CN104283948B - Server cluster system and its implementation of load balancing - Google Patents
Server cluster system and its implementation of load balancing Download PDFInfo
- Publication number
- CN104283948B CN104283948B CN201410512754.5A CN201410512754A CN104283948B CN 104283948 B CN104283948 B CN 104283948B CN 201410512754 A CN201410512754 A CN 201410512754A CN 104283948 B CN104283948 B CN 104283948B
- Authority
- CN
- China
- Prior art keywords
- node
- standby
- load
- information
- module
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
- H04L67/1004—Server selection for load balancing
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
- Computer And Data Communications (AREA)
Abstract
A kind of method that the present invention provides server cluster system and its load balancing is realized, system therein includes: host node, standby node and ordinary node；Host node includes information collection module, load computing module and connection distribution decision-making module；Standby node for reporting the information of the standby node, and the synchronizing information with host node to host node, and screens ordinary node, and after becoming host node so as to it, the ordinary node selected is as new standby node；Ordinary node for reporting the information of the ordinary node to host node, and when it becomes standby node, carries out standby node switching.Using the present invention, by using the server cluster system with standby usage node of netted networking, connection distribution is realized using the structure of standby usage node, reaches the function of sharing connection, be able to solve connection distribution delay, connect unstable and insecure problem.
Description
Technical field
The present invention relates to server cluster technical fields, more specifically, are related to a kind of server cluster system and its negative
Carry balance realizing method.
Background technique
With the frequent generation of the rising of car ownership, the aggravation of road congestion conditions and traffic accidents, more
Recognize that the deep driving behavior for understanding driver will be helpful to formulate more reasonable friendship come more organizations and individual
Logical rule and the significantly more efficient intelligent driving navigation system of design reduce the mesh that traffic accident improves traffic efficiency to reach
's.And above-mentioned analysis needs to establish on the basis of a large amount of travelling data, needs a large amount of intelligent terminals (including vehicle-mounted end
The equipment such as end, mobile phone terminal) constantly acquisition travelling data is uploaded to onboard servers and is analyzed in real time.Therefore, with
The continuous promotion of intelligent terminal and its portfolio, single server can no longer meet requirement, and server cluster becomes one kind
Need solution to problem.
For intelligent terminal, the important task of data collection is carry, needs constantly efficiently to acquire the driving number of vehicle
According to, and keep connecting with server, the data collected are transmitted to server-side in real time, this process is referred to as " data pendulum
It crosses ".In real-time " data ferry-boat " field, long connection type is generally used, i.e. holding intelligent terminal is online for a long time, to protect
Demonstrate,prove the real-time of data.Simultaneously as its data volume depends on the usage amount of vehicle, therefore certainly exist big with server end
Measure data interaction.Therefore, under this framework, how design server, on the one hand meet the long connection requirement of a large amount of intelligent terminals,
On the other hand guarantee that transmitting-receiving and the storage of mass data are normal, become a problem to be solved in the field.
In view of the above-mentioned problems, a solution is to use reverse proxy to realize load balancing.
In the load balancing scheme realized using reverse proxy, what is utilized is that Reverse Proxy realizes link point
Hair.Its specific method can be described as: Reverse Proxy receives the connection request come on automatic network first, and establishes with it
Connection, is then transmitted to internal server for the access request from client terminal subsequently received, and will be from internal server
The result of upper return is transmitted to the client terminal for initiating the connection request accordingly again.In the process, Reverse Proxy
A server externally is shown as, is then internally main entrance and the general export of connection request, is responsible for connection being distributed to inside
Each server.
The defect of this solution is: since Reverse Proxy is in the application layer in OSI seven layer model, because
This is necessary for each application service and specially develops a Reverse Proxy, limits reverse proxy load balancing in this way
The application range of technology.Currently, reversed proxy server is generally all used for the load balancing to web server, in addition, in reversed generation
In the load balancing that reason mode is realized, for act of agency each time, proxy server is switched on two connections, and one right
Outside, one internally therefore when concurrent connection request quantity is very big, the load of proxy server is also just very big,
It is all the bottleneck that proxy server itself becomes service performance under normal conditions in such network.
Another most common solution is the load balancing realized based on network address translation (NAT) technology.Network
It is in order to solve the problems, such as that IPv4 public network address quantity is reduced, internal multiple servers are external at the beginning of address translation technique is born
It is mapped as an individual server, to make internal multiple servers externally provide service using a public network IP.
In this solution realized based on network address translation (NAT) technology, the connection initiated from network is asked
It asks, carries out address conversion by address translation server, then be distributed on internal each server.It realizes in this way
Load balancing, although being able to solve reversed proxy server is located at application layer, it is still necessary to be directed to every kind of application service specially to open
The problem of sending out Reverse Proxy；And there is still a need for beat for request every time for the load balancing that network address translation mode is realized
Two internal and external connections are opened just to be able to achieve, thus it is very big in data when address translation server itself can be at
For performance bottleneck.
At the same time, it in the load balancing that network address translation mode is realized, is accomplished that dynamic when each request reaches
State selects the processing request of some internal server, and needs to be correctly found request source when processing result returns, it is therefore desirable to
Its conversion mapping table is retained to each request.Same reason is also required to the maintenance of request mapping table when request amount becomes larger
Vast resources is consumed, compared with reversed proxy server, it is easier to cause the performance issue of address translation server；Finally, in mesh
In the Address Translation services formula load balancing of preceding actual use, address translation feature is integrated in hardware switch mostly,
The only simple selection strategy realized, cannot support more optimized load balancing and more complicated application protocol, and
There is certain difficulty in extension.
Summary of the invention
In view of the above problems, the object of the present invention is to provide a kind of server cluster system and its load balancing realization sides
Method is realized by using the server cluster system with standby usage node of netted networking using the structure of standby usage node
Connection distribution, reach the function of sharing connection, with solves the problems, such as connection distribution be delayed, connect it is unstable and insecure.
On the one hand, the present invention provides a kind of server cluster system, including host node, standby node and ordinary node；Its
In,
Host node includes information collection module, load computing module and connection distribution decision-making module；
Information collection module, the load letter reported for storing the standby node being connected with host node and each ordinary node
Breath；
Computing module is loaded, for obtaining the load of node corresponding with load information according to the load information reported
Value；
Connection distribution decision-making module, for being attached distribution by HTTP redirection, wherein according to load computing module
The load value of the node of acquisition selects the node of optimal processing ability, the node as connection distribution；
Standby node, for reporting the load information of standby node, and the synchronizing information with host node to host node, and
Ordinary node is screened, the ordinary node selected is as new standby node；
Ordinary node will for reporting the information of the ordinary node to host node, and when it becomes standby node
It is switched to row standby node.
Furthermore it is preferred that scheme be, host node further include management conversation module；For managing customer terminal and group system
Between conversation history record, and be responsible for deleting between client terminal and group system that sluggish historical session is remembered for a long time
Record.Furthermore it is preferred that scheme be, in information collection module create a nodelist, nodelist for store with main section
The load information of the connected each node of point；Wherein,
Nodelist includes update mechanism, and update mechanism is for deleting offline node.
Furthermore it is preferred that scheme be that load information includes: cpu busy percentage, the memory for each node being connected with host node
Utilization rate, network bandwidth occupancy and socket connection occupancy；
In load computing module, to each node by the load computational algorithms of particular weights, obtain respectively its for
CPU intensive type, memory-intensive, network bandwidth be intensive and the load capacity of socket intensive request, and the knot that will acquire
Fruit updates to nodelist.
