module to the COM2 of message by RAP, and control module utilizes I then ²The C protocol header is repacked this message and by using I ²The group control panel inner link notice main control unit of C.Take place in case main control unit allows to connect, basic control unit is just by the inner I of group control panel ²C link and group manager direct communication, thus the current state of coupled group processor for example reported.And group control panel inner link can be as another kind of alternative communication link of working group's peer link of using SCSI and Ethernet.Equally for trouble-proof purpose, use based on the mouse of USB and replace mouse based on COM-1.Therefore, if fault has taken place the COM2 of RAP, so replace and provide data communication organizing between processor and the continuous basic control unit thereof with the COM1 of RAP.

Fig. 1 E has shown that 2 cascades group control panel together links to each other with 8 group processors that preferably are connected with working group.First group control panel (TP-408M just) and second group control panel (TP-408C just) are by public VGA, Audio, USB and I ²The C bus links to each other, but because TP-408C does not have main control unit, so the main control unit among the TP-408M will be managed all basic control units among the TP-408C.The group manager of controlling first group control panel also is the group manager of second group control panel.For with the communicating by letter of group manager, to coupled control module, this control module utilizes I to the COM2 that any group processor of second group control panel can be by RAP then transmission of messages ²The C protocol header is repacked this message and by inner I ²The C link is notified the main control unit in first group control panel.In case main control unit allows to connect, the basic control unit of second group control panel will be by the inner I of group control panel ²C link group manager direct and first group control panel communicates.According to same situation, the workgroup server array of any particular arrangement can be adapted to by single group of control panel or a plurality of cascade group control panel together.The front control plate of each group control panel can be encapsulated in each group frame or also can extend to an external control box so that the monitoring of a plurality of groups of control panels and control.The a plurality of groups of frames that the group processor of all working group server array is housed can be placed in the group rack (TeamRack), and this group rack can also be settled the other group server in the other group frame and further can also be equipped with a cable hut to settle cable and all turnover cables between all group frames.

Fig. 2 A is a theory diagram, the preferred workgroup server array of an embodiment of system according to the invention has been described, it comprises 8 group processors, 4 based on the group server of scsi disk and the group control panel of 2 cascades, and all above assemblies all are encapsulated in 2 group frames that can further be placed in the group rack.

Fig. 2 B is a theory diagram, the preferred workgroup server array of an embodiment of system according to the invention has been described, it comprises 4 group processors, 2 group server and 1 group control panel based on scsi disk, all above assemblies all are encapsulated in 1 group frame that can further be placed in the group rack.

Fig. 2 C is a theory diagram, the preferred workgroup server array of an embodiment of system according to the invention has been described, it comprises 12 group processors, based on the group server of scsi disk and the group control panel of 3 cascades, all above assemblies all are encapsulated in 3 and can further be placed in the group frames of organizing in the rack by 6 of two Workgroup Servers links links using two SCSI passages.

Fig. 3 A has described a preferred configuration with definition of data stream, and these data stream utilize 8 group processor workgroup server arrays shown in Fig. 2 A to carry out various basic functions.Preferably dispose according to this, 8 group processors can be divided into two classes on function: 1) application/file service processor (TP1-TP4), 2) database/file service/load balancing/fire wall processor (TP5-TP8).Each group processor has its main scsi disk group server, and it can be used as read-only group of server of other 7 group processors and operates, and this read-only group of server back is referred to as the assisted group server.Therefore, in bootup process, each group processor will be discerned system drive and a main group server and 7 the assisted group servers based on IDE, and these group servers need not use NOS reflection scheme just can be used as direct access type server.In addition, above-mentioned main group server and assisted group server by all group processor access that connect with working group can also be equipped with a plurality of fault-tolerant disk arrays and binary channels group server card with allocate communications amount on two SCSI passages.

Based on each can both be handled from the application oriented web inquiry and the generation based on HTTP of the Internet and can coil and the main batch transaction file of organizing on the server by writing system IDE among the group processor TP1-TP4 of application/file service.Based among the group processor TP5-TP8 of database/file service each can both handle from in-house network and extranets based on FTP or based on the web inquiry of the data base-oriented of proprietary real-time socket port and produce can writing system IDE dish and main group server on batch transaction file.

