Background technology
Appearing at of commercialization ISP (ISP) promoted the fast development of internet as computer network to a great extent.ISP is a user Internet access, inserts international telephone network as telephone operator for the user.The mode that most of ISP collect the charges with regard to this access service is similar to the mode that the user collects telephone expenses to telephone operator.Originally, ISP charges to the user by the surf time usually, as telephone operator collects long-distance telephone expenses.Now, most of ISP take the mode of fixing month access fee, and this is similar with the mode that telephone operator collects city telephone rate.But in fact, all these expenses are meterage usefulness, are according to collecting expenses of surfing Internet preset time, i.e. how many branches of per minute or every month how many units.
Many reasons cause the meterage mode of ISP and telephone operator that similarity is arranged.Computer internet utilizes identical high speed, the long Distance Transmission voice-and-data of high bandwidth communication channel main line business with international telephone network.Local telephone network is also passed through in data service between dial-up user and the ISP greatly.Subunit or branch offices that many large-scale ISP are telephone operators.As telephone operator, ISP can be subjected to government control as public network operator or government utility.Perhaps the most important thing is, have only very few company to provide the required backbone network of ISP to connect, and all these companies take the metering and billing mode to collect operating charges.It is important composition in the typical ISP expense that backbone network connects expense, and for non-North America ISP, can constitute its most expenses.About being connected with network, phone how to realize that metering and billing then is that many United States Patent (USP)s (as are numbered 3,764,747,5,187,710,5,303,297,5,351,286,5,745,884,5,828,737, theme 5,946,670,5956,391 and 5,956,697 patent).
Concerning ISP, there are many charging package softwares to can be used for registering the meterage accounts, such asRens.comXaCCT andInovaware.comISP Power.Other assists the software program of ISP network management also to develop, asLightspeedsystems.comIP Magic,Resonate.comInternet Service Management andLuminate.comMAMBA.The management of ISP and operation also become numerous articles and forum's main topic of discussion, the speech of delivering in the interworking between network forum that held on April 19th, 1999 such as Hursti and Jani that is entitled as " online and Service Management ".ByRackspace.comThe typical ISP that the website can be understood with the charge of particular month rate provides which software and hardware configuration, maintenance and support for providing business level online and website to manage.
Geoff Huston has gone through several big key element of ISP pricing strategy in " ISP Survival Guide: competitive ISP management tactics " the 13rd chapter 497-535 page or leaf of publishing in 1999.He finds to have comprised in the retail rate that ISP collects five big characteristics of ISP access service, i.e. access, time, capacity, distance and quality.If service fee greater than running cost, typically inserts price strategy is to adopt a month nontraffic sensitive, because this is convenient to carry out the characteristics that have simplicity, expandability and create competitive environment for the service provider.If based on running cost, take moon nontraffic sensitive then may bring unacceptable commercial risks, and certain rate structure that increases progressively based on the meterage of tight monitoring may be more desirable.Although Mr. Huston estimates, the ISP industry is along with being dominated and will constantly stablize and consolidate by large-scale operator, and he thinks that stablizing of Internet market constantly grows up strong and sturdy continuing to be subjected to and the influence of the service mode that develops.
One of differentiation is the minority picture to have occurredNetzero.comWithFreeInet.comSuch ISP, they provide free service to each terminal user.The management mode of these ISP is to appear at the advertising income that the headline advertisement on the user display screen produces after relying on user's access service always, rather than collects access fee and use.It is similar to the management mode of commercial broadcasting TV in many aspects, and under this management mode, advertising income has been born service fee.
Another product of traditional ISP service is to have developed application system provider (ASP), asApplicast.comWithUsi.net, and enhancement mode or enterprise solution provider (ESP), as cwusa.com and hostpro.net.Although to the not clearly definition of service that ASP and ESP provided, the host services office pattern similarity that its management mode and Electronic Data Systems etc. are carried out stipulates that a part of Computer Processing demand of enterprise is contracted out to the third party.ASP and ESP provide services such as relevant application program operation, Website development, managing eBusiness and server allotment at some of client, most of or whole demand, carry out charge regularly.With regard to server allotment, run application according to the client traditionally or the website clear and definite specific software and hardware configuration charge.The same with traditional ISP, software and hardware configuration is high more, and the support service that provides is many more, and every month expense is also high more.
At present the relevant internet of great majority charges and the patents of ISP all are absorbed in ISP is involved ecommerce travelling trader and buyer, i.e. dealing between the user of ISP is paid the bill in the chain, guarantees the safety of online transaction.Be numbered 5,794,221,5,845,267 and 5,899, in 980 the United States Patent (USP), enumerating this type of safety payment system example that involves ISP.Although these payment systems can be used in limited place, people the credit card information that utilizes the security server link to provide have been provided been have have been bought and sold on the net, and this has replaced many demands to these payment systems.