Furthermore it is preferred that scheme be that standby node includes standby node information reporting module, synchronizing information module and common
Node screening module；
Standby node information reporting module, for reporting load information to host node cycle；
Synchronizing information module for the load information in periodically synchronizing information collection module, and is stored；
Ordinary node screening module, for selecting to load in ordinary node the smallest when standby node becomes host node
It is alternative that node becomes it.
Furthermore it is preferred that scheme be that ordinary node includes ordinary node information reporting module and standby node switching module；
Ordinary node information reporting module, for reporting load information to host node cycle；
Standby node switching module will be new to what is gone out selected by it for after standby node becomes new host node
Standby node is given notice, and ordinary node is switched to standby node accordingly, and undertakes the responsibility of standby node.
On the other hand, the present invention also provides a kind of implementation of load balancing based on server cluster system, servers
Group system includes host node, standby node and ordinary node, wherein:
Receive to host node cycle coupled standby node, the load information that each ordinary node reports, and root
According to the load information reported, the load value of each node corresponding with load information is obtained；
When there is client terminal connection request, host node is attached distribution by HTTP redirection, wherein host node root
According to the load value of each node of acquisition, the node of optimal processing ability is selected, the distribution node as connection；
When host node breaks down, standby node carries out active-standby switch, becomes new host node, executes the duty of host node
Duty, while enabling new standby node；
New standby node is reported to load information in new host node, and the synchronizing information with new host node, together
When ordinary node is screened, select the alternative of new standby node.
Furthermore it is preferred that scheme be the session management mechanism that is provided in the master node, session management mechanism includes client's end
Conversation history between end and group system records, and is responsible for deleting inactive for a long time between the client terminal and host node
Historical session record, make selection connection distribution node optimal selection.
Furthermore it is preferred that scheme be to create a nodelist in the master node, the load information of each node is stored in
In nodelist, before the connection of customer terminal request,
Judge in nodelist with the presence or absence of offline node, if it exists offline node, then by this node from
It is deleted in nodelist；Offline node is standby node if it exists, then selects new standby node in ordinary node again,
And synchronizing information is carried out with host node；
If offline node is not present in nodelist, distribution is attached according to the connection request of client terminal.
Furthermore it is preferred that scheme be, during new standby node screens ordinary node, new spare section
Point selects the smallest node of load alternative as it from ordinary node.
Furthermore it is preferred that scheme be that load information includes: cpu busy percentage, the memory for each node being connected with host node
Utilization rate, network bandwidth occupancy and socket connection occupancy；
By the load computational algorithm of particular weights, each node calculates separately it for CPU intensive type, memory-intensive
The load capacity of type, network bandwidth intensity and socket intensive request, formula are as follows:
rl=f (cu,mu,nu,su)
Wherein, rlIndicate the surplus resources percentage of node l；
cu、mu、nu、suCpu busy percentage, memory usage, network bandwidth that the node received uploads is respectively indicated to occupy
Rate and socket connection occupancy.
Furthermore it is preferred that scheme be, according to the load information for receiving each node and uploading, to calculate surplus resources percentage
Than formula is shown below:
Wherein,
rl(k) the surplus resources percentage at k moment is indicated；
cu(k) k moment CPU usage is indicated；
cu(k-1) utilization rate of (k-1) moment CPU is indicated；
In (1-tu(k)-tconst) in, as t=s, tu(k) the socket connection utilization rate at k moment, t are indicatedconstIt indicates
The pre- allowance of socket resource, (1-tu(k)-tconst) indicate the remaining available socket resource situation of node t moment；
As t=m, tu(k) memory usage at k moment, t are indicatedconstIndicate the pre- allowance of memory source, (1-tu
(k)-tconst) indicate the remaining available memory source situation of node t moment；
As t=n, tu(k) the network bandwidth occupancy at k moment, t are indicatedconstIndicate the pre- allowance of network bandwidth, (1-
tu(k)-tconst) indicate the remaining available network bandwidth resources situation of node t moment.
It can be seen from the above technical scheme that server cluster system and its implementation of load balancing of the invention, with biography
The scheme of system is compared and is had the advantages that
Firstly, cluster networking uses netted networking structure, and have the progressive relationship of hierarchy between multiple nodes, one
Aspect improves service stability on the net from group, and being on the other hand conducive to select best performance node in the whole network becomes standby node,
To become host node under host node failure conditions, global optimum's election has been ensured；
Secondly, the primary server realized in cluster networking mode stays field, standby server association control, common server candidate
Layout, active and standby direct carry out synchronizing information, can when host node failure under the premise of guaranteeing to reduce synchronous information overhead
To realize that the seamless switching between host node and standby node, standby node are capable of the function of hosting node, single-point event is shielded
Barrier improves cluster and externally provides the availability of business；
Again, the connection distribution algorithms proposed in host node, under the resource situation for comprehensively considering each node, selection is most
The node of excellent processing capacity receives service request, can utilize the processing capacity of each node, connection distribution to greatest extent
On the one hand mechanism meets the needs of connection reasonable distribution, while greatly reducing the burden of host node server, takes winner
Business device can carry out active/standby backup co-ordination, make very big contribution for cluster overall stability, simultaneous session administrative mechanism can
To further decrease time delay when distribution connection, guarantee the reliability and timeliness of business.
To the accomplishment of the foregoing and related purposes, one or more aspects of the present invention include be particularly described below and
The feature particularly pointed out in claim.Certain illustrative aspects of the invention is described in detail in the following description and the annexed drawings.
However, these aspects indicate only usable some of the various ways in the principles of the present invention.In addition, of the invention
It is intended to include all such aspects and their equivalent.
Detailed description of the invention
By reference to the following description in conjunction with the accompanying drawings and the contents of the claims, and with to it is of the invention more comprehensively
Understand, other objects and results of the present invention will be more clearly understood and understood.In the accompanying drawings:
Fig. 1 is reverse proxy and the load balancing networking structure schematic diagram that network address transmission technology is realized；
Fig. 2 is the server cluster system networking structure schematic diagram according to the embodiment of the present invention；
Fig. 3 is the server cluster system logical construction block diagram according to the embodiment of the present invention；
Fig. 4 is the master node logic structural block diagram according to the embodiment of the present invention；
Fig. 5 is the server cluster system Application Example schematic diagram according to the embodiment of the present invention；
Fig. 6 is to be illustrated according to the implementation of load balancing process based on server cluster system of the embodiment of the present invention
Figure；
Fig. 7 is to be illustrated according to the client terminal of the embodiment of the present invention and the principle process of group system connection request process
Figure；
Fig. 8 is the workflow schematic diagram realized according to the load balancing of the server cluster system of the embodiment of the present invention.
Identical label indicates similar or corresponding feature or function in all the appended drawings.
Specific embodiment
In the following description, for purposes of illustration, it in order to provide the comprehensive understanding to one or more embodiments, explains
Many details are stated.It may be evident, however, that these embodiments can also be realized without these specific details.
In the load balancing networking structure that reverse proxy and network address transmission technology are realized, network construction form is adopted
With the tree of single-point entry, the load balancing networking knot that Fig. 1 shows reverse proxy and network address transmission technology is realized
Structure, as shown in Figure 1, the main entrance by a server as flow, by traffic distribution to other internal servers, then will return
The flow returned passes to connection originating end.Simultaneously as the task that portal server undertakes is more, be generally difficult to its into
Row active/standby backup, active-standby switch time are also difficult to meet the requirements, wherein active/standby backup is exactly standby node periodically to main section
The load information that point receives is backed up.