In addition, each all keeps a special-purpose working group database among group processor TP5 and the TP7, and this database is installed on its main group server.These two databases are essentially identical at one day final time.Each batch transaction file that the database that TP5 controlled can be produced according to group server 1-group server 4 in the time period of setting (t) in this day upgrades.The database that TP7 controlled is that the final time in this day upgrades according to all batch transaction files that group server 1-group server 4 was produced in this day.TP6 handles most of web inquiry based on the data base-oriented of FTP from in-house network, so that TP5 can be from 6 retrievals of group server and new database more in each t time period.TP5 also can upgrade database at once according to the data base querying based on proprietary real-time socket port from in-house network.TP8 is default group processor, just controls the group manager of those two group control panels.

On the basis of preferred server to configuration, workgroup server array of the present invention hereinafter is referred to as WSA, can realize a large amount of unique functional services.

Can realize a kind of method for optimizing, make group manager (TP8) can coordinate all and also produce based on the behavior of managing with group processors that working group connects about cooperation of WSA server and supervisory service.These behaviors comprise: monitor the inventory of each group processor, and disk space and cpu busy percentage, these can be produced by the OS that is installed on each group processor.These behaviors also comprise the intrusion that may take place on each group processor with working group's connection in addition, remove and failure warning.Each group processor will be packed routinely and based on the status information of management and the COM2 by RAP these information be sent to its control module, and this control module notice main control unit waits for that also the OK signal is with by the inner I of group control panel ²The C link sends the instruction of Autonomous Control unit.In case receive the OK signal, this specific group processor just sends the communication information from its control module to the control module of group manager, and the latter sends the COM2 of this status information by RAP again to group manager.Group manager will always keep about all supervisor status tables with the group processor of working group's connection.

Can realize a kind of method for optimizing about WSA inside front control plate switch service, make: one receives from himself or any group processor requirement and checks that a specific group processor is whether still in the request of operation, group manager just sends this request to main control unit, the main control unit general further sends diagnosis request to the control module of that specific group processor.If not response, main control unit will be given notice to the control module of group manager, and this control module can be sent to group manager to the COM2 of this notice by RAP.Afterwards, group manager can be this warning information by network link to the response that sends and wait for the operator based on the supervisor console of LAN.The operator can be by the control of copy on the operational management control desk computing machine or similar software adapter group manager.In addition, group manager is equipped with video capture card and public VGA bus also to link to each other with the NTSC converter, and the VGA of any group processor shows that can be caught the VGA that enters group manager again shows like this.Therefore, by send the screen display of the group processor that " selections " ask to indicate group manager to catch and break down to main control unit, main control unit also will allow the control module of group manager to carry out follow-up communicating by letter with the control module of the group processor that breaks down.The operator can also also take action in view of the above to the information that the group processor transmission keyboard that breaks down knocks, but also diagnostic file is deposited in group manager in order to further analyzing.If the group processor that operator's decision resets and breaks down, he just indicates group manager to send the Reset command to the control module of the group processor that breaks down.That specific control module will trigger the circuit that resets that directly links to each other with the group processor that breaks down it is resetted.Bootup process will be hunted down, and shows and is stored on the group manager, so that the operator at remote management console computing machine place can monitor and react step by step according to bootup process.And the technician can further analyze with the problem identificatioin place and releases solution according to the diagnostic file of storage.

Can realize a kind of method for optimizing about the on-the-spot front control plate switch service of WSA, make the front control plate of local execute-in-place person on can use group frame by workgroup device such as VGA monitor based on the group control panel, one group of loudspeaker, keyboard and mouse are watched, and control and any group processor that resets.In case receive the button request that requires " selection " and " resetting " on the control panel, these request signals directly link to each other with main control unit, if suitable, main control unit will at first be checked and utilize counting rate meter to guarantee not use conflict, relevant LED flicker will be set then.Trigger if button is carried out in decision, local operator triggers with regard to pressing the button again, has so just opened relevant LED.In case operation is finished, and just closes relevant LED.

