CN1441933A - Method and system for providing dynamic hosted sevvice management - Google Patents

Abstract

A hosted service provider for the Internet is operated so as to provide dynamic management of hosted services across disparate curstomer accounts and/or geographically distinct sites (80). A plurality of individual servers are allocated to a common administrative group defined for that customer account and configured to access software and data unique to that customer account to provide hosted services for that customer account. The system automatically monitors the performance and health of the servers in each administrative group. At least one server from a first administrative (52a) group is automatically and dynamically reallocated to a second administrative group (52b) in response to the automatic monitoring. Each administrative group includes a local decision software program that communicates with a master decision software program that determines when and how to dynamically reallocate servers to different administrative work groups in response to usage demands, available resources and service level agreements for different customer accounts.

Description

Be used to provide the method and system of dynamic hosted sevvice management

Technical field

The present invention relates generally to the data processing business field.More particularly, the present invention relates to be used to stride the method and system that different clients' accounts and/or diverse geographic location provide the dynamic management of trusteeship service.

Background of invention

The explosive increase of the Internet is to be promoted by the appearance of gyp service provider such as ISP (ISP), application service provider (ASP), independent software vendor (ISV), enterprise solution provider (ESP), managed service provider (MSP) and deposit mechanism to a great extent.Though the services set accurately that each service in these business is provided does not have clear and definite definition, but these service providers and deposit mechanism provide the service of customization usually, with satisfy the client aspect application program trustship, website exploitation, managing eBusiness and the server disposition some, great majority or whole demands so that earning erected cost and regular expense.For example, under the situation of server disposition, expense is this client's of trustship the application program or the specific hardware and software configuration of website appointment based on the client usually.For the present invention, term " trusteeship service " comprises various types of these services that these a series of service providers and deposit mechanism provide.For simplicity, these a series of service providers and deposit mechanism should be referred to as managed service provider (HSP).

Gyp HSP provides visit to the application program of trustship on the Internet to the user, provides being connected of caller by international telephony network and their plan as telephone operator to the client.HSP their application program of providing of trustship and the computer equipment of service is provided is commonly called server.The simplest form of server can be a personal computer, and it is connected to the Internet by network interface, and operates to the request that the client that satisfies client or this server makes and the specific software that designs.Supply a pattern for all that can be used to provide trusteeship service by HSP, most of HSP use the set of the server that is connected to internal network, this set is commonly called " server farm ", each station server is all carried out the load that exclusive task or group of server are shared a plurality of tasks, as mail server, Web server, access server, record keeping and management server.For example, under the situation of trustship website, the client's frequent with less website provides support in single Web server and by single Web server.Yet bigger website is trustship on the special-purpose Web server that service only is provided for this website usually.General background information about the Internet and HSP sees also GeoffHuston, ISP Survival Guide: the strategy that moves competitive ISP(1999).

Along with the demand to Internet service is increasing, need bigger capacity could satisfy this demand.A solution is to utilize computer system with better function as server.Use large scale computer and medium-size computer system to serve large-scale website and enterprise network as server.Most of HSP are just tending to not utilize these bigger computer systems, because expense height, the complexity of such system and lack dirigibility.On the contrary, HSP more is ready to utilize server farm, and comprising a large amount of individual pc servers, they are connected to public the Internet and connect or pool of modem, also visit public disc driver collection sometimes.For example, when HSP added new trusteeship service client, one or more personal computer server were manually added to the HSP server farm, and loaded appropriate software and data (for example, web content) for this client.So, HSP only is deployed as and supports needed this hardware-level of its current levels of clients.It is also important that HSP can collect installation fee to its client, this expense can compensate most of hardware cost.By utilizing this method, HSP needn't be in advance for having the large computer system expense of idle capacity, and idle capacity can not produce directly income for HSP.All enjoy the situation of serving from single bigger large scale computer with the client and compare, the server farm solution also provides solution more easily for the problem of striding different customer care safety and data integrity.If only load software at this client at a client's Servers-all, and only be connected to the data at this client, the security of this client's information can be guaranteed by physical isolation so.

For HSP, can provide a lot of software accounting program bags for the account, and collect the expense of these metering service, as from Tens.comXaCCT and from Inovaware.comHSP Power.Developed other software programs helping management HSP network, as from Lightspeedsystems.comIP Magic, from Resonate.comInternet Services Management and from Luminate.comMAMBA.The management of HSP and operation also are such as the Hursti in the internet symposial on April 19th, 1999, the article of Jani " management that accesses network and service provide " and so on and the theme of symposial.Typically provide various configurations hardware, software, maintenance and for example that the HSP that commerciality other access to the Internet of level and website custody support is provided with every month pay charge way can Rackspace.comFind.

Up to the present, the HSP method that makes up their server farm has two kinds.Homology personal computer system's (hardware and software) group that a kind of method is to use single manufacturer to provide.The personal computer system that the another one method is to use many different manufacturers to provide.The homology method provides the advantage that need only support the individual server platform for HSP, but also HSP is limited to this individual server platform simultaneously.Use the allos method of the system that different manufacturers provides more flexible, and for HSP provides the advantage of utilizing optimal server hardware and software platform for given client or task, but this dirigibility is that to increase with complicacy and the support challenge related with safeguarding a plurality of server platforms be cost.

No matter it is use any method to set up server farm, generally speaking still identical to the actual physics management of such server farm.When the client need increase or be reduced to the volume of services that their account provides, HSP added data storage and the interconnected part of network in the website that is directly connected to this client of HSP server farm manually server or therefrom deletes server.Add at needs under the situation of service, typical process will be some mutation of following situation: the order of (a) receiving the change service class from the trusteeship service client there, (b) HSP buys new server hardware to satisfy the change of request, (c) personnel of HSP are from physically installing new server hardware in the position at server farm place, (d) data storage of the cable of new server hardware being added to this website is connected with network, (e) software loading with server hardware arrives server, the personnel of HSP carry out a series of initialization steps with the configuration software that requires at this clients' accounts, and (f) new server that install and configuration fully joins the existing management group that the server of trusteeship service is provided at client's account.No matter be any situation, each server farm all has been assigned to specific client, and must be configured to satisfy the greatest requirements from the expection of this clients' accounts.

At first, need restart some or all existing servers in the management group of given clients' accounts and just can finish last step of this process, add this situation of new server in the management group because pointer in the existing server and table manually upgrade needs to be reflected in.This requirement has determined the change in the server hardware regularly to take place at the well-defined service window such as the late into the night on Sun..Recently, developed and expanded software and similar program such as MicrosoftWindows 2000, Microsoft Cluster Server, Oracle Parallel Server, window networking load balance service (NLB), making new server join existing management group at any time automatically, rather than at these well-defined windows.

In U.S. Patent No. 5,951, the example how new server joins existing management group has automatically been described in 694.In this patent, represented the Servers-all in the management group in the mapping table by the gateway server maintenance.Mapping table has identified the different service groups of management group, as mail service group, database service group, access server group etc.Gateway server will be routed to suitable service groups to the request of management group based on mapping table.New server can add one of them service groups to by load the appropriate software assembly on this server, after this, gateway server will be discerned new server, and add it to mapping table, and use the transaction journal of safeguarding for each service groups to make all the other servers maintenances in new server and this service groups synchronous.Perhaps, if the operating load of a service groups is heavy especially, the load of another service groups is light, then server can be switched to another group from a service groups.This patent has been described a software routines of carrying out on the management server of special use, this software working load balance scheme is revised mapping table, to guarantee the request more balancedly distribution in the various service groups of forming the management group to this management group.