For in the Address Translation services formula load balancing actually used at present, only simple selection strategy, Bu Nengzhi
The problem of holding more optimized load balancing and more complicated application protocol, the present invention by analysis realize by reversed proxy server
Load balancing, the schemes such as load balancing realized of network address transmission technology, analyzing the basis of all kinds of scheme advantage and disadvantage
On, in conjunction with business scenario, a kind of new server cluster system and implementation of load balancing are proposed, it is provided by the invention
It is netted networking structure that server cluster system, which uses, and on the basis of the agreement used in carefully analyzing, innovative
Ground dexterously realizes connection distribution using the specific function of agreement, has reached the function of sharing connection, connect distribution with tradition
Algorithm is compared, and is had small, the feature of stable connection that is delayed, is adapted to the characteristic that vehicle-mounted client terminal fast moves.
Hereinafter, specific embodiments of the present invention will be described in detail with reference to the accompanying drawings.
In order to illustrate the reticular structure of server cluster system provided by the invention, Fig. 2 shows implement according to the present invention
The netted networking structure of the server cluster system of example.
As shown in Fig. 2, the networking structure of server cluster system proposed by the present invention is netted networking structure, wherein tool
Three kinds of node types, host node, standby node and ordinary node are set in the cluster service framework of standby active and standby function.
Cluster networking of the invention improves traditional networking mode in terms of two, firstly, mitigating the work of portal server
Work amount burden allows to meet the requirement of master and backup network, realizes that primary server stays field, standby server association control is common to take
The layout for device candidate of being engaged in, shields Single Point of Faliure, greatly improves networking stability；Secondly, tree-like networking structure is replaced
For netted networking structure, service stability on the one hand is improved on the net from group, is on the other hand conducive to best performance in election the whole network
Node becomes standby node, to become host node under host node failure conditions, has ensured global optimum's election.
The netted group-network construction of cluster used in the present invention is explained in detail below.
From figure 2 it can be seen that proposing that the scheme of standby node is mainly based upon the portal service of current load balancing
The task that device undertakes is more, is generally difficult to carry out it active/standby backup, the active-standby switch time is also difficult to meet the requirements, therefore
Proxy server will necessarily become the important node of Single Point of Faliure under this mode；And it is active and standby in the present invention between synchronizing information
Be exactly for the ease of standby node when host node failure can rapid pipe connecting work, if only existing host node and common section in net
Point, to realize that seamless switching, principal point just need and net interior ordinary node to carry out synchronizing information, the information synchronized in this case
The expense of amount can be bigger in standby node synchronization overhead than only.Each node is briefly introduced below, it is whole to understand
Networking mode.
Host node is responsible for collecting the nodal information of each node in cluster, and calculates through overload, calculates each node
Load value, the connection request of host node managing customer terminal, to a large amount of client terminal connection carry out according to certain rules
Connection distribution；At the same time, host node adds session management mechanism, establishes between functional client terminal and server cluster system
Express passway further lowers time delay.
Standby node mainly realizes three functions, first is that the load information of oneself node is periodically reported to host node,
The foundation of connection distribution node is selected as host node, while standby node periodically loads letter out of host node synchronized clusters
Breath combines with host node and realizes active/standby backup, and standby node is switched to host node under host node abnormal conditions, avoids because of main section
The case where point exception causes group system that cannot externally provide business；Finally, standby node is responsible for electing performance in ordinary node
Optimal node, to become standby node after oneself becomes host node.
Ordinary node realizes load information period reporting mechanism, is responsible for reporting loading condition to host node cycle, meanwhile,
Most important task is to handle the link for redirecting the client terminal to come by host node, provides service for client terminal.
From the point of view of the composition function description of the server cluster system in Fig. 2, server cluster system provided by the invention
System, host node are responsible for balancedly breaing up the professional chief connection of smart client terminal in cluster on each node, are avoided single
Performance bottleneck brought by node is quickly redirected secondly, the session management mechanism on host node can be realized connection, is dropped
Low time delay；The united back mechanism of standby usage node, standby node can quickly become newly when can be realized host node failure
Host node, the offer business service that when group system failure node still can be stable.After standby node becomes host node,
New standby node will be selected, and entire group system becomes the progressive system having levels, as long as there is a server can normal work
Make, normal service can be provided out, maximizes the availability for improving cluster.
The netted networking structure provided according to the present invention, the present invention have carried out point the node in server cluster system
Class is always divided into host node, standby node and ordinary node three classes, and all kinds of nodes cooperate jointly, and the division of labor is clear, is externally providing
While service, cluster internal load balancing is realized, improves system overall stability.Fig. 3 shows real according to the present invention
Apply the server cluster system logical construction of example.
As shown in figure 3, the present invention provides a kind of server cluster system 300, comprising: host node 310, standby node 320
With ordinary node 330.
Wherein, host node 310 includes information collection module 311, load computing module 312 and connection distribution decision-making module
313。
Standby node 320 is used to report the load information of standby node, and the synchronizing information with host node to host node, with
And ordinary node is screened, the ordinary node selected is as new standby node.
Ordinary node 330 is used to report the load information of ordinary node to host node, and when it becomes standby node,
It is switched to standby node.
For the logical construction for the host node being described in detail in server cluster system, Fig. 4 shows real according to the present invention
The master node logic structure of example is applied, as shown in figure 4, host node 310 further comprises information collection module 311, load calculating mould
Block 312, connection distribution decision-making module 313 and session management module 314.
Specifically, information collection module 311 is used to storing the standby node being connected with host node and ordinary node reports
Load information.
Load computing module 312 is used to obtain the negative of node corresponding with load information according to the load information reported
Load value.
Connection distribution decision-making module 313 is used to be attached distribution by HTTP redirection, wherein calculates mould according to load
The load value for the node that block obtains, selects the node of optimal processing ability, the node as connection distribution.
Conversation module 314 is managed for the conversation history record between managing customer terminal and group system, and is responsible for deleting
Except sluggish historical session records for a long time between client terminal and group system.
Wherein, in information collection module 311 create a nodelist (node listing), nodelist for store with
The load information that the connected each node of host node reports；Wherein, nodelist includes update mechanism, and update mechanism is for deleting
Offline node.
Load information includes: that cpu busy percentage, memory usage, the network bandwidth for each node being connected with host node account for
With rate and socket (socket is the intervening software abstraction layer that application layer is communicated with TCP/IP protocol suite, it is a group interface,
For users, one group of simple interface be exactly all, allow socket to remove a group organization data, to meet specified agreement) connection account for
With rate；In load computing module 312, to each node by the load computational algorithms of particular weights, obtain respectively its for
CPU intensive type, memory-intensive, network bandwidth be intensive and the load capacity of socket intensive request, and the knot that will acquire
Fruit updates to nodelist.
Standby node 320 further comprises standby node information reporting module, synchronizing information module and ordinary node screening
Module.
Specifically, standby node information reporting module, for reporting load information to host node cycle.
Synchronizing information module for the load information in periodically synchronizing information collection module, and is stored.
Ordinary node screening module, for selecting to load in ordinary node the smallest when standby node becomes host node
It is alternative that node becomes it.
Ordinary node includes ordinary node information reporting module and standby node switching module.
Specifically, ordinary node information reporting module, for reporting load information to host node cycle.