Can realize a kind of method for optimizing, make that arbitrary remote computer can be according to External modem control group manager or any the group processor of the proprietary access code that translates by linking to each other with serial link based on working group about the long-range front control plate switch service of WSA.In case set up communication, remote computer can carry out all with based on the supervisor console computing machine identical functions of LAN.

Can realize a kind of method for optimizing, make that the peripheral hardware among the WSA can be by other group processor access of group manager and any about WSA equipment share service.When specific group processor need be visited arbitrary peripheral hardware such as printer, the group processor sent request message by the COM2 of RAP to its control module, and control module is by inner I ²C chain road direction main control unit sends request.If it is available that main control unit is found after having detected state and having utilized counting rate meter, it just allows this specific control module and main control unit to carry out subsequent communications, and main control unit will be by built-in parallel interface to the printer Data transmission that is connected.Other serial equipment also can be realized similar process.But for USB device, specific group processor sends request signal by the COM2 of RAP to its control module, and control module will send to main control unit to it.If main control unit is found behind the counting rate meter availablely having detected utilizing of USB device, it just beams back the OK signal to this control module, and this control module is just further opened the USB switch on the panel.So, the USB interface on this specific group processor will be directly links to each other with USB device such as camcorder based on working group by public usb bus.

Can realize a kind of method for optimizing about the anti-fault scheme service of WSA, make vital task assembly among the WSA as the group frame, the group control panel, group processor and group server or fault-tolerant or trouble-proof, use just can not be destroyed for vital task like this.

As for the group control panel, the vital task performance is relevant with its main control unit, and main control unit has two microprocessors, so if fault has taken place for first, second will replace it and give the alarm to group manager, and group manager is further notified supervisor console again.As for the group frame, it is fault-tolerant, because it has been equipped with dual power supply and outside UPS.As for the group processor, always have 4 anti-fault groups, just TP1 and TP2, TP3 and TP4, TP5 and TP6, TP7 and TP8.Because every group member has identical hardware configuration each other, if so each the group in one fault has taken place, another will replace it, vice versa.Therefore, if fault has taken place group manager TP8, TP7 will replace as group manager.And, TP1-TP2 pair with TP3-TP4 to all being anti-fault group.TP5-TP6 pair with TP7-TP8 to also all being anti-fault group.If TP1-TP2 is to fault has taken place, TP3-TP4 is to replacing, and vice versa.Same situation also be applicable to TP5-TP6 pair with TP7-TP8 right.

As for group server based on file service, always have 8 anti-fault groups, just: IDE1 in the group processor 1 and group server 1, IDE2 and group server 2, IDE3 and group server 3, IDE4 and group server 4, IDE5 and group server 5, IDE6 and group server 6, IDE7 and group server 7, IDE8 and group server 8.Therefore, if fault has taken place group server 1, other group processor still can be from group processor 1 acquired information on the IDE1.If fault has taken place IDE1, other group processor can be directly from group server 1 acquired information.Same situation also is applicable to other 7 anti-fault groups.As for the group server based on database service, the database on the group server 5 is controlled by TP5, and the database on the group server 7 is controlled by TP7, and they all are identical private database basically as previously mentioned.But if fault has taken place database-TP5, according to all relevant autoexecs collected from group server 1 to group server 8, database-TP7 will be upgraded and be ready at once service at once by group processor 7.

Can realize a kind ofly, make group processor among the WSA to organize control panel and reach load balancing by use based on using about the method for optimizing of WSA based on the load balancing service of using.In environment, based on the Internet of the query requests of using from the use http protocol based on web.The input inquiry traffic at first will be passed through router.Router sends all requests to group manager TP8 then.Then group manager by inner FTP port or the proprietary port of working group's peer link that uses Ethernet the incoming traffic load to TP1, TP2, TP3 and TP4 distribute.In the process that circulation is carried out, according to each cpu busy percentage and response time based on the group processor of using, group manager (TP8) keeps one based on round-robin load balancing state table, and the main control unit of group control panel keeps various important sign state tables.

