CN100534070C - A method and system for implementing application server load balancing - Google Patents

Abstract

The invention discloses a method to realize application server, comprising the following steps: a domain name analyzing server receives a request with application server cluster domain name sent by a user, and assigns an application server with an IP address mark to the user from the pre-stored application server IP address list corresponding to the domain name; the application server assigned to the user locates an application server for the user according to the current load information of each application server in the application server cluster stored in a meta-data base server, and feedbacks the application server information to the user. The located server stores the file data uploaded by the user in the storage device at the network side during the communication with the user. The invention can realize direct P2P data transmission, balance distribution operations, and make full use of every application server resource in the application server cluster.

Description

A method and system for implementing application server load balancing

technical field

The invention relates to the technical field of network storage, in particular to a method and system for realizing load balancing of application servers.

Background technique

The WebDisk service is mainly used for file upload and download. In order to provide better scalability and high availability, the application servers in the WebDisk system are usually deployed in clusters to achieve load balancing of the application servers. However, some existing load technologies have some problems when dealing with WebDisk services.

For example, the IP Virtual Server (IPVS) software in Linux Virtual Server (LVS, Linux Virtual Server), the three load balancing technologies used in it cannot completely solve the load problem of the load balancing scheduler (Load Balance Box) when uploading and downloading. The LVS project provides an LVS framework for implementing scalable network services, as shown in Figure 1. In this LVS framework, the load balancing scheduler is responsible for job scheduling, so that the load of the scheduled back-end real servers can be balanced; the real servers are used for Under the scheduling of the load balancing scheduler, it receives the request message and the feedback response message sent by the user to the user, and completes the upload and download of data. In the LVS architecture, the working principles and defects of the three methods of IP load balancing are as follows:

1.Virtual Server via Network Address Translation(VS/NAT)

Through network address translation, the load balancing scheduler rewrites the target address of the request message sent by the user, and distributes the request message to the real server at the backend according to the preset scheduling algorithm; the response message of the real server is dispatched through load balancing When the server, the source address of the message is rewritten, and then returned to the client to complete the entire load scheduling process.

The disadvantage of this method is that whether uploading or downloading, the data must pass through the load balancing scheduler. Obviously, it is a hot device in the system. In a large data upload and download system like WebDisk, it is easy to become a performance bottleneck.

2. Virtual Server via IP Tunneling (VS/TUN)

When NAT technology is used, both the request message sent by the user and the response message fed back by the real server must be rewritten by the load balancing scheduler, so that when the user requests more and more, the processing capacity of the load balancing scheduler will become bottleneck. In order to solve this problem, the load balancing scheduler forwards the request message to the real server through IP Tunneling, and the real server returns the response message directly to the user, so that the load balancing scheduler only needs to process the request message. In general, the network service response is much larger than the request message, so after using VS/TUN technology, the maximum throughput of the cluster system can be increased by 10 times.

The disadvantage of this method is that although the load balancing scheduler is bypassed when downloading data, the data must still pass through the load balancing scheduler when uploading data. It is still a hot device in the system when uploading data, so there is still a bottleneck question.

3.Virtual Server via Direct Routing(VS/DR)

The VS/DR method sends the request message to the real server by rewriting the MAC address of the request message sent by the user, and the real server directly returns the feedback message to the user. Like the above-mentioned VS/TUN method, the VS/DR technology can greatly improve the scalability of the cluster system. This method does not have the overhead of IP tunneling, and the real servers in the cluster do not have to support the IP tunneling protocol, but it requires that the load balancing scheduler and the real servers have a network card connected to the same physical network segment.

Compared with IP Tunneling, this method reduces the overhead of IP tunneling, but it still does not solve the bottleneck problem of hotspot devices when uploading data.

Contents of the invention

The present invention provides a method and system for implementing load balancing of application servers, which are used to solve the problem in the prior art that a load balancing scheduler tends to become a data transmission bottleneck when data is uploaded or downloaded.