Standby node switching module will be new to what is gone out selected by it for after standby node becomes new host node
Standby node is given notice, and ordinary node is switched to standby node accordingly, and undertakes the responsibility of standby node.
In addition, in the present invention, being described in detail by taking the server cluster system of five nodes composition as an example, Fig. 5 shows
Server cluster system Application Example structure according to an embodiment of the present invention is gone out.
As shown in figure 5, A is host node, B is standby node, and C, D, E are ordinary node.Wherein, host node A includes information
Collection module, load computing module, connection distribution decision-making module and session management module.Standby node B includes standby node letter
Cease reporting module, synchronizing information module and ordinary node screening module.Ordinary node C, ordinary node D, ordinary node E are wrapped respectively
Include ordinary node information reporting module and standby node switching module.
In the information collection module of host node A as shown in Figure 5, host node A is responsible for and standby node B and ordinary node
C, stable long connection status is kept between D, E, and the status information of each node is obtained with this.One is created on A node
Nodelist, to store the load information that each node being attached thereto reports, whenever A node receives B, C, D, E period
(status information includes: the load information of each node, i.e. cpu busy percentage, memory usage, net when the status information that property reports
Network bandwidth usage and socket connection occupancy), A node is saved the update of corresponding load information to this node
In nodelist, and distribute the selection gist for sentencing timed node using these information as connection.Nodelist includes update mechanism,
Update mechanism ensure that the ossified node by not online (offline) is deleted in due course.Such as: when host node A does not receive one for a long time
When the corresponding information of node, that is, thinks that the node breaks down, therefore the information of the node is rejected from nodelist, avoid
A node says that connection is distributed on the node when being attached distribution.
In the load computing module of host node A as shown in Figure 5, node that A node was collected in information collection module
Load information, including CpuUseRate, MemUseRate, NetworkUseRate, SocketUseRate have respectively represented every
Cpu busy percentage, memory usage, network bandwidth occupancy and the socket connection occupancy of a node.A node is according to collection
To load information the load value of each node is calculated, loading algorithm will using particular weights load computational algorithm,
That is: it is directed to each node, it is intensive for CPU intensive type, memory-intensive, network bandwidth intensity and socket to calculate separately it
The load capacity of type request is more reasonably attached distribution, calculated knot effectively to handle different request types
Fruit updates into the nodelist of corresponding node, is rapidly attached distribution convenient for subsequent.
In the connection distribution decision-making module of host node A as shown in Figure 5, when there is intelligent terminal connection request, selection
The node of optimal processing ability, and connection is distributed on the node, it is assumed that in tetra- nodes of B, C, D, E, selection is being loaded
The node that optimal processing ability is calculated in computing module is D node, then the connection of intelligent terminal is redirected to D section by A node
On point, by this scheme, breaing up on each node for equilibrium can will be connected, and the delay of processing can be reduced.Even
The detailed process and related algorithm for connecing distribution will be further described below.
In the session management module of host node A as shown in Figure 5, host node adds session management mechanism, management intelligence
The conversation history of terminal and server cluster system record, is quickly distributed, to greatest extent for the subsequent connection of intelligent terminal
Ground reduces request time delay.Session management module is responsible for deleting those sluggish conversation recordings for a long time, to guarantee the company of resurrection
Connecing can be broken up according to connection distribution algorithms, keep connection distribution more reasonable.
That is, session management module will delete those sluggish conversation recordings for a long time, it is due to intelligent terminal
It is not connected to for a long time with the node for connecting distribution, the load of the node in group system can change, connection distribution originally
The load of node can also change, therefore cannot be guaranteed that the node of original connection distribution is the section of optimal processing ability
Point, therefore this conversation recording is deleted, the node of new connection distribution is reselected, using the node for guaranteeing to select as optimal processing energy
The node of power.Therefore time delay when session management module further decreases distribution connection, guarantee intelligent terminal and group system it
Between business reliability and timeliness.
In the standby node information reporting module of standby node B as shown in Figure 5, standby node B is periodically to main section
Point A reports the load information of oneself, and load information includes cpu load, memory load, network bandwidth load and socket load
Deng facilitating host node A to be attached distribution and node administration using these information.
In the synchronizing information module of standby node B as shown in Figure 5, standby node B is periodically same from host node A
Nodelist information is walked, is stored in the memory of this node, to the information of group system between synchronous standby usage node.Spare section
Point B detects the presence of host node A using timing mechanism and event trigger mechanism: firstly, host node A leads timing detection
The presence of node A；Secondly, will be triggered to host node A when standby node B periodically reports information to fail to host node A
Presence detection, if testing result is host node, A is offline, and standby node B will be switched to host node, and mass-send notice collection
Each node in group's system, notifies the ordinary node selected in advance to become standby node, to continue to ask the connection of intelligent terminal
Offer service is provided.
In the ordinary node screening module of standby node B as shown in Figure 5, standby node B will be selected with longer cycle
Loading the smallest node becomes the alternative of oneself, and when oneself is as host node, the node is notified to become standby node, from
And the repairing of the networking structure of predetermined network is completed in the case where host node failure, continue the stability for guaranteeing network, externally
Service is provided.
In the ordinary node information reporting module of ordinary node as shown in Figure 5, ordinary node C, D, E are periodically to master
Node reports the load information of oneself, including cpu load, memory load, network bandwidth load and socket load etc., facilitates master
Node is attached distribution and node administration using these information.
In the standby node switching module of ordinary node as shown in Figure 5, when standby node B becomes new host node
Afterwards, it will give notice to the new standby node gone out selected by it, ordinary node C carries out standby node switching accordingly, including holds
Various functions and the service for carrying on a shoulder pole standby node, complete current connection request and return the result, then communicated with host node,
Continue to be managed the node in cluster.
The above-mentioned function for each node in server cluster system, in order to illustrate node each in server cluster system
Load-balancing function realization, Fig. 6 shows the load balancing according to an embodiment of the present invention based on server cluster system
Implementation method process.
As shown in fig. 6, the method that the load balancing provided by the invention based on server cluster system is realized, server set
Group's system includes host node, standby node and ordinary node.Wherein,
S610: coupled standby node is received to host node cycle, the load information that each ordinary node reports.
Wherein, it should be noted that create a nodelist in the master node, the load information of each node is stored in
In nodelist, load information includes: cpu busy percentage, memory usage, the network bandwidth for each node being connected with host node
Occupancy and socket connection occupancy.
S620: host node obtains the load value of each node corresponding with load information according to the load information reported.
Before subscriber terminal requests connection, judge with the presence or absence of offline node in nodelist, it is offline if it exists
Node then deletes this node from nodelist；Offline node is standby node if it exists, then again in ordinary node
New standby node is selected, and carries out synchronizing information with host node；If offline node is not present in nodelist, according to institute
The connection request for stating client terminal is attached distribution.
S630: when there is client terminal connection request, host node is attached distribution by HTTP redirection, wherein main
Node selects the node of optimal processing ability according to the load value of each node of acquisition, the distribution node as connection.
Wherein, it is provided with session management mechanism in the master node, session management mechanism includes client terminal and group system
Between conversation history record, and be responsible for deleting between client terminal and host node that sluggish historical session records for a long time,
Make the node optimal selection of the connection distribution of selection.
This is because client terminal with connect the node of distribution for a long time without business contact, it is each in group system
The load of node can change, and to guarantee that the node of service connection distribution is the node of optimal processing ability, therefore session be remembered
Record is deleted, and the node of new connection distribution is reselected.The setting of session management mechanism further decrease distribution connection when
Prolong, guarantees the reliability and timeliness of business between intelligent terminal and group system.