Because arbitrary group processor that connects with working group all can transmit the COM2 of important sign or the like by RAP by follow procedure to the control module of its connection, main control unit be repacked and be notified to control module will to data.In case main control unit allows the generation of connection, basic control unit just can download to data in the memory buffer unit of main control unit, and this impact damper can be distributed to various important sign state tables.According to these real-time state tables, which group processor main control unit can detect, and fault or overload may take place.If any has taken place in above-mentioned two kinds of situations, main control unit will be reported to group manager.If overload situations, group manager will be attempted this group processor is taken out from cyclic sequence at once, until receive the notice of main control unit once more, and till requiring to return that specific group processor in the cyclic sequence.If the situation of fault has taken place, group manager will be attempted to communicate by letter by the specific group processor foundation of working group's peer link and that.If not response, group manager " resetting " circuit that will notify main control unit to pass through RAP this group processor that resets causes part or all of recovery and carries out corresponding operating so.

Except circulation pattern, also have the intelligent algorithm of other type, as " minimum open circuit connection ", " the fastest detection time or response time ", " content type ", " number and other statistical figure of from application server, gathering that open circuit connects ".Because group manager TP8 can gather the information of these types one by one and detect the group processor break down by working group's peer link, therefore can intelligently carry out various algorithm and can not make a specific group processor overload also can not send to the group processor that breaks down to load.But if all group processors all are same kind, round-robin algorithm will be a best choice so, and group manager only need respond to the instruction from the main control unit of organizing control panel according to unusual situation and get final product.

Can realize a kind of method for optimizing, make that file and the database on any specific group server can directly visit and share between a plurality of groups of processors about WSA file and database service.Reach this point a plurality of read-only datas storehouse engine of similar number need be installed on a plurality of groups of processors of the assisted group server that is used for directly visiting, on main group processor a full powers database engine is installed, it has the absolute privilege that is applicable to the database on its main group server.In addition, group manager (TPB) is preserved a series of states and is utilized counting rate meter for the equipment that is connected to some extent.One of them table storage the tabulation main group processor of each group server the time, on any one group server, just do not have the situation that the dual data integrity that writes destroys like this and take place.But, since anti-fault, the existence of the factors such as temporary control and education change that the different operating in the different time zones need and be used to upgrade, and the main group processor of the group server that any is specific all may become another group processor.Group manager will always guarantee to have only a group processor can upgrade specific group server in preset time.

Can realize a kind of method for optimizing, make any unauthorized that can detect invade WSA about the WSA security service.Because group manager TP8 will receive all input requests and distribute loading in the group processor,, fire wall is installed to avoid any possible external attack so group manager must strengthen security.Basically, group manager TP8 can filter out any problematic input request by following measure: carry out based on SSL, visit encryption safe measure based on OS or higher level based on application, by working group's peer link of using Ethernet those legitimate request are redirected to those based on the group processor of using, are separated into two different places of safety.Each all will be replied based on the group processor of using, and this relates to the visit private database and by comprising that a correct implicit IP address with content-encrypt safety practice returns to the requestor to it again.Like this, group manager can be decrypted and it is redirected to the correct group processor of handling the front request content.The method of this adhesion port type is called as the persistent sessions based on following factor, and these factors for example comprise: be included in the user verifying device request protocol or the cookie that returns in source IP address and specific information.This adhesion port type method can also be carried out safely, and this point is necessary for the ecommerce that moves effectively based on web.

Can realize a kind of method for optimizing about the anti-failed services of WSA, make in a large number based on agency's management software program just TeamSoft can with merge based on the definition data structure of preferred disposition and all above-mentioned functions services of data stream.Have only current group manager that the server section of TeamSoft will be installed, remaining group of processor will be installed the client portion of TeamSoft.As long as a group processor operation is arranged, the computing machine of remote management console just can be controlled this group processor makes it serve as group manager, its any group processor that breaks down that just can reset like this, workgroup server array of the present invention just can be returned to normal operating condition.According to the anti-fault ability of TeamSoft, whether each group processor can still begin its detection in operation according to the corresponding assembly of its anti-fault by group manager.If not operation, this group processor will be born the task that its corresponding assembly that breaks down is born so.For example, if fault has taken place in TP5, group manager will the privilege of group server 5 and more the Task Distribution of new database to TP6.If fault has taken place TP6, group manager will be distributed to TP5 to the privilege of group server 6 and by the IP address of the IP address of TP5 rather than TP6 being notified with input request and the traffic of TP6 being redirected to TP5.TeamSoft also comprises the working group's diagnosis to the problem of built-in automatic corrigendum behavior.

Can realize a kind of method for optimizing, the feasible optimal performance that can obtain WSA by the value of adjusting some key parameters about the service of WSA Performance Detection.The performance of workgroup server array of the present invention and following three factors are closely related: the 1) operation of group manager fire wall, 2) based on the quantity of the group processor of using, 3) size of private database.If the fire wall action need that is installed among the group manager TP8 costs a lot of money the time, will reduce the input request quantity of per minute in execution contents decryption security and upper access security.But this problem can be solved based on the router of fire wall by connecting, and this router can be carried out the network layer filtration and filter on the upper strata.

If the quantity based on the group processor of using reduces, the output of per minute is replied number and also will be reduced.As for the consideration of database aspect, if application-centered database makes up according to the non-fidelity traffic, it tends to only provide ready-made information, and these ready-made information may increase by accident to satisfy the traffic based on non-fidelity.On the other hand, if make up database according to the fidelity traffic, database just will phenomenal growth so.Yet to reply page or leaf with formation be not a problem the required time to retrieve data from database, because the database on the group server is easy to visit and does not need to rely on any other group processor.

Therefore, two kinds of situations are arranged: 1) based on the application and 2 of non-fidelity) based on the application of fidelity.Under the situation based on the application of non-fidelity, the optimal performance of workgroup server array of the present invention depends on the number based on the group processor of using.According to the complexity of computing power and service, group processor can be handled the input request of X quantity and reduce output in one minute and reply and do not reduce service quality, the service quality that can accept (QOS).Therefore, four group processors can be with the input request of the operational processes 4X quantity of steady state (SS).The 6X if the peak value non-fidelity traffic constantly jumps, workgroup server array of the present invention still can adapt to the operation constantly of this peak value by TP6 and TP7 being assigned as based on the group processor of using and adding the round robin load-balancing algorithm of group manager operation.

And, shown in Fig. 2 C, workgroup server array based on 12 group processors, wherein 8 group processors that are based on application are arranged in 12, this server array can be handled the non-fidelity traffic of 8X quantity in steady state operation, handle the traffic of 10X quantity in peak value is operated constantly.If incoming traffic surpasses 10X and just needs second workgroup server array.

Under the situation based on fidelity, workgroup server array optimal performance of the present invention depends on based on the number of the group processor of using and based on the size of the database of fidelity.The quantity of the input request that is produced if the size of database is too big has surpassed the treatable scope of all group processors, the size that so just need reduce database is to satisfy the operation of steady state (SS), and the part that exceeds should be transferred to second workgroup server array.For example, a workgroup server array based on 12 group processors can be handled the traffic based on fidelity of 8X quantity, and the traffic of this 8X quantity can be converted to the user based on fidelity of the Y quantity that application-centered database can hold.At peak value constantly, the user of Y quantity will produce the traffic based on fidelity of 10X quantity, and this still satisfies acceptable QOS.

Workgroup server array of the present invention always can be readjusted the quantity of X and Y to guarantee acceptable service quality according to the collected information of group manager.Therefore, the performance index that are used for workgroup server array of the present invention are parameter X and Y, and can extrapolate optimum operating point and prediction in view of the above needs to increase problem of resource.

For higher bandwidth applications, service class is also just higher, and this can reduce the quantity of X and Y.But the QOS of workgroup server array of the present invention will be unaffected.

Under situation based on fidelity, in order to hold the input request of more application of the same race, Fig. 3 B has described a kind of Workgroup Servers of a plurality of workgroup server arrays based on single application that comprises and has trooped, and its each included server array has mutually exclusive database segment.Because each workgroup server array has the QOS ability,, whole Workgroup Servers also has the QOS ability so trooping.

So, Workgroup Servers troop can adapt to highly available, the application based on web of highly upgradeable vital task, this cluster of servers can comprise from first until n workgroup server array.Because router is based on fidelity, it can distribute correct incoming traffic according to correct IP address to correct group manager at once, because this information or be installed among their " cookies " of browser or be installed in the smart card based on chip, wherein this smart card can be used to access to netwoks and user rs authentication.

For situation based on non-fidelity, router and can utilize the name server (DNS) that URL is converted to the IP address built-in circulation ability input load to be distributed to a plurality of group managers of trooping together based on the Workgroup Servers of non-fidelity.So, it is unaffected just to have carried out the load balancing and the QOS that are used for based on the traffic of non-fidelity.This basis is added based on round-robin DNS based on the peculiar methods of the load balancing that Workgroup Servers is trooped, produced tangible benefit for removing the GSLB device, this load equalizer must be enough powerful in equally loaded and all web application servers of management, will produce the network switching overload that unnecessary Internet traffic makes interlayer like this.And if any group manager breaks down, DNS will initiate a message to the corresponding assembly of the anti-fault of this group manager.This correspondence assembly will replace former assembly automatically and handle incoming traffic from DNS, because DNS had both notified the IP address of group manager also to notify the IP address of its corresponding assembly.

No matter be for based on the situation of fidelity or for situation based on non-fidelity, the database server program run is got up all should be rapidly and simply, and do not need the intelligence of inherent complexity, fine because the fine and relative database that is defined based on the application of web also should be defined.The time that retrieve data spent should be short as far as possible, so that X and Y can be that bigger number is to produce more performance.

Because the input request from user/surfer may relate to a lot of different application based on web,, a plurality of different Workgroup Servers based on using troop so should being installed.Fig. 4 has shown that is used for a high degree of availability, the vital task in-house network of height upgradability, the preferred embodiment of the whole web server system that extranets and the Internet are used, it is also integrated a plurality of serial links troop with Workgroup Servers parallel link and create one desirable, the distributed computing environment of safety.

Except the region security of use based on the workgroup server array of fire wall, intercommunication mutually between the different operating cohort collection also can have based on SSL by use, the proprietary port of the content of OS or application and access security measure and being carried out safely, any like this PERCOM peripheral communication does not allow to visit any Workgroup Servers and troops.

And, by using FC-AL or the like that all group managers are linked at together, the group server of each workgroup server array, no matter be based on hard disk, tape, or CD can be converted to FC equipment, these FC equipment can be by any back-end data base processor access and maintenance based on SAN (storage area network).Make the application-centered file and the database server of each workgroup server array or be used for the Back end data center and can be equipped with more intelligent database engine like this based on the complicated file of SAN and the data cache server of database server.

In a word, present invention includes the assembly of a large amount of uniquenesses: 1) group processor, 2) group server and group server card, 3) the group control panel, 4) the group link cards, 5) the group frame, 6) the group rack.On the basis of these distinct elements, the present invention has also used the method for a large amount of uniquenesses to set up preferred workgroup server array.They are: 1) the WSA server is to method, 2) many working group's link methods of WSA, 3) coordination of WSA server and management method, 4) WSA inside, on-the-spot and long-range " front control plate " method of switching, 5) WSA equipment is shared method, 6) the anti-Failure And Recovery method of WSA, 7) the WSA load-balancing method, 8) WSA file/data share method, 9) WSA is based on the method for security, 10) WSA is based on the management method of TeamSoft, 11) WSA optimal performance detection method.And on the basis of the workgroup server array of these inventions, the present invention has also used a large amount of peculiar methods to set up preferred Workgroup Servers troop (WSC).They are: 1) WSC structural approach, 2) the WSC load-balancing method, 3) WSC is the data base method at center with the high-speed cache, 4) WSC is working group's data base method at center with the user rs authentication fidelity.At last, on the basis that the Workgroup Servers of these inventions is trooped, the present invention has used the method for a large amount of uniquenesses to set up the server farm based on web of preferred " the place ahead office ".They are: 1) the serial link method of a plurality of WSC, 2) the parallel link method of a plurality of WSC, 3) method that is connected with storage area network (SAN) of the WSC of a plurality of serials and parallel link.

Should be understood that now: the invention provides a kind of workgroup server array and relevant framework thereof, in the computing environment of a safe distribution, to set up various high degree of availability, the vital task cluster of servers of upgradability.

Those skilled in the art is easy to find some other advantage and the place that can revise.Therefore the present invention is not limited to some specific details with regard to its wideer scope, representative device, and some examples shown and that describe.Therefore only otherwise break away from the spirit and scope of applicant's general inventive concept, allow and such details generation deviation.The present invention defines according to following claims.

Claims (40)

1. multiprocessor computer system that comprises a plurality of groups of processors, each group processor has CPU computing platform more than, described platform have RAM and a plurality of respectively with the VGA monitor, loudspeaker, access port, USB, network link, working group's peer link, the module that peripheral drive links to each other with the Workgroup Servers link;

Described multiprocessor computer system also comprises at least one group control panel, this group control panel has a plurality of control CPU, and the group processor that each CPU is corresponding with links to each other, and each CPU controls one group of switch, these switches are used for selectively corresponding group of processor and USB device, the VGA monitor, loudspeaker, keyboard, mouse, modulator-demodular unit, printer links to each other with serial device

Wherein, the described Workgroup Servers link of described each group processor links to each other with one group of Workgroup Servers of sharing.

2. the computer system of claim 1, wherein each described Workgroup Servers all is a fault tolerant disk array.

3. the computer system of claim 1 further comprises at least one frame, can settle at least 4 described group of processors and at least 2 described Workgroup Servers.

4. the computer system of claim 1 comprises at least 8 described group of processors and at least 4 described Workgroup Servers; Described each group processor all communicates with described each Workgroup Servers.

5. the computer system of claim 1 further comprises one group of cascade described group of control panel together, is used to connect more described group of processor of more number.

6. the computer system of claim 1, wherein: will be selected from least one among described a plurality of control CPU, be appointed as first main control unit, and provide described group of processor access shared bus and equipment, and the management of the load balancing between described group of processor.

7. the computer system of claim 6, one of them described group of processor is designated as group manager, controls described group of control panel.

8. the computer system of claim 1 wherein has at least a described Workgroup Servers to be mainly used in the application file service, has at least another described Workgroup Servers to be mainly used in the database file service.

9. the computer system of claim 8, the wherein said Workgroup Servers that is used for the application file service provide a kind of ability for application oriented request service.

10. the computer system of claim 8, the wherein said Workgroup Servers that is used for the database file service provide a kind of ability of the request service for data base-oriented.

11. the computer system of claim 8, wherein each described Workgroup Servers is used to produce and can writes peripheral hardware to be used for the autoexec of anti-failed operation.

12. the computer system of claim 1, wherein Workgroup Servers of each group processor identification is a master server, and all other Workgroup Servers is a secondary server.

13. the computer system of claim 1, wherein said working group peer link are based on Ethernet or scsi interface, and described Workgroup Servers link is based on scsi interface.

14. the computer system of claim 1, one of them described group of processor is designated as group manager, monitors the state of remaining each group processor.

15. the computer system of claim 1, wherein each described Workgroup Servers can be at least one described group of directly visit of processor institute.

16. a workgroup server array system comprises:

A plurality of groups of processor groups, each group processor has CPU computing platform more than, and each group processor group comprises a plurality of groups of processors, and links to each other with the group control panel with a plurality of control CPU, is used for the monitor with VGA, keyboard, loudspeaker is connected with universal serial bus; Each described group of processor group is connected on the Workgroup Servers link; A plurality of direct access type Workgroup Servers link to each other with described group of processor group by described Workgroup Servers link; Described group of processor group is interconnected on working group's ethernet concentrator by working group's peer link; Described group of processor group also is connected on the LAN (Local Area Network) by working group's Ethernet switch.

17. the workgroup server array system of claim 16, wherein each described Workgroup Servers is a fault tolerant disk array.

18. the workgroup server array system of claim 16 further comprises: at least one frame is used to hold at least four described group of processors and at least two described Workgroup Servers.

19. the workgroup server array system of claim 16 comprises: at least 8 described group of processors and at least 4 described Workgroup Servers; Each described group of processor and each described Workgroup Servers communicate.

20. the workgroup server array system of claim 16 further comprises one group of cascade described group of control panel together, is used to connect described group of processor of increase.

21. the workgroup server array system of claim 20, wherein: will be selected from least one among described a plurality of control CPU, be appointed as first main control unit, and provide described group of processor access shared bus and equipment, and the management of the load balancing between described group of processor.

22. the workgroup server array system of claim 21, one of them described group of processor is designated as group manager, controls described group of control panel.

23. the workgroup server array system of claim 16 wherein has at least a described Workgroup Servers to be mainly used in the application file service, has at least another described Workgroup Servers to be mainly used in the database file service.

24. the workgroup server array system of claim 23, the wherein said Workgroup Servers that is used for the application file service provides a kind of ability for application oriented request service.

25. the workgroup server array system of claim 23, it is the ability of the request service of data base-oriented that the wherein said Workgroup Servers that is used for the database file service provides a kind of.

26. the workgroup server array system of claim 23, wherein each described Workgroup Servers is used to produce and can writes peripheral hardware to be used for the autoexec of anti-failed operation.

27. the workgroup server array system of claim 16, wherein Workgroup Servers of each group processor identification is a master server, and all other Workgroup Servers is a secondary server.

28. the workgroup server array system of claim 16, wherein said working group peer link are based on Ethernet or scsi interface, and described Workgroup Servers link is based on scsi interface.

29. the workgroup server array system of claim 16, one of them described group of processor is designated as group manager, monitors the state of remaining each group processor.

30. the workgroup server array system of claim 16, wherein each described Workgroup Servers can be at least one described group of directly visit of processor institute.

31. the workgroup server array system of claim 16, wherein: each group processor is associated with a described Workgroup Servers as master server, and is associated with remaining described Workgroup Servers as secondary server.

32. the workgroup server array system of claim 31, wherein each master server provides Full Featured database access for the group processor of selecting it, and other described group of processor provides the visit of read-only data storehouse to all.

33. the workgroup server array system of claim 16, described group of processor of some of them is based on application, some described group of processors are based on database, and some described servers are based on the server of application, and some are based on the server of database; This system further comprises:

Be used to adjust the device that is used to carry out the capability and performance of incoming traffic based on the quantity of group processor of using and server with optimization.

34. a group server is coordinated and management method, relate to multiprocessor computer system with a plurality of groups of processors, each group processor has CPU computing platform more than and a Workgroup Servers link, this link is used for the group processor is linked to each other with a plurality of direct access type group server of sharing, and this method may further comprise the steps:

Selecting first described group of processor is group manager;

Utilize described selected group manager to monitor the state of all remaining set processors;

Each group processor sends its status information to described group manager;

Corresponding to the described group of status information that processor received, state table of described group manager editor based on management.

35. the method for claim 34, wherein each described group of processor links to each other with public VGA monitor by the VGA link and further may further comprise the steps:

Utilize the diagnosis of described group manager monitoring to each described group of processor;

Catch arbitrary VGA link that the group processor of fault has taken place by group manager;

The group processor that described group manager is resetted fault has taken place.

36. the method for claim 34 further may further comprise the steps: distribute at least one group processor to be used for load balancing; Distribute at least one group processor to be used for database service; Distributing at least, another group processor is used for service-specific.

37. the method for claim 34 further comprises step: group processor pairing to provide fault-tolerant alternative by another that replaces this centering with a group processor.

38. the Workgroup Servers with a plurality of interconnective workgroup server arrays is trooped, each such server array comprises:

A plurality of groups of processor groups, each group processor has CPU computing platform more than, and each group processor group comprises a plurality of groups of processors, and links to each other with the group control panel, is used for the monitor with VGA, keyboard, loudspeaker is connected with universal serial bus; Each described group of processor group is connected on the Workgroup Servers link; A plurality of direct access type Workgroup Servers link to each other with described group of processor group by described Workgroup Servers link; Described group of processor group is interconnected on working group's ethernet concentrator by working group's peer link; Described group of processor group also is connected on the LAN (Local Area Network) by working group's Ethernet switch.

39. the Workgroup Servers of claim 38 is trooped, and further comprises a name server, is used to distribute incoming traffic.

40. the Workgroup Servers of claim 38 is trooped, wherein: described trooping can be carried out serial and be connected the upgrading that makes progress with parallel by trooping with a plurality of additional Workgroup Servers.

Images