Be numbered 5,819,092 United States Patent (USP) has been introduced a kind of instrument that develops software with expense set-up function on the net, make Website development merchant etc. can set up the online service charging structure, under this charging structure, charge to user and third party according to login online service, search or download of information.Be numbered 6,035,281 United States Patent (USP) and introduced a kind of charge system in many ways that the internet inserts that is applied to, its principle is to distribute corresponding cost according to the bandwidth of using in the function of predetermined accessed content and the access process to each side concerned.Though still some Internet access service is chargeed by access times, the demand of this instrument that accounts is not existed substantially, because being most message references, development trend can be undertaken by the internet that no this item is charged.
The european patent application that is numbered 0844577A3 has been introduced a kind of multistage marketing web server computer, whenever finishes a transaction on the server, and promptly " participant " to this multistage marketing program pays because of the deserved multistage marketing commission of marketing activity.Though the purposes of the webserver has been introduced in this application, its emphasis is not a How It Works of introducing ISP, but traditional multistage marketing arrangement realizes robotization, promptly pays the commission of every transaction to the in-house a series of individuals of multistage marketing.
Although the way to manage of ISP has had large increase and improvement, and developed many programs and instrument and assisted the ISP network operation, become the commercial network of dominance since the internet after, the service charge method of ISP does not have radical change.
Embodiment
Fig. 1 is the simplification functional diagram of the existing group ofserver 20 of server service provider, and this type of group of server is supported the host services requirement of a certain user account usually by existing software and hardware structure by static configuration.In the present embodiment, the group ofserver 20 of server service provider can be supported four different user accounts' host services.Group ofserver 20 is connected withinternet 22 by the network switch/router two 4, thenetwork switch 24 is connected with internal network switch/router 26 successively, form the Intranet of a certain user account in front end/content server 28 and rear end/calculation server 30, the data that all front end/content servers 28 and rear end/calculation server 30 are all exclusive with comprising this user account and the disk system 32 of software link to each other.Ardware feature according toserver 28 and 30, disk system 32 can be installed in the server shell, also can be installed in directly with eachserver 28 and autonomous device that 30 are connected in, also can be set to storage network (SAN) or network attached storage device (NAS) and be connected with 30 with a plurality ofservers 28.
Though this layout can make full use of ready-made hardware construction group ofserver 20, for multi-user account independently provides host services, but it has exposed several significant problems, and wherein greatest problem is that system resource allocation and scheduling between the different user account generally are static.For the unique user account is configured and manages, the keeper of server service provider needs carry out requirement forecast according to this user's request, is in view of the above on the system resource of this user account distribution fixed amount (be connected as server, disk and network etc.) again.
For example, be assumed to be a certain user account and designed a fairly simple website, when reaching expectation peak visit capacity, user account can require three front-end servers 28 simultaneously user's request to be handled, and the data base querying/renewal that requires 30 pairs of user's requests of a quaternary processor back-end server to be produced is handled.At this website, may utilize hardware technology such as F5 Big-IP, Cisco LocalDirector or Foundry ServerIron or form load balancing service software solutions such as (WLBS) that user's request is averaged distribution in front end/content server 28.In addition, also back-end data base/calculation server can be gathered to reach certain tolerance usually.The Cluster Server of Microsoft and many software products such as parallel server of Oracle make the website that has a plurality of servers can overcome contingent hardware fault in the normal running.In addition, system monitoring instruments such as Tivoli Enterprise, HP OpenView can be notified the keeper when group ofserver 20 breaks down detecting.Though these instruments are enough to the unique user account of certain website is carried out the host services management, all can not carry out the host services management to the different user account.
Prolong and use example, the website of supposing this user account is an e-commerce website, and its peak load is designed to per minute and handles 5000 transactions.The peak load design of supposing other user account websites in the group ofserver 20 again is respectively per minute 10000,15000 and 5000 transactions.As shown in Figure 2, each user account is planned and is disposed the serious waste that may cause whole group ofserver 20 capacity for the peak load of dealing with expection.Even handle multi-user account's group ofserver 20 unnecessary aggregate capacity is arranged, increase but this unnecessary capacity can not be used for the accident of processing hardware fault or account's peak load.For the resource of certain user account configuration is used for and can only be used for this account.When one of them front-end server generation hardware fault of first user account, network traffic will be directed to other front-end server 28.The back network traffic takes place remain unchanged or increase if the just busy and fault of user account takes place before the hardware fault, then remaining front-end server 28 is because of handling work before it in addition from the extra traffic overload rapidly of failed server.Under optimal situation, the system management software of group ofserver 20 can find that server breaks down, and gives the webmaster by pager and/or Email send server failure message.If the webmaster in time receives information and be in the scene, then dismountable faulty hardware is installed the standby hardware among the stock, and line again again for this user account setting with suitable software is installed, and is incorporated new hardware into remaining front-end server 28 again.If All going well, whole process can be finished in one hour.But, if failure message fails in time to receive that the webmaster is or not failed server group location, or not being used for replacing the standby hardware of trouble unit in the stock, this process just needs the longer time.Simultaneously, during this period, the response time of user's login account can prolong, and user account more is easy to generate other hardware faults.