S640: when host node breaks down, standby node carries out active-standby switch, becomes new host node, executes main section
The responsibility of point, while enabling new standby node.
S650: new standby node is reported to load information in new host node, and same with the information of new host node
Step, while ordinary node is screened, select the alternative of new standby node.
Above-mentioned steps S640 and step S650 is the workflow of active-standby switch.When host node breaks down, spare section
Point is how to replace host node to carry out fulfiling host node responsibility.Host node and the united back mechanism of standby node, Neng Goushi
Standby node can quickly become new host node when existing host node failure, and server cluster system still is able to when something goes wrong can
Steadily to provide business service.
Host node is described more fully below and goes out detailed process of the node of optimal processing ability as the distribution node connected.
The present invention is used and is backed up to host node in server cluster system, this back mechanism and reversed proxy server
And the load balancing of network address change technology realization is compared and is had the following advantages: on the one hand, the main section in server cluster system
Point does not need to undertake application layer work as Reverse Proxy；On the other hand, do not need yet as network address transmission that
Sample foundation is bi-directionally connected, therefore reduces the workload of host node, individual node failure risk is reduced, but also active and standby realization
Become possible, further increases cluster stability.The work that portal server (i.e. host node) connection distribution is described below is former
Reason and respective algorithms.
It is well known that after HTTP request reaches server-side, being handled by server-side, and return in http protocol
HTTP is responded, this response can carry HTTP return code, and each return code has contained different meanings, as shown in table 1.
Table 1
| HTTP return code | Meaning |
| 1** | Retain |
| 2** | Indicate that request is successfully received |
| 3** | It redirects |
| 4** | Client terminal mistake |
| 5** | Server-side mistake |
Wherein, show that network site has been replaced in requested service with the HTTP return code of 3 beginnings, need client whole
End end re-initiates request, and new chained address will return together in HTTP is responded.
For host node using 307 in HTTP return code, meaning is to be redirected using interim, shows that requested service is faced
When more preferable position.Had using the benefit temporarily redirected: desired network address can be returned in HTTP response, then allowed
Client terminal initiates the connection request to new address；Secondly as being temporarily to redirect, client terminal is asked in the new connection of initiation
When asking, the address after redirection will not be saved, when so as to each connection request, host node can all receive client first
The connection request of terminal is attached distribution further according to the loading condition of each node, ensure that the realization of clustering design function.
Fig. 7 shows the principle process of client terminal according to an embodiment of the present invention Yu group system connection request process.
In the present embodiment, using specific vehicle-mounted client as client terminal.
As shown in fig. 7, vehicle-mounted client sends connection request to host node, host node is obtained most by connection distribution algorithms
The ordinary node C of the connection distribution of excellent processing capacity, vehicle-mounted client are general by the optimal processing ability returned according to host node
The information of logical node C sets up professional chief connection to ordinary node C, and then vehicle-mounted client and ordinary node C start at business
Reason.
In the present invention, host node is attached distribution by HTTP redirection, is attached point by HTTP redirection
Hair has following benefit: host node only needs to carry out the processing of a HTTP request, not to offer service is applied, also without internal
Connection is established, the process resource of host node is dramatically saved, alleviates the burden of host node；Simultaneously as HTTP redirection
The new address carried is the address of the internal node by being calculated, and has unpredictability, and function is equal to network visit
Used identifying code when asking can shield the ineffective access initiated by automatic machine to a certain extent, therefore have anti-DDoS
The ability of network attack improves the safety of system entirety.
It is that the most important thing (that is: selects optimal processing ability to the selection for loading the smallest node in connection distribution algorithms
Node), can be not only related to connection request quickly be handled, be also related to the stability of group system entirety, below
Load minimum node selection algorithm will be described.
Due to will create a nodelist on the primary node, real-time update every loading condition of each node, often
The data four-tuple collected on a node, specially CPUUsage, MemUsage, NetworkUsage and
SocketUsage respectively indicates the consumption level of CPU on node, memory, network bandwidth and Socket connection quantity.According to even
Connection type is divided into CPU intensive type, memory-intensive and bandwidth intensive by the difference for connecing request.Therefore, added by specific
The load computational algorithm of power, each node will calculate separately its processing capacity to different connection requests, as follows:
rl=f (cu,mu,nu,su)
Wherein, rlIndicate the surplus resources percentage of node l；When remaining cpu resource is bigger, show its work that can be undertaken
It measures more, then more connections can be distributed to respective nodes up.cu、mu、nu、suRespectively indicate the node received
Cpu busy percentage, memory usage, network bandwidth occupancy and the socket connection occupancy of upload, value is bigger, shows corresponding
Resource occupation it is bigger, node load is heavier.According to the corresponding data received, final surplus resources percentage is calculated,
Its formula is shown below:
Wherein, rl(k) the surplus resources percentage at k moment is indicated；cu(k) k moment CPU usage is indicated；cu(k-1) table
Show the utilization rate of (k-1) moment CPU；The CPU impact factor considers two moment, estimates CPU using the method for discreet derivative
Change curve.
In (1-tu(k)-tconst) in, as t=s, tu(k) the socket connection utilization rate at k moment, t are indicatedconstIt indicates
The pre- allowance of socket resource, (1-tu(k)-tconst) indicate the remaining available socket resource situation of node t moment；Work as t
When=m, tu(k) memory usage at k moment, t are indicatedconstIndicate the pre- allowance of memory source, (1-tu(k)-tconst) indicate
The remaining available memory source situation of node t moment；As t=n, tu(k) the network bandwidth occupancy at k moment is indicated,
tconstIndicate the pre- allowance of network bandwidth, (1-tu(k)-tconst) indicate the remaining available network bandwidth resources of node t moment
Situation.
It should be noted that under normal circumstances, the pre- allowance of socket resource and network broadband resource is all 0, show this
The utilization rate of resource can reach 100%, and correspondingly, the pre- allowance of memory source cannot be 0, show the use of memory source
Rate cannot reach 100%, need to retain certain space.
That is, the surplus resources percentage r at k momentl(k) it indicates, respectively by cu、mu、nu、su4 influences
The influence of the factor, each impact factor are different to its influence mode；And each impact factor is to its influence mode
Different, if one of impact factor reaches capacity, if cpu surplus is 0, then calculated resource end value will become
0, then the node will not obtain resource allocation, ensure that the correct response of the stability and access request of the node.In 4 shadows
It rings in the factor, according to the different characteristics of every class resource, corresponding influence mode is also different.
By taking socket is connected as an example, 1-su(k)-sconstIndicate the impact factor of socket connection；suIt (k) is the k moment
Socket connection occupancy；sconstFor the pre- allowance of socket resource.Under normal circumstances, socket resource and network resource
Pre- allowance be all 0, show that the utilization rate of these resources can reach 100%, and correspondingly, the pre- allowance of Mem resource cannot
It is 0, shows that the utilization rate of memory source cannot reach 100%, this is the normal operation in order to guarantee node, is required to as memory
Retain certain space, in order to avoid node is caused to crash.
Wherein, it should be noted that in the calculating of impact factor, more special is the calculating of cpu resource, from system
For level, cpu resource is very valuable resource, while being the resource of utilization rate moment variation again, can neither be with simple pre-
The mode of spacing wastes cpu resource, and the curve that improves of CPU must be taken into consideration, while also needing to consider the complexity calculated, with
Just the computing resource of host node is saved in periodically a large amount of calculating.