The invention provides a system for realizing load balancing of application servers, including:

an application server cluster comprising at least one application server;

The metadata database server stores the current load information of each application server in the application server cluster;

The domain name resolution server stores the corresponding relationship between the domain name of the application server cluster and the IP address of the application server in the cluster, and is used for when the user accesses the application server cluster based on the domain name of the application server cluster, according to the stored corresponding relationship: The user assigns an application server identified by an IP address in the application server cluster;

The application server assigned to the user in the application server cluster locates an application server for the user according to the current load information of each application server stored in the metadata database server, and feeds back the application server information to the user;

A storage device, configured to store file data uploaded by the user during communication with the located application server.

The present invention provides a method for realizing application server load balancing, comprising steps:

The domain name resolution server receives the request from the user carrying the domain name of the application server cluster, and

Assign an application server identified by an IP address to the user in the pre-stored list of application server IP addresses corresponding to the domain name;

The application server assigned to the user locates an application server for the user according to the current load information of each application server in the application server cluster stored in the metadata database server, and feeds back the application server information to the user;

During the process of communicating with the user, the located server saves the file data uploaded by the user in the storage device on the network side.

The present invention eliminates the bottleneck problem of the hotspot device of the load balancing scheduler, and realizes direct, point-to-point data transmission; at the same time, it can distribute jobs in a balanced manner according to the load conditions of each application server in the application server cluster, making full use of the application server cluster. of each application server resource.

Description of drawings

Fig. 1 is the framework of Linux Virtual Server in the prior art;

Fig. 2 is the general structural diagram of WebDisk system in the embodiment of the present invention;

FIG. 3 is a flowchart of a WebDisk load balancing service in an embodiment of the present invention.

Detailed ways

The present invention will be described in detail below in conjunction with the accompanying drawings.

As shown in FIG. 2 , the WebDisk system proposed by the present invention includes: an application server cluster 201 , a domain name resolution server 202 , a metadata database server 203 and a storage device 204 .

Application server cluster 201, such as application server 1 in the figure ... application server n constitutes application server cluster 201, and all application servers can be used as load balancing schedulers, responsible for job scheduling; domain name resolution server (DNS) 202 , storing the corresponding relationship between the domain name of the application server cluster and all or part of the IP addresses of the application servers in the cluster, for when the user accesses the application server cluster based on the domain name of the application server cluster, in a round-robin (Round Robin) manner In the application server cluster 201, an application server identified by an IP address is assigned to the user as a load balancing scheduler; the metadata database server 203 is used to store metadata in the WebDisk system, and the metadata includes the application server cluster 201 The current load situation of each application server, wherein the current load situation of each application server can be reported by the corresponding application server periodically or according to other conditions, and the data stored in the metadata database server 203 can be shared by all application servers and distributed to users The application server as the load balancing scheduler obtains the application server with the lightest current load as the real communication server configured to the user by querying the current load situation of each application server stored in the metadata database server 203; the storage device 204, It is used to store business data (mainly uploaded file data) in the WebDisk system, which is shared by all application servers. It can be various DAS, NAS, and SAN storage devices, which are used to store users and communicate with the configured real communication server , uploaded file data, etc.

The application server cluster is a large, virtual server consisting of multiple application servers. Users access the cluster through the same domain name. In this way, the number of application servers is transparent to users, that is, users only feel A virtual large server exists, and it is not known how many application servers there are.

Each application server in the application server cluster 201 reports its current load status according to a certain strategy (such as time interval strategy, load change strategy, etc.), and records the load status information in the metadata database server 203 of the system. For example, if the load of the application server changes and exceeds a certain load change threshold, it can report and update its own load status information; no matter whether the load of the application server has changed, the application server must be within the specified time interval Report its own load status once; if the application server has not reported its own load status beyond the specified time interval, the system will regard the status of this application server as unavailable, and the server as the load balancing scheduler will no longer report to the It assigns jobs.

The WebDisk system load balancing business process that the present invention proposes is shown in Figure 3, comprises the following steps:

Step 1. The user initiates a request, and the request carries the domain name of the application server cluster accessed by the user;

Step 2, the domain name resolution server (DNS) assigns an application server IP to it in a round-robin (RoundRobin) manner according to the domain name provided by the user;

In this step, the domain name of the application server cluster and the IP list of each application server in the cluster corresponding to the domain name and the IP list of each application server in the cluster (at least one application server IP, and at most all application server IPs) are stored in DNS. These registered application servers It will act as a load balancing scheduler and be responsible for job scheduling; DNS locates client requests to an application server identified by an IP address in the IP list in a round-robin manner. For example, the domain name of the application server cluster is webdisk.com.cn, and the IP addresses of the application servers in the cluster corresponding to the domain name are IP1, IP2, and IP3. These three application servers will act as load balancing schedulers; DNS will The user request is located on the application server identified by an IP address among IP1, IP2 and IP3. Each server in the cluster is both an application server and a load balancing scheduler.

Step 3, establishing a communication connection between the client and the application server located above;

Steps 4.1-4.2, the located application server obtains the latest load status table of each application server from the metadata database, and finds out the application server with the lightest load;

Step 5, the located application server feeds back the information of the found application server with the lightest load to the client;

Step 6: The user end relocates the job to the application server with the lightest load, performs data transmission, and saves the file data uploaded by the user in the storage device 204 on the network side.

In Figure 3, the dotted line is the control information flow, and the amount of information data in this part is relatively small; the solid line is the actual business data flow, which refers to uploading and downloading file data in the WebDisk business, and the data volume is very large.

Since the bottleneck problem of the hotspot device of the load balancing scheduler is eliminated, the present invention realizes direct, point-to-point data transmission; resources of each application server, thus simplifying the WebDisk system, improving the efficiency of data upload and download, and facilitating the realization of load balancing.

Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these modifications and variations.

Claims (8)

1, a kind of system that realizes application server load balancing is characterized in that, comprising:

The application server cluster that includes at least one application server;

The metadata database server stores the current load information of each application server in the described application server cluster;

Domain name resolution server, store the corresponding relation of described application server cluster domain name and this middle application server IP address of trooping, being used for when the user visits this application server cluster based on described application server cluster domain name, is the user distributes an IP address designation in application server cluster application server according to the described corresponding relation of storing;

Be assigned to user's application server in the application server cluster, according to the current load information of storing in the described metadata database server of each application server, for the user locatees an application server, and this Application Server information is fed back to the user;

Memory device is used for storing the user and the described application server that is positioned communicates the file data that process is uploaded.

2, the system as claimed in claim 1 is characterized in that, reports the metadata database server by the load information that self is current of each application server in the described application server cluster and stores.

3, system as claimed in claim 2 is characterized in that, described each application server regularly reports the metadata database server with the load information of self; Or

Each when the load information of self changes, self load information is reported the metadata database server.

4, the system as claimed in claim 1 is characterized in that, the domain name resolution server distributes an IP address designation in the wheel mode of crying in application server cluster application server is given the user.

5, the system as claimed in claim 1, it is characterized in that, the described application server that is assigned to the user is according to the current load information of storing in the metadata database server of each application server, for the user locatees the lightest application server of present load.

6, a kind of method that realizes application server load balancing is characterized in that, comprises step:

Domain name resolution server receives the request that carries application server cluster domain name that the user sends, and

Storage in advance with this domain name corresponding application server IP address list in distribute the application server of an IP address designation to give this user;

Be assigned to user's application server,,, and this Application Server information fed back to the user for this user locatees an application server according to the current load information of each application server in the application server cluster of storing in the metadata database server;

Described located server is communicating in the process with described user, and the document data saving that the user is uploaded is in the memory device of network side.

7, method as claimed in claim 6 is characterized in that, the described application server load information current according to each application server that is assigned to the user is for the user locatees the lightest application server of present load.

8, method as claimed in claim 6 is characterized in that, the domain name resolution server is mode with wheel, storage in advance with this domain name corresponding application server IP address list in distribute the application server of an IP address designation to this user.

Images