A lot of patents described single troop or the management group in server in the technology of balance operating load.U.S. Patent No. 6,006,529 have described and have comprised that the security that is subjected to master server control and the software of heartbeat signal layout troops, and at this, have distributed a public IP address for all making cluster members, and this is trooped in the execution load balance.United States Patent(USP) Nos. 5,537,542,5,948,065 and 5,974,462 have described the various operating load balanced distributions of the multisystem computer disposal system with shared data space.Operating load is distributed in multiple servers and can also finish by insert intermediate system between client and server.U.S. Patent No. 6,097,882 have described in the reproducer system that inserts between the client and server to be redirected the IP packet based on server availability and operating load between client and server pellucidly.

Also used various technology to coordinate many computing machines in single the trooping or the operation of server.U.S. Patent No. 6,014, the operation of trooping that the 669 use lock step decentralized configuration files of having described the multiple servers in single the trooping are carried out.U.S. Patent No. 6,088,727 have described the control of trooping in the shared data space multicomputer environment.Other patents have been described and how to be used the single reflection in I/O space to coordinate many computing machines.U.S. Patent No. 5,832,222 have described the single reflection that how to use the I/O space coordinates the geographical distributed computers system that goes up.U.S. Patent No. 6,067,545 described have shared metadata management, the distributed file system of replicated setup database and territory load balance, server is added under the control of configuration database and leave individual domain.

Although these method improvements the management of the server in management group, territory or the shared data space, do not have ability with these technological expansion outside the group of server that defines for common operation system or public shared data space and link with them.In general, this restriction is not regarded as problem, because all these methods are at management in the computer network of single company and the bigger enterprise computing system realized.Although these methods can be used for managing the server of the certain accounts that is assigned to given client by HSP, the neither one method can make the HSP management that one group of server of trusteeship service is provided to a plurality of accounts of different clients in these methods.

Also developed the system of the operation that is used to manage bigger enterprise computing system, as from the OpenView of Hewlett-Packard, from the Unicenter TNG of Computer Associates, from the Tivoli of IBM, from the Mamba of Laminate, and from BMC Software, the Patrol of Inc..In general, these system side overweight inventory management and the Software deployment control problem that a large amount of computing machines of moving run in single company or mechanism.Some systems in these Operational Management Systems comprise the performance monitoring solution, by the performance of the in-house server of network inquiry, with the demand of determining other resources or load are distributed again.The method of similarly " passing through network " also is used to the report and the management function that provide concentrated.Plan is that the Tivoli service provides management platform for a good example of such Operational Management System that HSP uses, this platform comprises user management module, software dispatch module, checklist module, enterprise console, security module, enterprise manager module, the assembly of this module in network be added to all these assemblies are provided after the network can self-defining view, can make operating load balanced operating load proposal program in the server in shared common data space.All these modules all use " passing through network " communication plan to operate, and this scheme relates to the agency on the various nodes in the network, and their collection status and event information are also reported other modules.In case the nextport hardware component NextPort of new node is from physically being added to network, the Tivoli service provides the various modules of management platform can more automatically take over and manage those assemblies.Yet, remain a manual procedure in essence from the process of physically hardware of new node being added to network, can be to finish with previously described same way as.

Aspect the physical hardware of management composition computer system, developed the fault of the whole bag of tricks with the nextport hardware component NextPort in the automatic compensation computer network.U.S. Patent No. 5,615,329 have described the typical case of redundant hardware layout, and this layout uses the special-purpose main and secondary computer system that separates to realize teledata volume shadow, under the situation that main computer system breaks down, the secondary computer system takes over main computer system.The problem of the mirror image of these types or volume shadow layout is that costing an arm and a leg and causing waste, particularly secondary computer system of they is idle under the standby mode of the fault of waiting for main computer system.U.S. Patent No. 5,696,895 have described a solution at this problem, in this solution, a series of servers are arranged, each station server all moves their task, but each station server also all is specified in the backup of serving as this server under the situation that a certain station server breaks down in other servers.This layout can make task be carried out to continue on backup server, though performance can descend by two-server.Other examples of such solution comprise Epoch Point of Distribution (POD) server design and USI Complex WebService.Being used to the nextport hardware component NextPort of these services is provided is predefined calculating pod, comprises load balance software, the fault of the nextport hardware component NextPort in all right compensation management group of this software.Even under the situation of using so predefined calculating pod, the physics of such pod is prepared and is installed to the management group to spend a week just can finish.

All these solutions can be managed and balance operating load and get around hardware fault in the management group automatically based on existing hardware calculated capacity; Yet, also do not develop the solution that more hardware resources can be deployed to automatically the management group.If know the potential demand to the more hardware resources in the management group in advance, modal solution is based on comes pre-configured hardware resource to the highest demand of the prediction of the resource of this group.Although this solution can make the management group make corresponding response during peak demand, for the additional hardware resources that satisfies this peak demand distribution can not get abundant use at most of other times.As a result, provide the cost of the trusteeship service of management group to increase owing to the hardware resource utilization of this group is not enough.

A solution to the demand of more trusteeship services is Cable﹠amp; Internet Shock Absorber (ISA) service that Wireless provides.ISA service is distributed to client's static Web content and is positioned at Cable ﹠amp; There are one or more caching servers of point (POP) in each on the Wireless Internet backbone network.Can be directed to caching server and each POP position to exempt this function in the server from the management group that trusteeship service is provided for this client to the request of this static Web content.Yet the buffer memory static Web content is as the abiogenous something of ingredient by the Internet distribution information.A large amount of users being arranged under the situation of given IP Address requests static information, be common in a plurality of positions on the Internet with this information cache.In essence, the ISA service can make the client on one's own initiative at the buffer memory that starts static Web content on the Internet.Improve the potential of performance that static Web content is provided although this solution has, this solution is not suitable for the trusteeship service of a lot of other types that relate to interactive mode or multidate information content.

Though the mode to management HSP has been made effective enhancing, though developed many programs and instrument helping the operation of HSP network, the basic fundamental that HSP is used for creating with the physical resource of maintenance server field not have the too big variation of generation.Therefore the mode of a kind of more effective operation HSP need be provided, and this mode can be improved the mode of the physical resource of management server field.

Summary of the invention

The present invention is the method and system that is used for runing by this way the managed service provider of the Internet, so that stride the dynamic management that different clients' accounts and/or diverse geographic location provide trusteeship service.For each clients' accounts in many clients' accounts, many individual servers are assigned to the public management group into this clients' accounts definition.Each management group all is configured to visit peculiar software of this clients' accounts and data, so that provide trusteeship service for this clients' accounts provides the Internet.The performance and the operation conditions of the server in each management group of automatic monitoring.For responding automatic supervision, at least one station server in the first management group is re-assigned to second management group automatically and dynamically.Redistributing automatically and dynamically of server is that mode by such is finished: the newly assigned server of attaching most importance to is provided with the initialization pointer with peculiar software of the clients' accounts of visiting the second management group and data, reinitializes the server of redistributing then so that they join the second management group when restarting.Under the preferable case, the independent out-of-band communication channel of Connection Service device monitors with striding the management group by being used for specially for the performance of the server in each management group and operation conditions.Each management group all comprises local decision-making software program, this program communicates with main decision-making software program, and main decision-making software program determines when and how for response user demand, available resource with dynamically redistribute server to different management work groups with the service-level agreement of each clients' accounts signing.

In one embodiment, be used to a plurality of clients' accounts to provide the system of the dynamic management of trusteeship service to comprise five servers that in operation, are connected to an Intranet at least.Each station server all comprises the Host Administration circuit, and this circuit provides in other servers outside the Intranet at least one at least for communication channel.Have at least four station servers to carry out local decision-making software program, this sequential monitoring server is also striden communication channel and is passed on status information.Have at least two-server to be assigned to the first management group of first clients' accounts, and be configured to visit peculiar software of this first clients' accounts and data, so that provide trusteeship service for this clients' accounts by the Internet; Have at least two-server to be assigned to the second management group of second clients' accounts in other servers, and be configured to visit peculiar software of this second clients' accounts and data, so that provide trusteeship service for this clients' accounts by the Internet.Under the preferable case, have at least a station server to carry out main decision-making software program, this program is from other server collect status information, and responsive state information dynamically is re-assigned to the second management group with at least one station server in the first management group at least.

With only limit in the environment of single clients' accounts, to work or require large-scale and expensive department of computer science unify the existing load balance system of common operation system or shared data space different be, the present invention can dynamically extend across a plurality of different clients' accounts and redistribute server, so that with both economical and flexible services device field layout provides trusteeship service.The ability of a plurality of management groups of a plurality of clients of support of the present invention can carry out intelligence between different clients' accounts and dynamic server resource distributes.

Description of drawings

Fig. 1 is the simplified block diagram of prior art layout of the server farm of managed service provider.

Fig. 2 is the server capacity of prior art server farm of a plurality of clients' accounts of trustship and the diagrammatic representation of the relation between the Internet traffic.

Fig. 3 is the simplified block diagram according to the layout of server farm of the present invention.

Fig. 4 is the simplified block diagram that is similar to Fig. 3, has shown server dynamically is re-assigned to second clients' accounts to solve hardware fault from first clients' accounts.

Fig. 5 is the simplified block diagram that is similar to Fig. 3, has shown server dynamically is re-assigned to second clients' accounts to solve the situation that user demand increases from first clients' accounts.

Fig. 6 is the block scheme according to the first-selected embodiment of the assembly of server farm of the present invention.

Fig. 7 is a decomposition diagram according to the first-selected embodiment of the hardware of server farm of the present invention.

Fig. 8 shows the block scheme of the present invention for the level type relation of each software layer of given clients' accounts utilization.

Fig. 9 is the block scheme of striding the embodiments of the invention of realizing in different geographic position.

Figure 10 is the diagrammatic representation of the server capacity and the relation between the Internet traffic of server farm of the present invention when a plurality of clients' accounts of trustship.

Figure 11 is the block scheme that shows the first-selected embodiment of main decision-making software program of the present invention.

Figure 12 is the diagrammatic representation of the different service-level agreement layout of three of a given clients' accounts.

Figure 13 is the diagrammatic representation of server capacity and the relation between the Internet traffic of multi-site embodiment of the present invention.

Figure 14 is the block scheme of leading the decision-making software program that shows that the Control Network switch is connected with storage unit.

Figure 15 is the block scheme that shows the first-selected embodiment of local decision-making software program.

Figure 16 is the diagrammatic representation that the operating load in each measurement module of local decision-making software program under different loading conditions is measured.

Figure 17 is that the request of local decision-making software program is added server or the diagrammatic representation of the decision surface that the deletion server generates from the management group in the management group.

Embodiment

With reference to figure 1, the figure illustrates a simplification function view of the existing server farm 20 of a managed service provider.Such server farm typically uses ready-made hardware and software component structure, and these assemblies are configured to support the trusteeship service requirement of given clients' accounts statically.In this embodiment, the server farm 20 of managed service provider is supported the trusteeship service of four different clients' accounts.Server farm 20 is connected to the Internet 22 by the network switch/router two 4.The network switch 24 is connected to internal network switch/router 26 again, and they have constituted Intranet between the front end/content server 28 of given clients' accounts and rear end/calculation server 30.All front end/content servers 28 and rear end/calculation server 30 all are connected to the disk system 32 that comprises peculiar data of this clients' accounts and software.The physical property that depends on the hardware of server 28,30, disk system 32 can be positioned at the server shell, and perhaps disk system 32 can be arranged in the physically separated unit that is directly connected to each station server 28,30 or is attached to multiple servers 28,30 as memory attaching network (SAN) or network attached storage (NAS) configuration.

Although this layout is used the server farm 20 that ready-made hardware makes up can provide trusteeship service for a plurality of independently clients' accounts well, a plurality of serious problems have been exposed in such layout.The most serious problem is to distribute between different clients' accounts and the character of the general static state of deployment system resource in these problems.For configuration in this complex with manage single clients' accounts, the keeper of HSP need based on the expection requirement of this client's demand will such as the system resource server, disk, the network linking some fixedly rank be exclusively used in specific clients' accounts.

For example, be assumed to be that any given clients' accounts has designed simple relatively website so that under the peak load of expection, clients' accounts may require three front-end servers 28 with the request of process user and four-processor back-end server 30 to handle data base querying/renewal that these requests generate.For such website, can use hardware based technology, as F5 Big-IP, Cisco Local Director or Foundry ServerIron, or use solution based on software, stride front end/content server 28 user's request that balancedly distributes as the software of Windows load balance service (WLBS) or equivalence.In addition, back-end data base/calculation server 30 forms usually and troops that other is fault-tolerant so that some level to be provided.Many software products are arranged, and as MicrosoftCluster Server, Oracle Parallel Server etc., they can make the website with multiple servers overcome contingent hardware fault during normal running.In addition, such as TivoliEnterprise, HP OpenView, or the like the system monitoring instrument can work as when in server farm 20, detecting fault and give notice to the keeper.Though these instruments go for managing trusteeship service in the single clients' accounts of given website, neither one can be used to stride different clients' accounts management trusteeship services in these instruments.

Under this routine situation, the website of supposing this clients' accounts is the e-commerce site for the peak load design of handling 5000 affairs of per minute.In addition, the website of also supposing all the other clients' accounts in the server farm 20 is for the peak load design of handling 10,000,15,000 and 5000 affairs of per minute respectively.As shown in Figure 2, each clients' accounts must be designed and dispose and capacity serious waste in the whole server farm 20 may be caused with the peak load of handling expection.May have extra aggregate capacity although handle the server farm 20 of a plurality of clients' accounts, this extra capacity among account can't be used to respond hardware fault or the unexpected situation about increasing of peak load among another account.For the resource of particular customer account configuration is exclusively used in this account and only is exclusively used in this account.Wherein front-end server 28 at first clients' accounts takes place under the situation of hardware fault, and Web communication will be routed to remaining front-end server 28.If remain unchanged after clients' accounts busy and Web communication before hardware fault takes place is being broken down or increase, so remaining front-end server 28 will owing to service before them operating load and from the redirected additional communication of failed server and very fast quilt overloads.Under the situation of the best, the system management software that is used for server farm 20 will find that server breaks down, and send first message to the website keeper by pager and/or Email, point out that fault has taken place server.If the webmaster web receives message in time and is positioned on this website, the webmaster web can be from physically removing the nextport hardware component NextPort that breaks down so, stock's standby nextport hardware component NextPort for this purpose is installed, reconnect new nextport hardware component NextPort, be this clients' accounts configuration and installation appropriate software, and make new nextport hardware component NextPort join remaining front-end server 28 again.This process can be hopeful to finish in one hour time in less than.If message can not receive in time, if the webmaster web is not positioned at the website at server farm place, if perhaps there is not stock's standby hardware to can be used to replace trouble unit, this process even will spend the longer time so.During this period, descend for response time of access customer account's user, during this period of time, another hardware fault takes place in clients' accounts more and more easily.

Under the demand of clients' accounts being increased and surpass situation for the peak demand of the expection of this clients' accounts configuration, the Web that load balance equipment does not have resource to can be used to redistribute this increase communicates by letter.All servers 28,30 will move under peaking capacity.The result reduces response time of clients' accounts significantly and may occur the request that can't handle in time " serve unavailable " responded.Although can't provide service to the consumer in time is that the people is undesirable, but it is perhaps manageable, the problem of the enterprise in other environment, the additional problem that generates " serving unavailable " message for the website is, no matter if what the former thereby lasting existence of such message, the Internet may begin this information is propagated into a lot of intermediate nodes in the network.As a result, understand website " unavailable " and will be forwarded to other websites to the request of back owing to these intermediate nodes.In case clients' accounts is saturated or overload, the consumer who not only receives " serving unavailable " message can not get service, and many other consumers also possibly can't visit this website.

With reference now to Fig. 3,, uses description to provide the server farm 40 of the dynamic management of trusteeship service below to a plurality of clients' accounts.The same with existing server farm 20, server farm 40 comprises the network switch 44 so that set up interconnection between server farm 40 and the Internet 22.Yet different with existing server farm 20 is that the member of server 46 manages under the control of engine group manager 48.Each station server 46 all is connected to the Internet by the network switch 44 with programming mode and is connected to the stateless computing equipment of disk storage system 50.In one embodiment, server 46 is connected to disk storage system 50 by optical-fibre channel storage area network (SAN).Perhaps, server 46 can be connected to disk storage system 50 by network attached storage (NAS) layout, switchable layout in length and breadth or any similar interconnection technique.

Shown in Figure 4 and 5, engine group manager 48 is responsible for distributing automatically stateless server 46, these servers of account configuration for distributing then between a plurality of clients' accounts.This carries out in the following way: with the server-assignment of given clients' accounts to being this clients' accounts public management group 52 definition and that be configured to visit peculiar software of this clients' accounts and data.As what hereinafter will describe, engine group manager 48 monitors each management group automatically, and response automatically monitors the server 46 among the first management group 52-a ' be re-assigned to automatically and the dynamically second management group 52-b.This is to finish by such mode: use engine group manager 48 to be the server 46 redistributed from the first management group 52-a ' initialization pointer is set to visit peculiar software of clients' accounts and the data of the second management group 52-b, reinitialize the server 46 redistributed ' so that the server of redistributing 46 ' the join second management group 52-b then.Be used for to server farm 20 add hardware resources or therefrom remove the existing process of hardware resource different be that the present invention can make the server 46 redistributed ' can use new management group 52 in a few minutes.Basically, making the server 46 redistributed ' get back to needed unique effective time of on line state will be that to restart the needed institute of server that server 46 ' needed time and load balance and/or the software of trooping identification restarts free.Be appreciated that load balance software is more common on front end/content server, and troop software or troop software and load balance software be combined in rear end/calculation server use more commonly used.To use term " load balance software " to refer to any combination in these possible combinations.

In one embodiment, the server of redistributing 46 ' join the automatically second management group, because the software of the second management group 52-b comprises load balance software, the server that this software responses is got back on line state (promptly reset and energized) or got back to off-line state (promptly resetting and powered-down) adds server to the management group automatically or therefrom removes server.As previously mentioned, this load balance software today widely people know and be extensive use of; Yet existing load balance software is merely able to add server or therefrom remove server to single management group.In this embodiment, engine group manager 48 utilize the function of the load balance application software that can buy on the current market in case stride different management groups 52 dynamically redistribute server 46 '.Perhaps, in case the server 46 that agency in the operating system software of single management group or subroutine can be responsible for redistributing ' get back to server 46 ' integrated second management group 52-b that enters into that on line state is about to redistribute.In another embodiment, engine group manager 48 can be published to renewal in the available server tabulation of each management group 52.

Under the preferable case, engine group manager 48 is set to the suitable copy that starts reflection software and configuration file with the pointer in each station server 46 of management group 52, is included as operating system and application program that this management group 52 is set up.When the server of redistributing 46 ' when quilt restarts, its pointer is reset pointing to startup reflection software and the configuration file of the second management group 52-b by engine group manager 48, rather than the startup of first management group 52-a reflection software and configuration file.

In general, each management group 52 is all represented website or the similarly trusteeship service that the server farm 40 of specific clients' accounts provides.Though different clients' accounts can be by same enterprise or relevant commercial entity pays expense, be appreciated that relatedly with given clients' accounts, therefore organizing 52 related data and softwares with given management will be that this clients' accounts is peculiar.With utilize large computer system to a plurality of clients trusteeship service is provided and use single common operation system realize the service provider of timesharing of resource of large computer system different be, each management group 52 all comprises exclusive software, comprises the traditional operating system software that does not expand to outside the server 46 that is assigned to management group 52.Distributed method of the present invention allows to use the better simply traditional software application and the operating system that can be installed on relatively cheap individual server.So, the individual element of forming management group 52 can be made up of the relatively cheap hardware server that can buy on the market and standard software program.

Fig. 6 and 7 has shown the first-selected embodiment according to the assembly of server farm 40 of the present invention and hardware.Though come first-selected embodiment of the present invention is described with regard to this hardware, but be appreciated that, notion of the present invention is equally applicable to use the server farm of all traditional servers realizations, comprise current 1U or 2U packaged type server, if those servers have Host Administration circuit or its equivalent, and are as mentioned below.

Under the preferable case, the hardware of server farm 40 is extendible engines 100, and it is made up of the server board 102 that can buy on a large amount of market, and each server board is all as the engine blade 132 that is positioned at the rack 110 of power saving and compression.Engine blade 132 vertically is positioned at the front 112 of rack 110, and can dismantle.The plane 130 with holes, centre that is positioned at rack 110 provides utility power and controls peripheral signal to all engine blades 132.The I/O signal of each engine blade 132 is connected to the interface card 134 that is positioned at rack 110 rear portions by the hole in the plane 130 with holes.The I/O signal will be connected to the Internet 22 or disk storage 50 by suitable interface card 134 and by the network switch 44.Under the preferable case, these different communication paths use independent interface card 134.

Extendible engine is because general blade cradle structure 103 can be held the dissimilar server board 102 in the same rack 110.The dissimilar mainboard 102 that can buy on the market is installed on the interchangeable blade carrier structure 103, and this carrier structure 103 provides the identical mechanical interface of rack 110.The custom-designed host pci plate that can be inserted into various types of mainboards 102 can be connected to interface card 134 by plane 130 with holes.The multiple power source signal that redundant heat interchange high-efficiency power 144 is connected on the plane 130 with holes.Motherboard 104 comprises management circuit, and this circuit is assigned to power supply signal by simulation ATX power management protocol the server board 102 of this engine blade 132.It is following with cooling engine 100 that removable fan disk 140 is installed in engine blade 132.Under the preferable case, rack 110 holds the multirow engine blade 132 in the base plate combination 128, and base plate combination 128 comprises a pair of daughter backplane 129 that is laminated to each other together and is positioned at power frame 146 tops, and power frame 146 has power supply 144.Under the preferable case, rack 110 also will comprise rack-mounted ethernet network switch 44 and 147 and be connected to the memory mapping machine 149 of disc driver 50 by fiber channel network.

Be appreciated that, although be that description is of the present invention like this: single rack 110 comprises engine blade 132 and server board 102, they have constituted various servers 46 with appropriate application program software, these servers are assigned to the first management group 52-a, with the second management group 52-b, each management group all has two engine blades 132 at least, still, server farm 40 also can hold the client's of any amount management group 52, and this will depend on the sum of the server 46 in the server farm 40.Under the preferable case, a plurality of racks 110 can be integrated together the sum with the server 46 that increases given position.Just as will be discussed, a plurality of racks 110 that are positioned at diverse geographic location can also be linked at together ingredient as the individual server field 40 of operation under 48 controls of engine group manager.

In the embodiment of first-selection, the server board 102 of each engine blade 132 can be equipped with the up-to-date processor of Intel, SPARC or PowerPC design, and each processor can be supported the standard operation system environment such as Windows NT, Windows 2000, Linux or Solaris.Each engine blade 132 can hold one or more server board 102, and each server board can be according to current ATX form factor or the uniprocessor or the multiprocessor design of the form factor that in the future industry is can be receptible new.Under the preferable case, communication channel 106 realizes as controller zone network (CAN) bus, and the communication path of this bus and the network switch 44 or storage switch 149 separates.Second wrong backup communication channel 106 ' to guarantee the fault-tolerant and redundant communication path of group manager software 48 also is provided.

In traditional server, pointer and start configuration information be by the hand switch setting on the server board or be hard coded into PROM chipset on the server board, or be stored in the fixed position on the addressable local hard drive of server board.Management circuit on the motherboard 104 has suitable hook and is suspended to server board 102, so that actual pointer and other of providing of Host Administration circuit start configuration information.Engine blade 132 can also comprise local hard drive 107, can be by motherboard 104 these hard disks of visit, so that the information that is stored on the local hard drive 107 can be configured by communication channel 106 by motherboard.In addition, comprise electric power management circuit 108 under motherboard 104 preferable case, this circuit can come the application of power of Control Server plate 102 by simulation ATX power management sequence, thereby makes rack 1 10 use utility power.Under the preferable case, back channel Ethernet switch 147 also allows delivery applications program and data message between the various server board 102 in server farm 40, and needn't transmit these communications by the Internet 22.

In the embodiment of first-selection, each rack 110 all can comprise 32 engine blades 132 at most.In this configuration, the network switch 44 and 147 can comprise two 32 circuit switching ethernet network routers from Foundry.Under the preferable case, the network switch 44 and 147 allows to reconfigure between server 46 and the network switch 44 and 147 so that dynamically adjust by the IP address of change server.With regard to disk storage unit 50, two schemes are arranged.At first, can the form with crossbar switch 149 insert exclusive hardware and software between engine blade 132 and disk storage unit 50, they will take out the details of basic SAN storage hardware configuration.In this case, the link between disk storage unit 50 and each blade 132 will be passed to crossbar switch 149 by one group of software API.Perhaps, can use fibre channel media or the RAID storage box that to have bought on the market between blade 132 and disk storage unit 50, dynamically to connect.In these two schemes, one deck software in the engine group manager 48 is carried out necessary configuration adjustment to the connection between server blade 132 and the network switch 147 and the disk storage unit 50.In another embodiment, the part of server 46 can for good and all be connected to the network switch or disk storage unit reduces the switch cost, if server farm 40 will comprise the quantum of the server 46 that can't be reallocated all the time to one group of clients' accounts of certain portions support.In this case, the quantum of the server 46 of each management group 52 can for good and all be connected to the network switch 149 and the disk storage unit 50 of the association of this management group 52.

Again with reference to Figure 4 and 5, as can be seen, the trusteeship service that provides to a plurality of clients' accounts can dynamically be provided in server farm of the present invention system 40.As can be seen, have at least five station servers 46 in operation, to be connected to Intranet54.Under the preferable case, Intranet be server 46 is connected to constitute on the identical network switch 44 in the Internet 22 or on the server 46 similar network switchs such as the network switch 147 connected to one another, constitute.Each station server 46 has management circuit on motherboard 104, this circuit provides one of other servers 46 that separate with the network switch 44 and/or 147 Intranet 54 that form for communication channel 106 at least.

Have at least four station servers 46 to be configured to carry out local decision-making software program 70, this sequential monitoring server 46 is also striden communication channel 106 and is passed on status information.Have at least two-server to be assigned to the first management group of first clients' accounts in these servers 46, and be configured to visit peculiar software of first clients' accounts and data, to provide trusteeship service for this clients' accounts by the Internet.Have at least other two-server 46 to be assigned to the second management group of second clients' accounts, and be configured to visit peculiar software of second clients' accounts and data, to provide trusteeship service for this clients' accounts by the Internet.Have at least a station server 46 to carry out main decision-making software program 72, this locality decision-making software program 70 collect status information of this program from carrying out at other servers 46.In one embodiment, a pair of server 46 uses fault-tolerant coordination software to connect together to constitute the fault-tolerant/redundant processing platform of main decision-making software program.As mentioned below, main decision-making software program 72 responds the status information of collecting from this locality decision-making software program 70 at least with at least one station server 46 the first management group 52-a ' dynamically the be re-assigned to second management group 52-b.

The server 46 of two management groups 52 can think that any configuration of given clients' accounts appointment arranges.As shown in Figure 3, three station servers 46 of management group 52-b are configured to front-end server, and separate unit server 46 is configured the rear end/calculation server of clients' accounts for this reason.Be the violent increase of the peak period use activity of the clients' accounts that responds the second management group 52-b, main decision-making software program 72 defines and necessary the current server 46 that is used as the server of the first management group 52-a ' redistribute is used for the rear end/calculation server as the second management group 52-b.Below in conjunction with this decision-making is described in the description of the operation of this locality decision-making software program 72 is the first-selected embodiment how to obtain.After the process of just having described, the server 46 that main software program 72 indications of making a strategic decision will be redistributed ' dynamically be re-assigned to the second management group 52-b, as shown in Figure 4.

Though first-selected embodiment of the present invention is by with the server 46 among the first management group 52-a ' be re-assigned to the second management group 52-b to describe, should be appreciated that, the present invention can also provide an available server 46 ' public pond, server in this public pond is current not to be assigned to given management group 52 and can to redistribute, and not necessarily requires they are withdrawn from from work management group 52.For example, can set up server farm 40 and distribute six station servers to give each clients' accounts in four different clients' accounts with 32 station servers 46, one station server 46 is carried out main decision-making software program 72, and remaining initial unappropriated seven station server 46 sets up a pond 56, and can be assigned to be in four management groups 52 of this server farm definition any group.Because the work of distribution server is just dynamically finishing at run duration at server farm 40 in the management group according to the present invention, first-selected embodiment of the present invention uses this pond 56 as a buffering, and there is no need at first the server 46 that will redistribute ', server 46 ' adding management of redistributing was organized for 52 needed times thereby further shorten from its existing management group 52.In one embodiment, pond 56 can have warm server and cold server simultaneously.Warm server is the server 46 that has been configured to specific management group 52, therefore will make this warm server add the management group, there is no need it is restarted.Cold server is the server that does not dispose to specific management group 52, therefore will make cold server join the management group, it must be restarted.

Should also be understood that the new management group of the server 46 redistributed ' can be assigned to singly or as have many servers of redistributing 46 ' group be re-assigned to the second management group 52-b from the first management group 52-a simultaneously.How the network switch 44,147 and storage switch 149 are configured to adapt in such environment of dynamically redistributing, will also be understood that, but if must reduce the dynamic-configuration on the network switch 44,147 and/or the storage switch 149 port quantity or if to do like this be desirable, multiple servers 46 can be used as a group and redistributes together.

An apparent in view advantage of the present invention is that the server among the management group 52-a is reconfigured to the process of the second management group 52-b and will the ingredient of this server as the second management group 52-b could be put into operation then from server 46 ' removing of redistributing all states with the particular customer account association of the first management group.This is very effective security mechanism with regard to a nature is provided, to prevent deliberately or by mistake to visit the data of different clients' accounts.Unless server 46 or 46 ' be the member of given management group 52-a, this server will have no idea to visit data or the information of different management group 52-b.The present invention adopts to utilize shared storage space and/or common operation system to stride that different clients' accounts provides the complexity that must realize in the mainframe servers of trusteeship service or other the bigger server systems and the software security function that has potential problems, but kept the advantage that between clients' accounts, adopts simple physics to isolate common in traditional server farm layout, not only can do like this, but also can be at needs or have an opportunity to use better under the situation of certain hardware and make this hardware obtain reconfiguring automatically and dynamically.Unique accessing points of this mandate that reconfigures and control is by main decision-making software program 72 and by out-of-band communication channel 106.

As shown in figure 14, under the preferable case, each station server 46 is connected to the Internet 22 with programming mode under main decision-making software program 72 controls.Main decision-making software program 72 also switches the server 46 redistributed ' with the part of the disk storage unit that is connected to the peculiar software of clients' accounts of storing the second management group and data in operation.Under the preferable case,, use the out-of-band communication channel 106 that separates with Intranet 54 by the network switch 44, so that pass on the part of the status information of main decision-making software program 72 utilizations at least owing to consider safety, fault isolation and bandwidth isolation factor.In the embodiment of first-selection, communication channel 106 is serialization controller Local Area Network (CAN) buses of moving with the bandwidth of 1Mb/s in rack 106, and auxiliary backbone network also moves with the 1Mb/s bandwidth between different rack 106.Be appreciated that, have the independent Intranet that uses Internet agreement (IP) agreement to communicate and to be used as communication channel 106, the serial management interface of replacement such as the CAN bus, embodiment although it is so is overproof designs for the rank and the complicacy of the communication channel 106 needed communications that are connected to motherboard 104.Realize communication channel 106 although can be used as the ingredient of Intranet 54, because the reason of safety, fault isolation and bandwidth isolation aspect, such mode is not first-selected.

Fig. 8 has shown the block scheme of level type relation of an embodiment of various data that the present invention utilizes for given clients' accounts and software layer.Customer data and database 60 have constituted the basic layer of this hierarchical structure.Also can come managing customer data 60 with a plurality of examples of striding the storage unit that constitutes storage system 50 by integrated Web data management software layer 62.Troop and/or load balance application software 64 has constituted the software of the website that is considered to the client usually and the top layer of data.Load balance software 66 is combined multiple servers 46 as the ingredient of public management group 52.The a plurality of examples that also have traditional operating system software 68 all have one on each station server 46.Perhaps, load balance software 66 and operating system software 68 ingredient that can be used as the common software routine package in the single management group 52 comes integrated.On traditional operating system software 68, be engine operation software 48 of the present invention, be used to stride a plurality of clients' accounts 52-a and 52-b and come management resource.

In one embodiment of the present of invention as shown in Figure 9, the server 46 that is assigned to the first management group 52-a is positioned at first website 80, and the server that is assigned to the second management group 52-b is positioned at geographical second website 82 distant apart from first website 80 of going up.In this embodiment, this system further comprises the layout that copies to second website 82 to the data of the major general first management group 52-a automatically.In the embodiment of first-selection, use the communication channel 84 separate with the network switch 44 that the data of the disk storage unit 50-a in first website 80 are copied to disk storage unit 50-b in second website 82.This layout has dual purpose.At first, duplicating of data provides redundancy and backup protection, thereby carries out disaster recovery under the situation of first website, 80 generation disasters.Secondly, data in second website 82 are duplicated and also can be made the present invention comprise the server 46 that is arranged in second website 82 in available server pools, and main decision-making software program 72 can satisfy situation to the demand increase of first client's trusteeship service with them by dynamically these servers being re-assigned to the first management group 52-a.

Under the preferable case, the coordination between the main decision-making software program 72 of first website 80 and second website 82 is to finish by the global decisions software routines 86 that the main decision-making software program 72 that uses with each website communicates.This modularization layout can make main decision-making software program 72 lay particular emphasis on the server resource of managing in the given website, and extend such notion: allow each website 80,82 service outside the venue outside global decisions software routines 86 amount of the claim, or provide available service outside the venue, its mode is identical with the mode of the outer server of local decision-making software program 70 amount of the claim, makes server can be used for being re-assigned to main decision-making software program 70 at given website.

Under the preferable case, multi-site embodiment of the present invention utilizes SAN or the NAS storage networking software that can buy on the market to realize two-layer data redundancy and duplicating layer aggregated(particle) structure.As shown in Figure 9, the disk storage unit 50 in first website 80 has been safeguarded the working version 74 of the customer data of first clients' accounts.Redundant data protection use such as data image, book shadow or the RAID data protection is set up the backup version 76 of customer data of first clients' accounts of first website 80.Network software utilizes communication channel 84 to generate second backup version 78 of the customer data of first clients' accounts that is positioned at second website 82.Under the preferable case, use with the network switch 44 to the Internet 22 be connected the communication channel 84 of separating redundant communication path can be provided, and reduce the influence of generation second backup version 78 necessary background communication activities to greatest extent.Perhaps, the backup version 78 of customer data that is positioned at first clients' accounts of second website 82 can carry out route by the network switch 44 and the Internet 22.In another embodiment, the extra backup version of customer data can duplicate at extra site location to further expand the function of system, is not re-assigned to the clients' accounts that needs these resources will fully use the server of the clients' accounts of server resource.

As shown in figure 10, never the server that fully uses the clients' accounts of server resource of the present invention is re-assigned to the ability of the clients' accounts that needs these resources, can make the resource of server farm 40 provide trusteeship service, thereby obtain more effective use to a plurality of clients' accounts.For each clients' accounts 91,92,93,94 and 95, distributed server 46 fully to each clients' accounts, so that safeguard metastable a small amount of overload bandwidth for each clients' accounts.Different with existing server farm is, at present, change to the hardware resource that is assigned to given clients' accounts is that unit carries out with hour, day or week, and the present invention can realize the server resource that dynamically distributes is on demand carried out minute being the change of unit.Figure 10 has also shown and utilizes a plurality of diverse geographic locations to place the advantage of the various piece of server farm 40.As can be seen, because the time difference between the time zone between site location 80 and the site location 82, the time on clients' accounts 94 and 95 use peak is with the asynchronism(-nization) on the use peak of other clients' accounts 91,92 and 93.The present invention can utilize these mistimings of using the peak that the server capacity that rises and falls is assigned to the site location that is in the use peak period from other not frequent site locations of activity.

In an embodiment of as shown in figure 13 multi-site of the present invention configuration, the site location 80,82 and 84 that has three separation at least, under the preferable case, they are positioned at each other removes individual hour position of N+1 apart from being at least 24, and wherein N represents the quantity of the different site location in the multi-site configuration.In the embodiment of the site location 80,82 with three separation and 84, under the preferable case, site location distance each other is eight hours.The time difference that separation on this geographic position realizes allow to be positioned at all three websites clients' accounts the use pattern by each station server at given position can not utilize from one or more other position dynamic compile by the sum of server under the situation of the needed lazy weight of server redistributed and serve.To experience the advantage of doing like this when using rank night more obvious when site location 80, and the server in this site location 80 can dynamically be re-assigned to experience and use other site location 82 of level daytime.Simultaneously, site location 84 is experienced and is used rank at dusk, may be fit to or also may be not suitable for from then on server is re-assigned to another position or carries out opposite operation in the position.In general, site location is to arrange like this: so that at first borrow capacity from the site location that is positioned at later time zone (i.e. the east of this website), and unnecessary capacity can be used to the site location that is positioned at time zone (i.e. the west of this website) early.Set up other preferences according to the use pattern in past and the use pattern of prediction.

With reference now to Figure 11,, will the first-selected embodiment of main decision-making software program 72 be described.Main decision-making software program 72 comprises resource database 150, service-level agreement database 152, main decision logic module 154 and scheduler module 156.Main decision logic module 154 can access resources database 150 and service-level agreement database 152, and the information in status information and resource database 150 and the service-level agreement database 152 is compared to determine whether dynamically server to be re-assigned to second clients' accounts from first clients' accounts.Scheduler module 156 is linked to main decision logic module 154 in operation, so that when main decision logic module 154 is sent indication, dynamically redistribute server, its mode is, uses communication channel 106 to attach most importance to newly assigned server 46 ' be provided with initialization pointer with the peculiar software of clients' accounts and data of visiting the second management group 52-b and reinitialize the server 46 redistributed ' so that at least one station server joins management group 52-b.Under the preferable case, scheduler module 156 comprises one group of concatenate rule 160 and the one group of personality module 162 at each station server 46.Concatenate rule 160 provides the instruction that particular server 46 is connected to the given network switch 44 or data storage cell 50.Personality module 162 has been described the details of specific software configuration of the server board 102 of the management work group that will add clients' accounts to.In case scheduler module 146 is determined to redistribute server, it will assess concatenate rule 160 collection and one group of personality module 162 with definite server 46 that will be dispatched to this cura specialis group 52 that how to make up.

Understanding another mode that how the present invention dynamically to extend across different accounts trusteeship service is provided and be a part of looking server 46 is the pond that is assigned to many virtual servers, can dispose software and the data of these servers with visit cura specialis group 52 selectively.Determine need be when specific management group 52 be added servers 46 when scheduler module 146, and it just is assigned to a station server this management group automatically from the virtual server pond.On the contrary, can abandon one station server 46 if scheduler module is determined a management group, then this server of abandoning will be added to the virtual server pond that can be used for being re-assigned to different management groups.When from then on angle was of the present invention, then as can be seen, 48 operations of group manager software were with the server in this many virtual servers pond in real time or as required strategic point " manufacturing " or create one or many virtual servers.As previously mentioned, the pond of virtual server can be warm pool or cold drop, or their any combination.According to concatenate rule 160 collection and personality module 162 is made or the structure virtual server so that allow the management group utilization of needs.

In this embodiment, main decision logic module 152 is connected to supervisor console 158 in operation, and this control desk can show the information of relevant main decision-making software program, and accepts accounts maintenance and progress information is updated to various databases.Record keeping software module 160 is integrated in the engine group manager 48, so that keep accounts based on the server tracks of distributing to given clients' accounts.Under the preferable case, when the service-level agreement server based on the client dynamically is re-assigned to certain clients' accounts, collect higher trusteeship service rate to this clients' accounts.

Figure 12 has shown the diagrammatic representation of the service-level agreement layout that three of a given clients' accounts are different.In this embodiment, for providing trusteeship service to sign service-level agreement in section preset time such as month.In first rank 170 that shows, clients' accounts has the capacity of the connection support trusteeship service that is 640,000 whiles.If clients' accounts does not need to redistribute server to support will to collect other erected cost of this level of the capacity of promise to this client so greater than the capacity of the capacity of first rank, 170 promises.In the second level 172 that shows, clients' accounts can be dynamically extended into the double capacity of capacity of first rank 172.In the embodiment of first-selection,, engine group manager 48 supports the capacity of second level 172 in case dynamically being re-assigned to clients' accounts with server, to satisfy the demand on the use peak that is higher than expection, will collect higher rate for needing extra that time of using to clients' accounts.In addition, can also collect nonrecurring charge, so that start the higher service class of second level 172 expressions to clients' accounts.In one embodiment, the rate of the expense of second service class 172 is manyfolds of the rate of first rank 170.Second level 172 is illustrated in the guaranteed expansion rank that the client can use in given a period of time.At last, the third level other 174 provides optional extra expansion service rank, can use this rank to provide trusteeship service for clients' accounts.In this embodiment, the third level other 174 provides the service class of the many manyfolds of service class of first rank 170.In one embodiment, for the extra-service rank of this expansion is provided, host computer system uses foregoing multi-site layout so that introduce needed number of servers, so that satisfy this other service of level.Under the preferable case, collect the second high rate of time period of the extra-service of the expansion that is re-assigned to this clients' accounts to clients' accounts.In one embodiment, in the preset time of extra the 3rd service class 174 that expansion is provided to certain clients' accounts section, the rate of the expense of the 3rd service class 174 is manyfolds of first rank 170.In addition, also can collect nonrecurring charge, so that start the 3rd service class 174 in the section at any time in preset time to clients' accounts.When preset time, section finished, the client can revise the service class of given clients' accounts.

As shown in figure 12, the service-level agreement higher expection that responds given clients' accounts uses the peak to increase 50% from very first time section to second time period.Under the preferable case, the time period of the service-level agreement of given clients' accounts will be unit with the moon, can advise to the client, whether be willing to be intended to following accounting segment change service-level agreement.Though this example is to demonstrate with connection simultaneously, should be appreciated that, the service-level agreement of given clients' accounts can be signed with various performance measurements, the resource, the income of generation or their any combination that utilize as simultaneously connection, hit rate, data quantity transmitted, the quantity of affairs, tie-time, different Application Software Program.Be further appreciated that service-level agreement can provide different promise ranks for the different resource type such as front-end server, back-end server, network connection or disk storage unit.

With reference now to Figure 15,, will the block scheme of embodiment of the first-selection of local decision-making software program 70 be described.Each module is all carried out independently assessment to the operation of the particular server that moves local decision-making software program 70 in a series of measurement modules 180,181,182,183 and 184.The output of these measurement modules is provided to the polymerizer module 190 of local decision-making software program 70.Predictor module 192 is response time and the probability that various requests produce expection.Fail 194 by means of the priority that the main decision-making software program 72 in the service-level agreement database 152 provides, fuzzy inference system 196 determines whether to make the request into management group 52 interpolation engine blades 104, or does not make the proposal of abandoning or removing the engine blade from management group 52.The request of adding then or removing blade is communicated to main decision-making software program 72 by communication channel 106.In one embodiment, carry out on polymerizer module 190 each station server 46 in given management group 52, and carry out on 196 individual servers 46 in given management group 52 of predictor module 192 and fuzzy reasoning module, the output of various measurement module 180-184 is passed to specified server 46 by communication channel 106.In another embodiment, polymerizer module 190, predictor module 192 and fuzzy reasoning module 196 can be carried out on the multiple servers in the given management group, produce the request information necessary of adding or removing blade so that reach redundancy or distributed treatment.

Under the preferable case, polymerizer module 190 is finished the balancing run of striding various measurement module 180-184 according to following formula:

B _k＝[(∑T _ki/w _k)-min _k]*100/(max _k-min _k)-50

I=1 is to w _k

T wherein _KiBe i the needed time of request of measuring type k, w _kBe the window size of measuring type k, min _kBe the expection minimum time of measuring type k, and max _kBe for measuring the maximum time of type k tolerance.Then, the request rate B of balance _kBe passed to the fuzzy reasoning module 196 of predictor module 192 and local decision-making software program 70.To window size that measure type k be set reducing the useless invasion of any measurement module 180-184 to greatest extent, and allow to respond in time and suitably increase simultaneously the user demand of management group 52.

Figure 16 has shown the sample that the operating load among each measurement module 180-184 is measured under different loading conditions.The lasting predictable estimation of the probability of response time of expection and this response time is provided without any single operating load measurement as can be seen.So, fuzzy reasoning module 196 must be considered three basic parameters: the response time of the prediction of various requests, the priority of these requests, and their probability of happening.These three Considerations integrate to make definitely with all for fuzzy reasoning module 196, whether ask to add blade or remove blade from management group 52.Be the example of a fuzzy inference rule below:

(if priority is higher) and (probability is bigger) and (response time of expection is too high), (blade that amount of the claim is outer) so.

Under the preferable case, the net result of fuzzy reasoning module 196 will produce decision surface, and this surface is reflected in the probability of this response time of the needs management group therewith 52 of the outer server of amount of the claim in response time of expection.Figure 17 has shown the example of such decision surface.

Declaratives of the present invention are subjected to the protection of the Copyright Law.The copyright owner allows explanation of the present invention is faxed and copied as patent and trade mark office file or record, but other modes will keep all copyrights.

Though described the first-selected embodiment of automatic system of the present invention, be appreciated that and make a lot of changes and variation that scope of the present invention will be defined by claim.

Claims (26)

1. an automated process is used to run the service provider of the Internet so that the dynamic management of trusteeship service is provided, and comprising:

For each clients' accounts in many clients' accounts:

Multiple servers is provided, and these servers are assigned to the public pipe of this clients' accounts

The reason group, and be configured to visit peculiar software of this clients' accounts and data, think

This clients' accounts provides the trusteeship service to the Internet;

Automatically monitor each management group; And

Response monitors automatically, and at least one station server in the first management group is re-assigned to second management group automatically and dynamically, comprising:

For said at least one station server is provided with the initialization pointer to visit second management

The peculiar software of clients' accounts and the data of group; And

Reinitialize said at least one station server so that said at least one clothes

The affair device joins the second management group.

2. method according to claim 1, it is characterized in that, by an Intranet link together, and the step of monitor management group partly is to be finished by the communication channel that is different from Intranet of this management group to the many servers that are assigned to each management group automatically in operation.

3. method according to claim 1, it is characterized in that, the many servers that are assigned to the first management group are positioned at first website, the many servers that are assigned to the second management group are positioned at geographical second website distant apart from first website of going up, and monitoring step further comprises to the data of major general's first management group and copies to second website automatically automatically.

4. method according to claim 1 is characterized in that, the step of dynamically redistributing is to respond the parameter that monitors and be combined in each clients' accounts that defines in the service-level agreement database automatically to carry out.

5. method according to claim 1, it is characterized in that, automatically monitoring step detects the wherein fault of a station server in the second management group, and dynamically distributes a wherein station server in the first management group to replace the failed server in the second management group at least.

6. method according to claim 1, it is characterized in that, automatically monitoring step predicts that the operating load of the server in the second management group increases, and the operating load that dynamically distributes the wherein station server in the first management group to increase with redistribution between the server of the bigger quantity in the second management group at least.

7. method according to claim 1 is characterized in that, the step that the initialization pointer is set utilizes the information of safeguarding in the personality module of each clients' accounts to determine the initialization pointer.

8. method according to claim 1, it is characterized in that, each station server is connected to the Internet with programming mode, and the step of dynamically redistributing comprises that further switching said at least one station server is connected to the Internet with the ingredient as the second management group in operation.

9. method according to claim 8, it is characterized in that, each station server further is connected to a disk storage unit with programming mode, and the step of dynamically redistributing further comprises switches the part of said at least one station server with the disk storage unit of the peculiar software of clients' accounts that is connected to the storage second management group in operation and data.

10. method according to claim 1, it is characterized in that, the step of dynamically redistributing further comprises when said at least one station server dynamically is re-assigned to this clients' accounts, for trusteeship service is collected the charges to clients' accounts with higher rate.

11. one kind is used to the system that dynamic management is provided to the Internet trusteeship service that a plurality of clients' accounts provide, comprises:

Have at least five station servers to be connected to an Intranet in operation, each station server has the Host Administration circuit, and this circuit provides in other servers outside the Intranet at least one for communication channel;

Have at least four station servers to carry out local decision-making software program, this sequential monitoring server is also striden communication channel and is passed on status information;

Have at least two-server to be assigned to the first management group of first clients' accounts, and be configured to visit peculiar software of first clients' accounts and data, think that this clients' accounts provides the trusteeship service to the Internet;

Have at least two-server to be assigned to the second management group of second clients' accounts, and be configured to visit peculiar software of second clients' accounts and data, think that this clients' accounts provides the trusteeship service to the Internet, and

Have at least a station server to carry out main decision-making software program, this program is from other server collect status information, and responsive state information at least one station server in major general's first management group dynamically is re-assigned to the second management group.

12. system according to claim 11, it is characterized in that, main decision-making software program is dynamically redistributed said at least one station server, its mode is, using communication channel is that said at least one station server is provided with the initialization pointer with peculiar software of the clients' accounts of visiting the second management group and data, and reinitializes said at least one station server so that said at least one station server joins the second management group.

13. system according to claim 11, further be included in the network switch that is connected in the operation between the Internet and each station server, it is characterized in that each station server is connected to the Internet with programming mode under the control of main decision-making software program.

14. system according to claim 11, further comprise the disk storage unit that is connected to Servers-all with programming mode, it is characterized in that main decision-making software program switches the part of said at least one station server with the disk storage unit of the peculiar software of clients' accounts that is connected to the storage second management group in operation and data.

15. system according to claim 11, it is characterized in that, the many servers that are assigned to the first management group are positioned at first website, the many servers that are assigned to the second management group are positioned at geographical second website distant apart from first website of going up, and this system further comprises the device that is used for copying to automatically to the data of major general's first management group second website.

16. system according to claim 11 is characterized in that, main decision-making software comprises:

Resource database;

The service-level agreement database;

Main decision logic module, can access resources database and service-level agreement database, and the information in status information and resource database and the service-level agreement database compared, to determine whether that said at least one station server dynamically is re-assigned to second clients' accounts from first clients' accounts; And

In operation, be linked to the scheduler module of main decision logic module, when sending indication, main decision logic module dynamically redistributes said at least one station server, its mode is, using communication channel is that said at least one station server is provided with the initialization pointer with peculiar software of the clients' accounts of visiting the second management group and data, and reinitializes said at least one station server so that said at least one station server joins the second management group.

17. system according to claim 16 is characterized in that, scheduler module further comprises one group of concatenate rule and the one group of personality module at each clients' accounts.

18. system according to claim 11 is characterized in that local decision-making software program comprises many measurement modules with output, these outputs are pooled in the fallout predictor routine, with the response times and the probability of the expection of determining this server.

19. system according to claim 18, it is characterized in that, this locality decision-making software program of given server further comprises the fuzzy logic inference system of the output that is connected to the fallout predictor routine at least, to start the request of adding or deleting server from the management group related with this server.

20. system according to claim 19, it is characterized in that, the balance of main decision-making software program is from interpolation or the request of deletion server and the information in resource database and the service-level agreement database of all local decision-making software programs, to determine whether that the said at least one station server in the first management group dynamically is re-assigned to the second management group.

21. an automated process is used to run the service provider of the Internet so that the dynamic management of trusteeship service is provided, and comprising:

For each clients' accounts in many clients' accounts:

Many servers are provided, and these servers are assigned to the public pipe of this clients' accounts

The reason group, and be configured to visit peculiar software of this clients' accounts and data, think this

Clients' accounts provides the trusteeship service to the Internet;

Set up a pond of forming by many virtual servers, so that can optionally be configured to visit the software and the data of each clients' accounts in many clients' accounts;

Automatically monitor each management group;

Response monitors that automatically at least one virtual server of moving distribution comprises to add many servers of the first management group:

For said at least one virtual server is provided with the initialization pointer to visit second

Peculiar software of the clients' accounts of management group and data; And

Reinitialize said at least one virtual server so that said at least one

Station server joins the first management group.

22. method according to claim 21 is characterized in that, response monitors automatically, creates at least a portion in the pond of being made up of many virtual servers on demand.

23. method according to claim 22 comprises that further response automatically monitors the station server in the many servers that discharge the second management group at least automatically and distributes this described at least one station server to the pond of virtual server.

24. method according to claim 21 is characterized in that, the step that the initialization pointer is set utilizes each management to organize the information of safeguarding in the peculiar personality module.

25. method according to claim 21, it is characterized in that software and data that the step that the initialization pointer is set can prevent the virtual server visit to be associated with any clients' accounts outside those clients' accounts related with the management group of it having been distributed virtual server.

26. method according to claim 21 is characterized in that a plurality of virtual servers are assigned to the first management group simultaneously.

Images