In conclusion estimating the change curve of CPU using the method for discreet derivative, and leave a blank as the pre- of CPU
Between consider, when ascendant trend is presented in CPU curve, first derivative is greater than 0, is used, obtains using pre- allowance as CPU
1-2c outu(k)+cu(k-1) impact factor calculation formula, conversely, when downward trend is presented in CPU usage curve, single order is led
Number not reserve utilization rate space since CPU will obtain more surpluses for negative value for it at this time.It is pre- using curvilinear path
The CPU usage amount of survey method estimation is not only able to satisfy the estimation requirement of CPU usage, while can utmostly improve Shangdi
The utilization rate of CPU, lifting system overall efficiency have received good effect.
In order to further illustrate how to realize load balancing between the node of server cluster system, Fig. 8 shows root
According to the workflow of the implementation of load balancing of the server cluster system of the embodiment of the present invention.In the present embodiment, with tool
The vehicle-mounted client of body is as client terminal.
As shown in figure 8, the implementation of load balancing of server cluster system shown in the present embodiment, including walk as follows
It is rapid:
S801: start；
S802: host node whether there is failure? failure if it exists executes step S803；Failure if it does not exist executes step
S808；
S803: active-standby switch, standby node become new host node；
S804: new standby node enables；
S805: standby node information reporting；
S806: with host node synchronizing information；
S807: standby node carries out ordinary node screening；After the completion of active-standby switch, the responsibility of host node is fulfiled；It executes again
Step S808；
S808: host node cycle receives other node load information reportings；
S809: host node calculates the load value of each node according to the loading condition reported；
It is there not line node in S810:nodelist? if there is not online node, step S811 is executed；If no
Not online node executes step S815；
Does is S811: not online node standby node? if standby node, step S812 is executed；If it is not, executing
Step S814；
S812: this node is removed from nodelist；
S813: it is synchronous to reselect new standby node row information of going forward side by side, then executes step S815；
S814: removing this node from nodelist, then executes step S815；
S815: when there is vehicle-mounted client connection request, select the node of optimal processing ability as the section of connection distribution
Point；
S816: the node for the connection distribution that vehicle-mounted client and host node are selected starts to carry out operation flow；
S817: terminate.
The process of one specific embodiment of the above-mentioned implementation of load balancing for server cluster system.To sum up institute
State, the present invention for intelligent terminal high concurrent long connection the characteristics of, propose this long connection, high concurrent server-side load balancing
Implementation method and server cluster system and connection distribution algorithms, server cluster system of the invention use netted networking knot
Structure, and have the progressive relationship of hierarchy between multiple nodes, on the one hand service stability, another party are improved on the net from group
Face, which is conducive to best performance node in election the whole network, becomes standby node, to become host node under host node failure conditions, protects
The election of Zhang Liao global optimum.
Also, active/standby backup mechanism is used in cluster networking mode, and the seamless switching between active and standby may be implemented, it is spare
Node is capable of the function of hosting node, shields Single Point of Faliure, improves cluster and externally provides the availability of business.
Again, the connection distribution algorithms proposed select the node of optimal processing ability to receive service request, can be maximum
The processing capacity using each node of limit, simultaneous session administrative mechanism can further decrease time delay when distribution connection,
The reliability and timeliness of guarantee business.Therefore, the present invention proposes the possibility that the method that load balancing is realized has been widely used
Property, even more important effect will be played in all kinds of similar projects.
Describe the server cluster system proposed according to the present invention and its load in an illustrative manner above with reference to attached drawing
Balance realizing method.It will be understood by those skilled in the art, however, that the server cluster system proposed for aforementioned present invention
And its implementation of load balancing, various improvement can also be made on the basis of not departing from the content of present invention.Therefore, of the invention
Protection scope should be determined by the content of appended claims.
Claims (9)
1. a kind of server cluster system, including host node, standby node and ordinary node；Wherein,
The host node includes information collection module, load computing module and connection distribution decision-making module；
The information collection module, for storing the standby node being connected with the host node and that each ordinary node reports is negative
Information carrying breath；
The load computing module, the load information for reporting according to obtain section corresponding with the load information
The load value of point；
Decision-making module is distributed in the connection, for being attached distribution by HTTP redirection, wherein is calculated according to the load
The load value for the node that module obtains, selects the node of optimal processing ability, the node as connection distribution；
The standby node, for reporting the load information of standby node, and the information with the host node to the host node
It is synchronous, and the ordinary node is screened, the ordinary node selected is as new standby node；Wherein, described standby
It include standby node information reporting module, synchronizing information module and ordinary node screening module with node；
The standby node information reporting module, for reporting load information to the host node cycle；
The synchronizing information module for periodically synchronizing the load information in the information collection module, and is stored；
The ordinary node screening module, for selecting to load most in ordinary node when the standby node becomes host node
It is alternative that small node becomes it；
The ordinary node, for reporting the load information of ordinary node to the host node, and when it becomes standby node
When, it is switched to standby node；Wherein, the ordinary node includes ordinary node information reporting module and standby node switching
Module；
The ordinary node information reporting module, for reporting load information to the host node cycle；
The standby node switching module will be to going out selected by it for after the standby node becomes new host node
New standby node is given notice, and ordinary node is switched to standby node accordingly, and undertakes the responsibility of standby node.
2. server cluster system as described in claim 1, wherein
The host node further includes management conversation module, for the conversation history between managing customer terminal and the group system
Record, and it is responsible for deleting between the client terminal and the group system sluggish historical session record for a long time.
3. server cluster system as described in claim 1, wherein
A nodelist is created in the information collection module, the nodelist is connected for storing with the host node
The load information that reports of each node；Wherein,
The nodelist includes update mechanism, and the update mechanism is for deleting offline node.
4. server cluster system as claimed in claim 3, wherein
Load information includes: that cpu busy percentage, memory usage, the network bandwidth for each node being connected with the host node account for
With rate and socket connection occupancy；
In the load computing module, to each node by the load computational algorithms of particular weights, obtain respectively its for
CPU intensive type, memory-intensive, network bandwidth be intensive and the load capacity of socket intensive request, and the knot that will acquire
Fruit updates to the nodelist.
5. a kind of implementation of load balancing based on server cluster system, the server cluster system include host node,
Standby node and ordinary node, wherein
Receive to the host node cycle coupled standby node, the load information that each ordinary node reports, and root
According to the load information reported, the load value of each node corresponding with the load information is obtained；
When there is client terminal connection request, the host node is attached distribution by HTTP redirection, wherein the main section
Point selects node of the node of optimal processing ability as connection distribution according to the load value of each node of acquisition；
When the host node breaks down, the standby node carries out active-standby switch, becomes new host node, executes host node
Responsibility, while enabling new standby node；Wherein, the standby node includes that standby node information reporting module, information are same
Walk module and ordinary node screening module；The standby node information reporting module, for being reported to the host node cycle
Load information；The synchronizing information module for periodically synchronizing the load information in the host node, and is stored；
The ordinary node screening module, for selecting to load in ordinary node the smallest when the standby node becomes host node
It is alternative that node becomes it；
The new standby node is reported to load information in the new host node, and same with the information of new host node
Step, while the ordinary node is screened, select the alternative of the new standby node；Wherein, the ordinary node packet
Include ordinary node information reporting module and standby node switching module；The ordinary node information reporting module is used for described
Host node cycle reports load information；The standby node switching module, for becoming new main section when the standby node
After point, it will give notice to the new standby node gone out selected by it, ordinary node is switched to standby node accordingly, and undertakes standby
With the responsibility of node.