When the demand of receiving when user account surpasses the expectation a maximum demand that sets, then do not have resource to can be used for load balancing, redistribute the network traffic of increase.Theserver 28 that this moment is all and 30 is with the full load operation, and consequently the user account response speed reduces greatly, and may feed back the information of " serving unavailable " to the request that can not in time handle.Though can not be problem, still a may command for enterprise for the user provides timely service.Occurring " serving unavailable " this problem of information on the website then more bothers, if no matter what reason of this information occurs always, the internet may pass to this information the numerous intermediate nodes in the network, and these intermediate nodes can think website " unavailable " and the subsequent request that will mail to this website turns to.So, the user who not only receives " serving unavailable " information can not get service, and when this user account in a single day saturated or transship after may cause many other users can not visit this website.
Our Fig. 3 now, wherein group ofserver 40 provides the host services dynamic management for multi-user account.Identical with existing group ofserver 20 is, group ofserver 40 is connected withinternet 22 by thenetwork switch 44, but different be that group ofserver 40 is to be managed by group of server manager 48.Everystation server 46 is the statelessness computing equipment, connects the internet by programming by thenetwork switch 44, and is connected with disk storage system 50.In one embodiment,server 46 is connected withdisk storage system 50 by fiber channel SAN.Perhaps,server 46 can arrange by NAS, convertible arrange in length and breadth or any similar interconnection technique is connected withdisk storage system 50.
As shown in Figure 4 and Figure 5, group ofserver manager 48 is responsible for distributing automaticallystatelessness server 46 and is that the account who distributes carries out server configures in multi-user account.This process is by finishing the server-assignment of a certain user account to thepublic management group 52 that aims at this user account definition and be configured to visit exclusive software of this user account and data.As described later, the group of server manager monitor each management group automatically and according to automatic monitored results automatically withserver 46 from the management group 52-b of first management group 52-a dynamic assignment to the second again.This is that the initialization pointer of the server 46 ' redistributed from first management group 52-a by group ofserver manager 48 setting is gone up the exclusive software and the data of this user account to visit second management group 52-b, and then the newly assigned server 46 ' of counterweight reinitializes and it added second management group 52-b finish.The present invention is different with existing group ofserver 20 hardware resources interpolation or delete procedure, and it can in minutes make the server of redistributing 46 ' bring into use in new management group 52.Basically, with the server redistributed 46 ' online consuming time maximum be exactly to restart the server required time thatserver 46 required times and load balancing and/or the software of trooping identification restart.You can find that load balancing software is used for and being connected of front end/content server more, and troop software or troop software and the integrated software of load balancing software then are used for and being connected of rear end/calculation server more.This term of load balancing software is used in reference to acts as for what this type of integrated software.
In one embodiment, because the software of second management group 52-b comprises load balancing software, it can be on line machine (as reset, start shooting) or off-line machine (as reset, shut down) state according to server is that the management group is added or the deletion server automatically, and the server of therefore redistributing 46 ' can add in second management group automatically.The front was carried, the nowadays well-known and widespread of this load balancing software, but existing load balancing software only can add or delete server in single management group.In the present embodiment, group ofserver manager 48 utilizes the function of the balanced application software of present commercial load indifferent management groups 52server 46 to be carried out dynamic reallocation.Another kind method is, in case the server redistributed 46 ' online after, the Agent of operating system software or subroutine can be incorporated into the server of redistributing 46 ' in second management group in the single management group.In another embodiment, group ofserver manager 48 can be announced the renewal tabulation of available server in eachmanagement group 52.
Preferable is that the pointer of everystation server 46 of group ofserver manager 48management groups 52 is set to the suitable copy that 52 mounted navigational figure softwares and configuration file (comprising operating system and application program) are organized in this management.When the server of redistributing 46 ' restarts, its pointer just serviced device group manager be re-set as start software and the configuration file that points to second management group 52-b, rather than point to navigational figure software and the configuration file of first management group 52-a.
Generally speaking, the website that provided by the group ofserver 40 at a certain particular user account place of eachmanagement group 52 representatives or similarly host services.Although same company or relative commercial entity can be paid different user accounts, and be relevant with certain specific user, thereby also relevant with certain curaspecialis group 52 data and software are exclusive by this user account.Thereby provide the service provider of host services different by the timesharing of the large-scale mainframe computer system resource of single public operation system implementation as the multi-user with using large-scale principal computer, eachmanagement group 52 has its distinctive software (comprising the routine operation system software), can not exceed the scope of theserver 46 of having distributed to second management group 52.This location mode of the present invention makes people can install and use simple conventional application software and operating system on the individual server of cheapness.Like this, the single component of formingmanagement group 52 just can select for use market that the hardware server and the standard software program of the cheapness of sale are arranged.
Fig. 6 and Fig. 7 are for setting up a preferred embodiment of group ofserver 40 required parts and hardware according to the present invention.Though when introducing the preferred embodiment of the present invention, should be mentioned that this hardware, but need understand that conception of the present invention is equally applicable to use all General Servers group of server of (1U or the complete server of 2U that provide at present are provided), as long as above server has Host Administration circuit or the similar device that hereinafter will introduce.
Preferablely be, theheight expansion superserver 100 that the hardware of group ofserver 40 is made up of massmarket server board 102 on sale, every block of plate is installed in power saving, the space-efficient rack 110 as superserver blade 132.Superserverblade 132 vertically is installed in the front side 112 of rack 110, and detachable.In the centre of rack 110 a break-through plate 130 is arranged, utility power is provided and controls the peripheral signal that is delivered on all superserver blades 132.The input/output signal of everysuperserver blade 132 is passed to theinterface card 134 that is installed in rack 110 back by the hole on the break-through plate 130.Input/output signal will be sent tointernet 22 ormagnetic disk memory 50 through thenetwork switch 44 by the appropriate interface card.Preferable is that different communication paths uses independent interface card.
Because public blade mainboard carrier 103 is arranged, scalable superserver can hold dissimilar server board in same rack 110.Marketdissimilar mainboards 102 on sale are installed on the utility blade mainboard carrier 103, and this carrier provides unified mechanical interface for rack 110.Be connected withinterface card 134 by break-through plate 130 through can the peg grafthost pci plate 104 ofpolytype mainboard 102 of special design.Redundant heat interchange efficient power 144 is connected with utility power signal on the break-through plate 130.On themotherboard 104 management circuit is arranged, power supply signal is distributed to theserver board 102 of thissuperserver blade 132 by emulation ATX power management protocol.Removable fan bracket 140 is installed in thesuperserver blade 132 times, is used for to superserver 100 coolings.Preferablely be, rack 110 can hold many row'ssuperserver blades 132 on chassis assembly 128, and chassis assembly comprises a pair of parabasal plate 129, overlays the top of the power frame 146 that power supply 144 is provided.Preferablely be that rack 110 also comprises rack-mountedEthernet switch 44 and 147 and thestorage switch 149 that links to each other with disc driver by fibre channel networks.If will understand the detailed description ofhigh scalability superserver 100 in the relevant preferred embodiment of the present invention, see also the patented claim that is entitled as " the super server in high expansion internet " of awaiting the reply together of previous submission, draw herein and be reference.
Also to understand, though introduce mention when of the present invention be equipped withsuperserver blade 132 andserver board 102 Single Cabinet 110, with thevarious server 46 of the common formation of suitable applications software, distribute to first management group 52-a and second management group 52-b, have twosuperserver blades 132 separately at least, but group ofserver 40 can be held any user'smanagement group 52, and number of users depends on the sum ofserver 46 in the group of server 40.Preferable is a plurality of racks 110 can be integrated the sum that expands a certain server in station 46.As described later, also a plurality of racks 110 that are distributed in the different location can be coupled together a part asindividual server group 40, by 48 controls of superserver group manager.
In the preferred embodiment, can assemble the up-to-date processor of Intel, SPARC or PowerPC on theserver board 102 of eachsuperserver blade 132, wherein every kind of processor is all supported standard operation system environment such as Windows NT, Windows 2000, Linux or Solaris.Everysuperserver blade 132 can comprise one ormore server board 102, and the new shape factor that every block server plate can may adopt according to the shape factor or the industry of current ATX in the future is designed to uniprocessor or multiprocessor.Preferable is thatcommunication channel 106 uses controller zone network (CAN) bus, separates with the communication path of thenetwork switch 44 or storage switch 149.In addition, also can provide thestandby communication channel 106 of second failure to allow fault-tolerant and to be thatmanager software 48 provides the redundancy communication path.
In General Server, pointer and startup configuration information are provided with by the hand switch on the server board, or write PROM chipset on the server board, or be stored in the fixed position on the local hard drive that server board can visit with the form of fixed code.Management circuit on themotherboard 104 is designed to suitably be connected withserver board 102, and in fact make pointer and other start configuration information is provided by the Host Administration circuit.A superserver blade mainboard can select to install a localhard drive 107 bymotherboard 104 visits, makes motherboard can pass through the information ofcommunication channel 106 configuration store on this hard disk drive.In addition, also have electricpower management circuit 108 on themotherboard 104, can use rack 110 utility powers by emulation ATX power management sequence, control is delivered to the application ofserver board 102 power supplys.Preferablely be, backpassage Ethernet switch 147 can be implemented in the group ofserver 40 AC applications program and data message between eachserver board 102, and need not byinternet 22 information to be sent.
In a kind of preferred embodiment, each rack 110 can be installed nearly 32 superserver blades 132.In this configuration, thenetwork switch 44 and 147 can be installed two Foundry circuit switching ethernet routers 32.Preferable is that thenetwork switch 44 and 147 is by revising thereconfigurable server 46 of server ip address and treating being connected between the dynamic network switch of adjusting 44 and 147.With regard todisk storage unit 50, two kinds of selections are arranged.First kind of selection is that the form of exclusive software and hardware withcrossbar switch 149 embedded between thedisk storage unit 50 ofsuperserver blade 132 and intercepting SAN storage hardware configurations details, anddisk storage unit 50 is carried out communication with being connected by kit software API andcrossbar switch 149 of 132 of every blade mainboards.The another kind of selection is to utilize market fibre channel switch on sale or RAID storage box dynamically to set up being connected betweenblade mainboard 132 and the disk storage unit 50.In two kinds of methods, the software layer in the superserver group manager carries out necessary configuration adjustment to the connection betweensuperserver blade 132 and thenetwork switch 147 and the disk storage unit 50.In another embodiment, if the user account collection that part ofserver group 40 is supported comprises the server that can not reallocate 46 of certain radix always,part server 46 can forever be connected to reduce the exchange cost with the network switch or disk storage unit so.In this case, theserver 46 of this quantum can forever be connected withdisk storage unit 50 by thenetwork switch 149 relevant with thismanagement group 52 in eachmanagement group 52.