6. as claimed in claim 5 based on the implementation of load balancing of server cluster system, wherein in the host node
In be provided with session management mechanism,
The session management mechanism includes the conversation history record between the client terminal and the group system, and is responsible for deleting
Except sluggish historical session records for a long time between the client terminal and the host node, make the section of the connection distribution of selection
Point is optimal selection.
7. as claimed in claim 5 based on the implementation of load balancing of server cluster system, wherein in the host node
One nodelist of middle creation, the load information of each node is stored in the nodelist, in the subscriber terminal requests
Before connection,
Judge in the nodelist with the presence or absence of offline node, if it exists offline node, then by this node from described
It is deleted in nodelist；Offline node is standby node if it exists, then selects in the ordinary node again new spare
Node, and synchronizing information is carried out with the host node；
If offline node is not present in the nodelist, it is attached point according to the connection request of the client terminal
Hair.
8. as claimed in claim 5 based on the implementation of load balancing of server cluster system, wherein
The load information includes: cpu busy percentage, memory usage, the Netowrk tape for each node being connected with the host node
Wide occupancy and socket connection occupancy；
By the load computational algorithm of particular weights, each node calculates separately it for CPU intensive type, memory-intensive, net
The load capacity of network bandwidth intensive and socket intensive request, formula are as follows:
rl=f (cu,mu,nu,su)
Wherein, rlIndicate the surplus resources percentage of node l；
cu、mu、nu、suRespectively indicate cpu busy percentage, memory usage that the node that receives uploads, network bandwidth occupancy with
And socket connection occupancy.
9. as claimed in claim 8 based on the implementation of load balancing of server cluster system, wherein
The load information uploaded according to each node is received, calculates surplus resources percentage, formula is shown below:
Wherein, rl(k) the surplus resources percentage at k moment is indicated；
cu(k) k moment CPU usage is indicated；
cu(k-1) utilization rate of (k-1) moment CPU is indicated；
In (1-tu(k)-tconst) in, as t=s, tu(k) the socket connection utilization rate at k moment, t are indicatedconstIt indicates
The pre- allowance of socket resource, (1-tu(k)-tconst) indicate the remaining available socket resource situation of node t moment；
As t=m, tu(k) memory usage at k moment, t are indicatedconstIndicate the pre- allowance of memory source, (1-tu(k)-
tconst) indicate the remaining available memory source situation of node t moment；
As t=n, tu(k) the network bandwidth occupancy at k moment, t are indicatedconstIndicate the pre- allowance of network bandwidth, (1-tu(k)-
tconst) indicate the remaining available network bandwidth resources situation of node t moment.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410512754.5A CN104283948B (en) | 2014-09-26 | 2014-09-26 | Server cluster system and its implementation of load balancing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410512754.5A CN104283948B (en) | 2014-09-26 | 2014-09-26 | Server cluster system and its implementation of load balancing |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104283948A CN104283948A (en) | 2015-01-14 |
| CN104283948B true CN104283948B (en) | 2018-12-07 |
Family
ID=52258421
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410512754.5A Active CN104283948B (en) | 2014-09-26 | 2014-09-26 | Server cluster system and its implementation of load balancing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104283948B (en) |
Families Citing this family (47)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105989279A (en) * | 2015-02-13 | 2016-10-05 | 上海通用识别技术研究所 | Domino dedicated password cracking system based on multi-hierarchy GPU (Graphic Process Unit) cluster |
| CN104768196B (en) * | 2015-04-21 | 2019-03-15 | 南京先锋硕通无线技术有限公司 | The method of cluster group seamless switching |
| CN107615734B (en) * | 2015-05-26 | 2021-01-26 | 爱唯思有限公司 | System and method for server failover and load balancing |
| CN105007233B (en) * | 2015-07-13 | 2018-02-27 | 互联网域名系统北京市工程研究中心有限公司 | A kind of method that distribution address is loaded based on Dynamic Host Configuration Protocol server cluster |
| CN106453120B (en) * | 2015-08-05 | 2019-06-07 | 北京网御星云信息技术有限公司 | A kind of dynamic cluster method and system |
| CN105338078B (en) * | 2015-10-26 | 2019-01-01 | 北京百度网讯科技有限公司 | Date storage method and device for storage system |
| CN105516245B (en) * | 2015-11-25 | 2018-12-14 | 国家计算机网络与信息安全管理中心 | A kind of system and implementation method of the load balancing based on stream |
| CN105577759A (en) * | 2015-12-15 | 2016-05-11 | 东软熙康健康科技有限公司 | Server node allocation method and device |
| CN106933675A (en) * | 2015-12-31 | 2017-07-07 | 华为技术有限公司 | A kind of method of process task, manager, server and system |
| CN105430105B (en) * | 2016-01-05 | 2019-01-18 | 北京卓识达软件有限公司 | A kind of data processing method and system of car-mounted terminal |
| CN106060123B (en) * | 2016-05-20 | 2019-08-16 | 深圳市永兴元科技股份有限公司 | The collecting method and distributed data system of distributed data system |
| CN106027649B (en) * | 2016-05-20 | 2019-08-16 | 深圳市永兴元科技股份有限公司 | The collecting method and distributed data system of distributed data system |
| CN106716974B (en) * | 2016-07-28 | 2020-08-04 | 深圳前海达闼云端智能科技有限公司 | Access distribution method, device and system |
| CN107689878A (en) * | 2016-08-04 | 2018-02-13 | 创盛视联数码科技（北京）有限公司 | TCP length connection SiteServer LBSs based on name scheduling |
| CN106331755B (en) * | 2016-08-17 | 2019-10-08 | 广州酷狗计算机科技有限公司 | The method and apparatus for handling Media Stream Caton |
| CN107800551A (en) * | 2016-08-31 | 2018-03-13 | 北京优朋普乐科技有限公司 | Redis group systems and its method, the client for improving reliability |
| CN106549796A (en) * | 2016-09-27 | 2017-03-29 | 努比亚技术有限公司 | Resource control method and host node that a kind of firmware space is downloaded |
| CN107145406A (en) * | 2017-05-14 | 2017-09-08 | 四川盛世天成信息技术有限公司 | A kind of disaster-tolerant backup method and system based on Clustering |
| CN107172193A (en) * | 2017-06-20 | 2017-09-15 | 郑州云海信息技术有限公司 | A kind of load-balancing method and its device based on cluster |
| CN107257298A (en) * | 2017-07-27 | 2017-10-17 | 郑州云海信息技术有限公司 | A kind of fault handling method and device |
| CN107453909A (en) * | 2017-08-08 | 2017-12-08 | 北京盛华安信息技术有限公司 | The processing method of security incident |
| CN107592215A (en) * | 2017-08-31 | 2018-01-16 | 郑州云海信息技术有限公司 | A kind of method for managing message-oriented middleware |
| CN107682406B (en) * | 2017-09-08 | 2020-08-25 | 北京三快在线科技有限公司 | Method, device and system for processing service |
| CN107689925B (en) * | 2017-09-28 | 2020-01-14 | 平安科技（深圳）有限公司 | Load balancing optimization method and device based on cloud monitoring |
| CN107948260A (en) * | 2017-11-15 | 2018-04-20 | 郑州云海信息技术有限公司 | Main monitoring node selecting method and device in a kind of distributed type assemblies |