Again referring to Fig. 4 and Fig. 5, you can find that group of server of thepresent invention system 40 can carry out dynamic management to multi-user account's host services.You can see that also an Intranet 54 connects five station servers at least.Preferablely be that Intranet is by theidentical network switch 44 ofConnection Service device 46 andinternet 22 or realizes forming on the similar network switchs such as thenetwork switch 147 ofserver 46 interconnection.On themotherboard 104 of everystation server 46 management circuit is arranged all, communication channel is provided, theserver 46 that stands on theIntranets 54 that the network switch 44 and/or 147 forms with an independence for Taiwan carries out communication at least.
At least dispose fourstation servers 46 and carry out this locality decision-making software program 70 ofmonitoring server 46, and bycommunication channel 106 transmit mode information.Have at least two-server 46 to distribute to first management group 52-a of first user account, and be configured to visit exclusive software of first user account and data, for the internet of this user account provides host services; Have at least other two-server 46 to distribute to second management group 52-b of second user account, and be configured to visit second software and data that user account is exclusive, for the internet of this user account provides host services; Have at least astation server 46 to carry out main decision-making software program 72, this program is collected in the status information of this locality decision-making software program 70 of operation on other servers 46.In one embodiment, a pair ofserver 46 utilizes fault-tolerant coordination software to constitute dependent server jointly, as the fault-tolerant/redundant processing platform of main decision-making software program.As described later, main decision-making software program 72 according to the status information of collected this locality decision-making software program 70, is given second management group 52-b at least to major general's onestation server 46 from first management group 52-a dynamic assignment again.
Theserver 46 of twomanagement groups 52 can be distributed in any configuration of a certain user account appointment.As shown in Figure 3, threestation servers 46 of management group 52-b are set to front-end server, and astation server 46 is set to the rear end/calculation server of this user account.For responding the significantly increase of second management group 52-b user account peak use amount, main decision-making software program 72 is thought and is necessarycurrent server 46 as first management group 52-a is redistributed to second management group 52-b as rear end/calculation server.The first-selected framework that forms this decision-making will be introduced with the make a strategic decision operation ofsoftware program 72 of this locality.Next, as shown in Figure 4, main decision-making software program 72 indications are given second management group 52-b withserver 46 dynamic assignment again.
Be re-assigned to second management group 52-b though when introducing the preferred embodiments of the present invention, mentionserver 46 from first management group 52-a, but be construed as the present invention and also can be used for providing one group of still unallocatedpublic servicer 46 tocertain management group 52, these servers can be redistributed, and need not to require to withdraw from from the management group of moving 52.For example, can set up a group ofserver 40 that constitutes by 32station servers 46, distribute six for four user account each households, onestation server 46 is carried out main decision-making software program, still unappropriated sevenstation servers 46 of remainder constitute server library 56, and they can distribute in fourmanagement groups 52 of this group of server any one.According to the present invention, in group ofserver 40 operational processs, be dynamic to management set of dispense server, therefore, the preferred embodiments of the present invention use this server library 56 as impact damper.Because reallocation server 46 ' need not to withdraw from from themanagement group 52 at current place earlier, adds the time required in themanagement group 52 to so further reduced the server 46 ' of will reallocating.In one embodiment, existing hot server also has cold server in the server library 56.Hot server refers to be theserver 46 of a certain cura specialis configuration set, and therefore hot server adds the management group and need not to restart.Cold server is meant the stillunallocated server 46 of giving a certain curaspecialis group 52, so must restart before adding the management group.
Should understand, reallocation server 46 ' can be distributed to a new management group separately, also can be used as a group and redistribute, promptly simultaneously some the servers of redistributing 46 ' are re-assigned to second management group 52-b from first management group 52-a.How configuration network switch 44,147 andstorage switch 149 are with in the content that realizes dynamic reallocation, but be construed as if desired or wish to reduce the port of dynamic-configuration on the network switch 44,147 and/or thestorage switch 149,multiple servers 46 can be redistributed simultaneously as a group.
Of the present invention one big advantage is: server reconfigures to the process of second management group 52-b from first management group 52-a, in the server of redistributing 46 ' with first management organize all relevant status informations of certain particular user account before entering second management group 52-b with full scale clearance.Yet this provides a kind of from effective security mechanism, avoids intentionally or the data between unintended access different user account.Unlessserver 46 or 46 ' belongs to a certain management group 52-a, otherwise this server can't be visited data or the information of another management group 52-b.Storage space is shared in master server or other large server system utilizations and/or the common operation system provides host services for the different user account, must carry out complexity and have the software security function of potential problems, and the present invention has kept the advantage that the General Server group arranges that the simple physical that had is separated the different user account, simultaneously at needs or more can carry out dynamic assignment to this hardware again automatically during good utilisation hardware.Can only be by visiting these mandate that reconfigures and controls via main decision-making software program 72 withouter communication channel 106.
As shown in figure 14, everystation server 46 is connected withinternet 22 by programming, is controlled by main decision-making software program 72.Main decision-making software program 72 also is converted to the server of redistributing 46 ' with the exclusive software of storing second management group user account and is connected with the section data disk storage unit.For the consideration of aspects such as safety, fault isolation and bandwidth isolation, would rather use theouter communication channel 106 of the band that separates with Intranet 54 by thenetwork switch 44, carry out communication with the partial status information of main decision-making software program utilization at least.
In a kind of preferred embodiment,communication channel 106 is serialization controller Local Area Network (CAN) buses, operation inrack 106, and bandwidth is 1Mb/s, also in 106 operations of different racks, bandwidth is 1Mb/s to the secondary main line.The independent Intranet that adopts the IP agreement to carry out communication can be used forcommunication channel 106, and be not used in serial management interface such as CAN bus, but the designed function of this embodiment has surpassedcommunication channel 106 desired communication grade and the complexities that are connected with motherboard 104.Thoughcommunication channel 106 can be used as the part ofIntranet 54, consider for factors such as safety, fault isolation and bandwidth isolation, do not advocate this method.
Block diagram shown in Figure 8 has reflected that the present invention is applied to the hierarchical relationship between employed various data of a certain user account and the software layer.User data and database 60 constitute the basic unit of this hierarchical organization.The a plurality of storage unit that can select to add 62 pairs of formations of network data management softwarelayer storage system 50 are carried out user data 60 management.Troop and/or load balancing application software 64 constitutes the user website software that it is conventionally believed that and the top layer of data.Load balancing software 66 gathers a part aspublic management group 52 with multiple servers 46.Multiple routine operation system software is arranged, and eachserver 46 uses one.Another program is that load balancing software 66 and operating system software 68 can be integrated into the part of common software bag in the single management group 52.If want to understand the particulars of relevant operable SiteServer LBS framework, see also previous submission but the patented claim that is entitled as " the multiserver work allocation system that uses dynamic measurement system " of awaiting the reply together, draw herein and be reference.Above-mentioned routine operation system software 68 is a superserver of the presentinvention function software 48, and multi-user account 52-a and 52-b are carried out resource management.
As shown in Figure 9, in one embodiment of the invention, theserver 46 of distributing to first management group 52-a is positioned at first website 80, and the server of distributing to second management group 52-b is positioned at a distance of second far website 82 of first website 80.In the present embodiment, system also is provided with to the data of first management group of major general 52-a and copies to second website 82 automatically.In a kind of preferred embodiment, utilize the communication channel 84 separate with thenetwork switch 44 that the disk storage unit 50-b that the disk storage unit 50-a of first website 80 copies to second website 82 is got on.This set has two purposes: the first, and data are duplicated provides redundancy and backup protection, can carry out fault recovery when first website 80 breaks down; Second, the data of second website 82 are duplicated theserver 46 that makes the present invention will be positioned at second website 82 and are included the available server storehouse in, main decision-making software program may use these servers, satisfies the host services demand that first user increases by these servers of dynamic assignment again to first management group 52-a.
By coordinating the main decision-making software program 72 of first websites 80 and second website 82 with the make a strategic decision global decisions software routine 86 ofsoftware program 72 communications of each website master.This modularization layout makes main decision-making software program can manage the server resource of a certain website, and make website 80 and 82 to global decisions software routine amount of the claim outside remote service or with make a strategic decision server outsidesoftware program 70 amount of the claim or provide the roughly the same mode of the server that can be used for reallocating that available remote service is provided for the master of a certain website software program of making a strategic decision of this locality.
Preferable is that multi-site embodiment of the present invention utilizes market SAN on sale or NAS storage networking software to realize double-deck data redundancy and duplicates hierarchical system.As shown in Figure 9, the use version 74 of the user data of first user account is kept on thedisk storage unit 50 of first website 80.Take data image, data to hint obliquely at or redundant data safeguard measure such as RAID data protection is set up the backup version 76 of the user data of first user account on first website 80.It is second backup version 78 of user data generation of second first user account on the website 82 that network software utilizes communication channel 84.Use provides the redundancy communication path with the communication channel 84 that thenetwork switch 44 and being connected ofinternet 22 separate, and will generate the influence that second backup version, 78 required background communications activity are brought and reduce to minimum.In addition, the backup version 78 that is positioned at the user data of second first user account on the website 82 can pass through thenetwork switch 44 andinternet 22 transmissions.In another embodiment, the extra backup version of user data can duplicate at other website, server is given the ability of the user account that needs these resources with further expansion system from the inadequate user account of utilization of resources dynamic assignment again.
As shown in figure 10, the present invention is able to more efficiently utilization providing for multi-user account for resource that the ability of the user account need these resources makes group of server 40 from the inadequate user account of utilization of resources dynamic assignment again server the host services.For user account 91,92,93,94 and 95, each user account, the marginal overload bandwidth that each user account is kept relative stability are given in server 46 unified distributions.In the existing group of server, the variation of distributing to the hardware resource of certain user account is a unit with hour, day or week, and the present invention can reflect the variation of the server resource per minute that carries out dynamic assignment as required.Figure 10 has also reflected and has utilized a plurality of advantages that are distributed in the website localization part group of server 40 of diverse location.As can be seen, user account 94 and peak operating period of 95 are shifted the time of coming just because of having the time difference between website 80 and the website 82 from other user accounts 91,92 and 93 peak operating periods.It is poor that the present invention can utilize the time of peak between the operating period to shift, and the server capacity of not stopping circulation change is assigned to the website that is in the peak operating period from other little websites of current use amount.
As shown in figure 13, in the embodiment of multi-site configuration of the present invention, the time difference between at least three websites 80,82 and 84 present positions is preferably 24 ÷ (N+1) hours, wherein N represent multi-site dispose in the quantity of different websites.In the framework that three independent websites 80,82 and 84 are arranged, the time difference between the site location is preferably 8 hours.The user account of three websites of official post used pattern to gather when this distributing position was formed, server by some is served for it, if but every station server of certain website then needs the server of greater number can not use from server that one or more other websites dynamic assignment again come the time.When website 80 is in night during usage level, server just can be given from website 80 dynamic assignment again and be in the website 82 of usage level in the daytime, and just can embody this advantage this moment.Simultaneously, website 84 is in the usage level in night, may be fit to also may be not suitable for server is re-assigned to another website from this website, and vice versa.In general, a website is used capacity to the website that is positioned at one time zone, back (as the east of this website) earlier, and then seeks overhead provision to the website that is positioned at last time zone (as western part of this website).Can and estimate that the use pattern is provided with priority according to historical use amount.
Please referring to Figure 11, will introduce a preferred embodiment of main decision-making software program 72 here now.Main decision-making software program comprises resource database 150, service-level agreement database 152, main decision logic module 154 and scheduler module 156.Main decision logic module 154 accessible resource databases 150 and service-level agreement database 152, and the information of status information and resource database 150 and service-level agreement database 152 compared, so that determine whether will be with server from the user account of first user account dynamic assignment to the second again.Scheduler module 156 is connected with main decision logic module 154, initialization pointer by utilizing the server 46 ' that communication channel 106 setting redistributes is with the exclusive software and data of visiting user account among second management group 52-b and the newly assigned server 46 ' of counterweight reinitializes so that at least one station server is incorporated second management group 52-b into, and under main decision logic module 154 controls dynamic assignment server again.Preferable is that scheduler module 156 comprises that a cover connects the personality module of rule 160 and one cover at each station server 46.Concatenate rule 160 provides the instruction that connects certain particular server 46 and a certain network switch 44 or data storage cell 50.Personality module 162 is described the software arrangements to be added to the server board 102 of user account place management work group in detail.In case scheduler module 146 decisions are necessary to redistribute server, it will weigh a whole set of concatenate rule 160 and a whole set of personality module 162, to determine that how making up server 46 distributes to this cura specialis group 52.
Can also be from another angle explanation the present invention how for different accounts provide the DynamicHost service, Here it is is assigned to the virtual server storehouse withpart server 46, may carry out software and data that selectivity is configured to visit a certain cura specialis group 52.When scheduler module 146 was determined to addserver 46 for a certaincura specialis group 52, it was assigned to this management group with a station server automatically from the virtual server storehouse.Otherwise, if confirming the management group, scheduler module can discharge astation server 46, this d/d server will add the virtual server storehouse to so, can redistribute to other management group.From this point as can be seen, superservergroup manager software 48 can be in time or from the virtual server storehouse " manufacturing " or produce one or more virtual server as required.As previously mentioned, the virtual server storehouse can be heat reservoir, freezer or both combinations.The purpose of making or make up virtual server is to want the management group of server to use according to the supply and demand that requires of concatenate rule 160 and personality module 162.
In the present embodiment, main decision logic module 152 links to each other with supervisor console 158, accounts maintenance and lastest imformation that supervisor console can show main decision-making software program information and accept to handle in each database.Integrated account software module 160 in thesuperserver group manager 48 is in order to follow the tracks of the situation of chargeing by the server of distributing to certain user account.Preferable is when giving this user account according to user's service-level agreement with server dynamic assignment again, to calculate the host services expense of user account with higher rate.
Figure 12 has reflected three kinds of different service-level agreements of a certain user account.In the present embodiment, service-level agreement is defined in that (as one month) provides host services in a period of time.At the first estate 170, provide the capacity of 640,000 host services that connect simultaneously for user account.If user account does not need to redistribute the more high capacity of server support than the first estate 170 appointments, then charge to the user by the rate of this grade designated capabilities.In second grade 172, but the user account dynamic expansion, and support doubles the capacity of the first estate 170.In a kind of preferred embodiment, in case superserver group manager 48 is for the capacity of supporting second grade 172 and satisfy greater than the demand of estimating the peak use amount and give user account with server dynamic assignment again, will charge to user account by higher rate during the then extra use amount.In addition, also can disposablely collect the more high-grade service fee that starts 172 representatives of second grade to user account.In one embodiment, the service rate of second grade 172 is the several times of the first estate 170.Second grade 172 has been represented in certain period and can have been guaranteed to the expansion grade that the user provides.At last, the tertiary gradient 174 provides optional expansion to replenish the service of grade, can be used for providing host services for user account.In the present embodiment, the tertiary gradient 174 can provide the service that is higher than the first estate 170 several times.In one embodiment, be the service that provides this expansion to replenish grade, the multi-site layout that host computer system utilizes preamble to mention is added the server of requirement to satisfy the service of this grade.Preferable is redistribute the expansion supplemental services to user account during, to charge to this user account according to the higher rate in the second level.In one embodiment, providing expansion to replenish the tertiary gradient 174 viabilities to user account, the service rate of the tertiary gradient 174 can be than the first estate 170 big several times.Equally, to the disposable expense that starts the tertiary gradient 174 services of collecting of user account.When designated duration finished, the user can adjust the grade of service of agreement.
As shown in figure 12, service-level agreement for the higher expectation peak use amount of certain user account of response from having increased by 50% to the second phase between the first phase.Preferablely being, is unit with the moon during the service-level agreement of a certain user account, how to adjust service-level agreement to user's suggestion in the next one phase of chargeing simultaneously.Though this example illustrates to connect simultaneously, but the service-level agreement that is interpreted as certain user account can relate to various performances weighs factor, as connecting, hit rate, volume of transmitted data, number of transaction, tie-time, the resource of different application software program use, income or any composite factor of generation simultaneously.Will also be understood that the service that different brackets can be provided for dissimilar resources such as front-end server, back-end server, network connection or disk storage unit for service-level agreement.
Please see Figure 15, this block diagram introduction be the first-selected framework of this machine decision-making software program 70, shown a series of measurement modules 180,181,182,183 and 184 among the figure, each module is independently assessed the ruuning situation of the particular server of carrying out local decision-making software program 70.These measurement modules output to assessment result the collection modules 190 of this machine decision-making software program 70.Prediction module 192 generates the expectation response time and the possibility of different requests.The preferential input 194 that fuzzy EVAC (Evacuation Network Computer Model) 196 utilizes main decision-making software program 72 to provide from service-level agreement database 152, determine whether that request management group 52 is added or unloading superserver blade mainboard 104, the request with interpolation or unloading blade mainboard sends main decision-making software program 72 to by communication channel 106 then.In one embodiment, collection modules 190 is carried out on every station server 46 of certain management group 52, and prediction module 192 and fuzzy interference module 196 are only carried out on a station server 46 of certain management group 52, and the output of each measurement module 180-184 is delivered to given server 46 by communication channel 106.In another embodiment, in order to carry out redundancy or distribution process to generating to add or unload the required information of superserver blade request, collection modules 190, prediction module 192 and fuzzy interference module 196 can be carried out on the multiple servers of certain management group.
Preferable is that collection modules 190 realizes balance according to following formula in each measurement module 180-184:
BK=[(∑Tki/Wk)-mink]×100/(maxk-mink)-50
I=1 is to Wk
Wherein, TKiI request required time representing the K kind to measure, WkThe time size of representing the K kind to measure, minkRepresent the K kind to measure and estimate minimum time, maxkRepresent the K kind to measure the maximum permission time.The balance request rate B that drawsKSend the prediction module 192 and the fuzzy interference module 196 of local decision-making software program 70 to.The time size that the K kind is measured is set, reduces the unnecessary invasion of measurement module 180-184 as far as possible, can organize 52 user demand increases to management again simultaneously and make in time response fully.
What Figure 16 showed is the workload that each measurement module 180-184 records under the different loads condition.Measure without any a kind of workload as can be seen and can constantly estimate the possibility of estimating response time and this response time.Therefore, fuzzy interference module 196 must comprise three basic parameters: to the expectation response time of various requests, the priority level of request and the possibility of generation.Fuzzy interference module 196 is got up these three parametric synthesis, determines whether to ask to add or unload the superserver blade ofmanagement group 52 in view of the above.Below be the fuzzy example that disturbs rule:
(if priority level were urgent) and (possibility is for abundant) and (estimating that the response time is for too high), then (blade mainboard would be added in request)
Preferablely be, the net result of fuzzy interference module 196 is to form decision surface, promptly draws onmanagement group 52 is estimated the coordinate of the possibility that response times and this expectations response time take place and asks the needs of interpolation server.Figure 17 is the example of this decision surface.
Disclosure content of the present invention is protected by copyright.The copyright owner allows according to the file of patent and trademark office or the slavish copying disclosure content of the present invention of archives, but keeps the whole copyrights under other situation.
Though, will be appreciated that and to make a large amount of modifications and change, and determine scope of the present invention by claim at this preferred embodiment of having introduced automatic system of the present invention.