| CN108763312B (en) * | 2018-04-26 | 2021-07-06 | 大连理工大学 | Slave data node screening method based on load |
| CN110708179B (en) * | 2018-07-10 | 2022-06-07 | 福建省天奕网络科技有限公司 | Block chain-based data communication bridging method and storage medium |
| CN109474465A (en) * | 2018-11-13 | 2019-03-15 | 上海英方软件股份有限公司 | A kind of method and system of the high availability that can dynamically circulate based on server cluster |
| CN109634821A (en) * | 2018-11-13 | 2019-04-16 | 中科院微电子研究所昆山分所 | A kind of node availability detection method and device |
| CN109672728A (en) * | 2018-12-06 | 2019-04-23 | 航天恒星科技有限公司 | A kind of large-scale distributed clustered node information fast acquiring method |
| CN109728941A (en) * | 2018-12-20 | 2019-05-07 | 北京比新科技有限公司 | A kind of block chain leader election method and device thereof |
| CN109951331B (en) * | 2019-03-15 | 2021-08-20 | 北京百度网讯科技有限公司 | Method, device and computing cluster for sending information |
| CN110086726A (en) * | 2019-04-22 | 2019-08-02 | 航天云网科技发展有限责任公司 | A method of automatically switching Kubernetes host node |
| CN111400026B (en) * | 2019-11-15 | 2023-02-28 | 河海大学 | Distributed load balancing method based on master-slave backup technology |
| CN111343254B (en) * | 2020-02-17 | 2022-11-11 | 天津卓朗科技发展有限公司 | Client connection method and device and electronic equipment |
| CN111541753B (en) * | 2020-04-16 | 2024-02-27 | 深圳市迅雷网络技术有限公司 | Distributed storage system, method, computer device and medium for block chain data |
| CN111641522A (en) * | 2020-05-25 | 2020-09-08 | 杭州安恒信息技术股份有限公司 | Method, system and computer equipment for node switching |
| CN111565233A (en) * | 2020-05-28 | 2020-08-21 | 吉林亿联银行股份有限公司 | Data transmission method and device |
| CN112190924A (en) * | 2020-12-04 | 2021-01-08 | 腾讯科技（深圳）有限公司 | Data disaster tolerance method, device and computer readable medium |
| CN113156803A (en) * | 2021-02-03 | 2021-07-23 | 南京华鹞信息科技有限公司 | Task-oriented unmanned aerial vehicle cluster resource management and fault-tolerant control method |
| CN113472566A (en) * | 2021-06-11 | 2021-10-01 | 北京市大数据中心 | Status monitoring method of union block chain and master node status monitoring system |
| CN113950165B (en) * | 2021-12-03 | 2022-11-08 | 中电信数智科技有限公司 | Intelligent networking platform networking equipment connection method and device |
| CN114338727B (en) * | 2022-01-05 | 2024-03-12 | 中国建设银行股份有限公司 | Service cluster construction method and related device |
| CN114338696B (en) * | 2022-03-14 | 2022-07-15 | 北京奥星贝斯科技有限公司 | Method and device for distributed system |
| CN114978871B (en) * | 2022-04-08 | 2023-10-24 | 北京航天飞行控制中心 | Node switching method and node switching device of service system and electronic equipment |
| CN115001854B (en) * | 2022-07-18 | 2022-11-22 | 江苏艾盾网络科技有限公司 | Big data-based tracing-prevention server cluster management and control system and method |
| CN116112493A (en) * | 2023-02-09 | 2023-05-12 | 网易（杭州）网络有限公司 | Communication method, device, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1549978A (en) * | 2001-07-16 | 2004-11-24 | BEAϵͳ��˾ | Method and apparatus for session replication and failover |
| CN101521679A (en) * | 2009-04-03 | 2009-09-02 | 南京邮电大学 | Self-organizing method based on composite structured peer-to-peer network |
| CN101605092A (en) * | 2009-07-10 | 2009-12-16 | 浪潮电子信息产业股份有限公司 | A kind of content-based SiteServer LBS |
| CN102387218A (en) * | 2011-11-24 | 2012-03-21 | 浪潮电子信息产业股份有限公司 | Multimachine hot standby load balance system for computer |
| CN103117876A (en) * | 2013-01-24 | 2013-05-22 | 中兴通讯股份有限公司 | User state information synchronizing method and device |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100488146C (en) * | 2006-09-14 | 2009-05-13 | 华为技术有限公司 | Method for establishing point-to-point collection in P2P network and nodes in P2P network |
-
2014
- 2014-09-26 CN CN201410512754.5A patent/CN104283948B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1549978A (en) * | 2001-07-16 | 2004-11-24 | BEAϵͳ��˾ | Method and apparatus for session replication and failover |
| CN101521679A (en) * | 2009-04-03 | 2009-09-02 | 南京邮电大学 | Self-organizing method based on composite structured peer-to-peer network |
| CN101605092A (en) * | 2009-07-10 | 2009-12-16 | 浪潮电子信息产业股份有限公司 | A kind of content-based SiteServer LBS |
| CN102387218A (en) * | 2011-11-24 | 2012-03-21 | 浪潮电子信息产业股份有限公司 | Multimachine hot standby load balance system for computer |
| CN103117876A (en) * | 2013-01-24 | 2013-05-22 | 中兴通讯股份有限公司 | User state information synchronizing method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104283948A (en) | 2015-01-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104283948B (en) | Server cluster system and its implementation of load balancing | |
| CN107852604B (en) | System for providing Global Virtual Network (GVN) | |
| CN106209490B (en) | Select and monitor the method and system of multiple service key performance indicators | |
| US11463511B2 (en) | Model-based load balancing for network data plane | |
| CN103338150B (en) | Communication network architecture method for building up, device, server and router | |
| JP6883427B2 (en) | Diameter load and overload information and methods for virtualization, system and computer readable media | |
| CN103947156B (en) | Method, apparatus and communication network for root cause analysis | |
| CN102045363B (en) | Establishment, identification control method and device for network flow characteristic identification rule | |
| CN109600768A (en) | Management method, equipment and the system of network slice | |
| CN109067579A (en) | Towards the 5G network slice topology design of bottom layer node failure and reliable mapping method | |
| Huang et al. | Software-defined QoS provisioning for fog computing advanced wireless sensor networks | |
| CN104885431A (en) | Content based traffic engineering in software defined information centric networks | |
| JP7092813B2 (en) | Packet transmission method and equipment | |
| CN102780779A (en) | Gateway equipment and method and device for optimization of campus network export P2P (peer-to-peer) traffic | |
| CN104852859B (en) | A kind of aggregation interface method for processing business and equipment | |
| CN108880888A (en) | A kind of SDN network method for predicting based on deep learning | |
| CN112311896B (en) | Health examination method, device, equipment and computer readable storage medium | |
| CN106161053A (en) | A kind of SDN controller QoS manages system and method | |
| CN106533973A (en) | Method and system for distributing service message, and equipment | |
| CN110855424B (en) | Method and device for synthesizing asymmetric flow xDR in DPI field | |
| CN105391651A (en) | Virtual optical network multilayer resource convergence method and system | |
| CN109005126A (en) | The processing method and equipment of data flow | |
| CN102594606A (en) | Evolved intelligent node overlapping network system based on multi-agent | |
| CN106357723A (en) | Synchronous system and method for multi-cluster information caching based on cloud host | |
| CN108650337B (en) | Server detection method, system and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |