WO2009059456A1

WO2009059456A1 - Method for controlling load balance of network system, client, server and network system

Info

Publication number: WO2009059456A1
Application number: PCT/CN2007/003146
Authority: WO
Inventors: Hongguang Sun
Original assignee: Lucent Technologies Inc.
Priority date: 2007-11-06
Filing date: 2007-11-06
Publication date: 2009-05-14
Also published as: CN101849384A

Abstract

A method for controlling load balance in a network system, a client, a server, and a network system is provided. The method for controlling load balance in a network system that comprises a client cluster and a server cluster comprises the steps of: detecting a load condition of at least one server of the server cluster to determine a corresponding load level; transmitting a multicast message including the load level to the client cluster by multicast at the server; and determining a server to which a service request is to be sent at a client of the client cluster based on the multicast message. The load balance can therefore be performed at a client without dedicated load balance equipment, thus decreasing the risk of single point invalidation and simplifying the framework of the network system.

Description

METHOD FOR CONTROLLING LOAD BALANCE OF NETWORK SYSTEM, CLIENT, SERVER AND NETWORK SYSTEM

TECHNICAL FIELD OF THE INVENTION

The present invention relates to network communication technology, and particularly to a method for controlling load balance in a network system, a client, a server and a network system.

BACKGROUND OF THE INVENTION

Along with the development of network technology, the amount of network users is increased sharply. The demands of the users are satisfied by using a server cluster to provide an enormous amount of tasks and services to a lot of clients. The processing and communication activities can be distributed evenly to the servers in the server cluster by adopting load balancing technology among computer networks so as to extend the bandwidth of the servers, increase their throughputs, speed up the response of the servers, improve the utilization efficiency of the servers and other resources, and provide better access quality to the users.

Fig. 1 is a functional block diagram of an existing load balance system. As shown in Fig. 1, the existing load balance system comprises a client cluster 10 composed of a plurality of clients 100, 101 and etc., a load balance device 40 such as a load balancer, and a server cluster 20 composed of a plurality of servers 200, 201 and etc. The load balance device 40 is connected to the server cluster 20. A service request transmitted from a client 100 in the client cluster 10 is sent to the load balance device 40 through a network 30, e.g. an LAN or VLAN such as an Ethernet and a WLAN, and a WAN such as the Internet. The load balance device 40 selects a server from the server cluster 20 according to a special distribution algorithm such as polling, weighted polling, minimum connection number, fastest connection number and address hash algorithm, etc. The load balance device 40 transmits the IP address and port information of a selected server 201 to the client, thus the server 201 establishes a connection with the client 100 to process the service request from the client.

However, when the amount of the service requests increases to a certain extent, a bottleneck effect may occur on a single load balance device, and balancing devices by using a special device complicates the existing network frame and increases network maintenance cost. SUMMARY OF THE INVENTION

To solve the above problem in the prior art, the invention provides a method for controlling load balance in a network system, a client, a server and a network system.

According to an aspect of the present invention, there is provided a method for controlling load balance in a network system which comprises a client cluster and a server client, the method comprising the steps of: detecting a load condition of at least one server in the server cluster to determine a load level of the at least one server; transmitting a multicast message including the load level to the client cluster by multicasting at the server; and determining a server to which a service request is to be sent at a client of the client cluster based on the multicast message.

According to another aspect of the present invention, there is provided a client configured to receive multicast messages from a plurality of servers, comprising: a determining means for determining a server to which a service request is to be sent based on the multicast messages, wherein the multicast messages include corresponding load levels of the servers.

According to still another aspect of the present invention, there is provided a server configured to transmit a multicast message to the multicast group of a client cluster, comprising a detecting means for detecting a load condition of the server to determine its load level; and a multicast means for transmitting a multicast message including the load level to the client cluster.

According to still another aspect of the present invention, there is provided a network system comprising the above mentioned server and client.

According to still another aspect of the present invention, there is provided a program product comprising computer executable instructions which, when being executed on a computer device, makes the device to be configured to execute the above mentioned method.

BRIEF DESCRIPTION OF THE DRAWINGS

It is believed that the purposes, features and advantages of the present invention will be better understood from the following description of the detailed implementation of the present invention taken in conjunction with the accompanying drawings, in which:

Fig. 1 is a functional block diagram of an existing load balance network system;

Fig. 2 shows a functional block diagram of a load balance network system according to an embodiment of the present invention;

Fig. 3 shows a flowchart of a method for controlling load balance in a network system according to an embodiment of the present invention;

Fig. 4 shows a schematic diagram of a channel map according to an embodiment of the present invention;

Fig. 5 shows a flowchart of a method of load balance of a server according to another embodiment of the present invention;

Fig. 6 shows a flowchart of a method of load balance of a client according to still another embodiment of the present invention;

Fig. 7 shows a flowchart of updating at a client a load list of candidate servers or a channel map according to still another embodiment of the present invention;

Fig. 8 shows a functional block diagram of a server according to still another embodiment of the present invention; and

Fig. 9 shows a functional block diagram of a client according to still another embodiment of the present invention.

DETAILED DESCRIPTION OF EMBODIMENTS OF THE INVENTION

Next, various embodiments of the present invention will be described in detail in conjunction with accompany drawings.

According to an embodiment of the present invention, when there appear a lot of service requests, a new process can be used to control load balance in a network system. The basic concept is that the load conditions of one or more servers in a server cluster are multicasted to the clients in a client cluster using multicast technology, and a server in the server cluster is selected according to a special algorithm at a client, thus distributing the risk at a single node.

Fig. 2 shows a functional block diagram of a load balance network system according to an embodiment of the present invention. As shown in Fig. 2, the load balance system comprises a client cluster 10 composed of a plurality of clients 100, 101 and 102, etc. and a server cluster 20 composed of a plurality of servers 200, 201 and 202, etc. The client cluster 10 is a multicast group whose multicast IP address can be Addl, and the server cluster 20 can be another multicast group whose multicast IP address is Add2. A network 30 can be an Ethernet network, or other LANs or WANs supporting multicast. The client 100 in the client cluster 10 transmits a service request to a server in the server cluster 20 such as server 201 selected according to a special algorithm through the network 30 such as an Ethernet network.

Fig. 3 is a flowchart of a method for controlling load balance according to an embodiment of the present invention. As shown in Fig. 3, in step S301, the server 201 detects its own load condition which is for example categorized into different load levels such as light, moderate, high, extremely high, blocked or down according to a CPU, a memory and the value of a bandwidth as well as the tolerable response time of a system. The types of load levels are known to all hosts (including clients and servers). But each server has its own criteria for determining a load level. For example, the server 200 defines 95% of CPU/Memory occupation is of "extreme high load"; the server 201 is shared for other services and thus defines 50% of CPU/Memory occupation is of "extreme high load"; and the server 202 is of poor performance and thus defines 80% of CPU/Memory occupation is of "extremely high load".

Thereafter in step S302, the server 201 transmits a multicast message including its load level to the multicast group of the client cluster 10 whose multicast IP address is Addl. In the embodiment of the present invention, it is not necessary for the clients to know the specific capability of each server, but the load level of each server within the multicast messages from the servers. Of course, the multicast message can further include the specific attributes of the server or a timer duration. For a certain application, the information of a certain server can be customized, such as a timer duration, a server attribute, a load level and load level category criteria, etc., thus making load balance flexible. Specifically, by obtaining the attributes of corresponding servers from a multicast message, a client can take off the servers which do not have the attributes when maintaining a load list of candidate servers. For example, the server 200 can control a G.726 codec while the others can not, thus the attribute will be presented in the multicast message. An initial tag can be appended to the server attributes, which will be used by a client. If the initial tag keeps unchanged, then the client can suppose that the server attributes are unchanged, thus the client can ignore server attributes processing to save CPU time. For example, the pseudo codes are as follows:

<Load>l</Load> // Moderate Load

</Attribute> </server> Finally, in step S303, the client 100 in the client cluster 10 receives the multicast message, and when there appears a new service request in a request queue, selects a suitable server to transmit a service request. The server can be selected according to a conventional polling, weighted polling, minimum connection number, fastest connection number or address hash algorithm, etc. Of course, a load list of candidate servers can be maintained at the client, and as shown in table 1, the list can comprise the corresponding load level, attributes or IP address of a list to be selected to map to the server.

Table 1 Load List of candidate servers

Alternatively, the load list of candidate servers can also be a channel map, in which as shown in Fig. 4 each channel maps a server in a server cluster that accept service requests. The channel map contains related information of a target server obtained from the multicast message such as load level, server attributes, IP address, etc. In this case, the selection of a server by a client becomes the selection of a corresponding channel. Specifically, if a certain channel is selected to transmit a service request, then the request will be transmitted to the IP address of the target server. In the channel selection process, lighter load levels and preferred server attributes will be of higher priority.

It can be understood that in the embodiments of the present invention one or more servers in the server cluster can detect the load levels and transmit the multicast message, the principle is similar to that of the above embodiment.

It will be described in detail how a server in the server cluster 20 detects a load condition.

Fig. 5 is a flowchart of a method of load balance of a server according to an embodiment of the present invention. As shown in Fig. 5, the server can periodically detects its load condition, or detect its load condition based on a command such as a request, or detect the load condition under combination with the both. In step S507, the server detects whether the timer duration expires, and if so, then in step S508, it detects a load condition to determine a load level according to its own criteria. Next in step S509, the load levels, server attributes and other related information of the servers will be multicasted to the clients belonging to a same multicast group. In step S510, the timer is reset so as to start a next cycle. Referring to Fig. 5, the server can also notify the client when the load condition changes. In step S502, the server waits for a service request from the client. When the server receives the service request, it begins to process the request in step S503. Next, in step S504, the server detects the load condition and determines the load level according to its load criteria. In step S505, the server detects whether the current load level changes, and if so (step S505: Yes), continues to process the service request in step S506, but in the meantime, the server multicasts to all the clients in a same multicast group the new load levels of the server(step S509). If the current load level does not change (step S505: No), then the server continues to process in step S506. In the request processing course, the server adopts a first-one-being-responsible mechanism, i.e., regardless of the variation of the load condition of the server, when a client transmits a service request, it will be processed on the server, and even if the another server has a lighter load, the service request will not be transferred to the other server. Thus, a service response will not be shuffled between different servers, ensuring the smooth processing of the service request.

A "eager upgrade/lazy downgrade" mechanism may be introduced in the process. That is, when the load condition changes to make the load higher (for example, the load level changes from a lighter level to a higher level or a blocked level, which is also referred to as load upgrade in the present specification), a server in the server cluster 20 will multicast this load upgrade message as soon as possible so that the overload condition information of the server can reach the clients in time. The server in the server cluster will multicast the load downgrade information to notify the clients of the load lightening only after a certain interval when load lightening occurs (for example, the load level of the server changes from a higher level to a moderate level or a lighter level, which is also called load downgrade in the present specification). Any load level recovery during that interval will stop the load downgrade information multicasting. By using this mechanism, the overflow can be prevented from occurring, and an over fast variation of the server load condition under a critical condition can be prevented from occurring, thus lowering the burden on the system and the network.

Fig. 6 shows a flowchart of a method of load balance of a client according to another embodiment of the present invention. As shown in Fig. 6, in step S602, it is detected whether there exists a new service request in a queue.

When there appears a new service request, in step S603, the client will select the suitable server or channel based on a load list of candidate servers or a channel map according to a special algorithm such as a polling algorithm. The principle is the same as mentioned above, and will not be further described.

The load list of candidate servers or the channel map is updated periodically or based on the variation of parameters related to servers such as the load condition. As shown in Fig. 7, in step S701, a client in a client cluster monitors a multicast message at a client multicast IP address. In step S702, the client receives the message and obtains the up to date information of the parameters related to the servers such as the load condition. Next in step S703, the client updates the load list of candidate servers or the channel map according to such received parameter information as a server load level, for example it adds a new server option, deletes an existing sever option or updates an existing server option. For the channel map, it inserts a new channel, deletes an existing channel or updates the information for an existing channel.

Next, in step S604, the client transmits a service request to a selected server.

Under the same inventive concept, Fig. 8 shows a functional block diagram of a server for controlling load balance in a network system according to an embodiment of the present invention. The embodiment will be described in detail below in conjunction with Fig. 8.

As shown in Fig. 8, a server 200 in the server cluster 20 which controls load balance of the system comprises a detecting means 820 for detecting the load condition of the server to determine its load level; and a multicast means 821 for transmitting a multicast message including the load level to a client cluster.

The detecting means 820 can periodically detect the load condition of the server 200 to determine its load level, and can also detect its load condition based on a command such as a request to determine a corresponding load level or detect combination with foregoing load condition.

The multicast means 821 transmits a multicast message including information related to the server 200 such as the load level, server attributes and a timer duration to a client cluster 10 belonging to a same multicast group, and each client belonging to a multicast IP address monitors and receives the multicast message.

The multicast means 821 can be configured to, when the load condition changes to make the load higher (for example, the load level changes from a lighter level to a higher level or a blocked level, which is also referred to as load upgrade in the present specification), multicast this load upgrade message as soon as possible so that the overload condition information of the server 200 can reach the clients in time. The multicast means 821 can also be configured to multicast the load downgrade information to notify the clients of the load lightening only after a certain interval when load lightening occurs (for example, the load level of the server changes from a higher level to a moderate level or a lighter level, which is also called load downgrade).

The detecting means 820 and the multicast means 821 of the server for controlling load balance in a network system of the present embodiment can be implemented by a hardware circuitry such as an SLSI or a gate array, a semiconductor device such as a logical chip, a transistor, etc., or a programmable device such as a field programmable gate array, a programmable logic device ,etc, it can also be implemented by software executed by various types of processors, and can also implemented by the combination of the above mentioned hardware circuitry and software. The server for controlling load balance in a network system of the present embodiment implements in operation the method for controlling load balance in a network system at a server according to the embodiment as shown in Fig. 5.

Fig. 9 is a functional block diagram of a client according to another embodiment of the present invention. As shown in Fig. 9, a client 100 in a client cluster 10 of the present embodiment comprises a determining means 910 for determining a server to transmit a service request according to a multicast message.

The client 100 further comprises a maintaining means 911 for maintaining a load list of candidate servers or a channel map. The maintaining means 911 can be separated from the determining means 910, or integrated into the determining means 910.

The determining means 910 of the client 100 can determine a sever according to a conventional polling, weighted polling, minimum connection number, fastest connection number and address hash algorithm, etc. The determining means 910 can further select a sever to which a service request is to be transmitted by referring to the load list of candidate servers or channel map provided by the maintaining means 911.

The maintaining means 911 can maintain a load list of candidate servers or a channel map corresponding to the load list of candidate servers according to information related to the server such as the load level, server attributes and etc. The maintaining means can be configured to update the list or the channel map by performing the following operations on the load list of candidate servers or the channel map:

-add a new server option (channel);

-delete an existing sever option (channel); and/ or

-update an existing server option (channel).

The determining means 910 and the maintaining means 911 of the client for controlling load balance in a network system of the present embodiment can be implemented by a hardware circuitry such as an SLSI or a gate array, a semiconductor device such as a logical chip, a transistor, etc., or a programmable device such as a field programmable gate array, a programmable logic device, etc., it can also be implemented by software executed by various types of processors, or combination of the above mentioned hardware circuitry and software. The client for controlling load balance in a network system of the present embodiment implements in operation the method for controlling load balance in a network system at a server according to the embodiment as shown in Fig. 6.

In the embodiment of the present invention, since the clients in the client cluster 10 can distribute service requests to the selected servers to achieve load balance in a network without a third party responsible for the distribution of services, the structure of the network system is simplified and the risk that a single point is invalid is distributed. And in this structure, the server only has its load condition detected and is not used to achieve load balance, thus simplifying its structure.

In addition, because of the adoption of the multicast technology, it becomes very easy to add or delete a server in a server cluster during operation. When a server is to be added into a server cluster, it only needs to added it into a server multicast group, and transmit its information to the IP address of the client multicast group. When a server needs to quit, it may transmit the information to the IP address of the client multicast group appropriately. If a server quits exceptionally, the client can obtain its condition by failing to detect its multicast message periodically or by failing to send a service request to the server. Correspondingly, the client can update the load list of candidate servers or the channel map to delete the server.

In the network of the embodiment of the present invention, the monitoring of the whole system can be conducted simply by just transmitting a request to a client or server multicast IP address to reach each host in the domain.

Further, in the network system of the embodiment of the present invention, because of adopting multicast and being based on a frame without a third party, testing on a specific server in the server cluster can be facilitated by defining certain special server attribute, which will direct all traffic to the specific server without shutting down other servers, thus simplifying the maintenance and management of the network system.

While the method for controlling load balance in a network system, the client, the server and the network system of the present invention are described in detail in conjunction with specific embodiments, the present invention is not limited to these embodiments. It can be understood by a person skilled in the art that various changes, alternatives and modifications can be made without departing from the spirit and scope of the present invention, defined by the attached claims.

Claims

1. A method for controlling load balance in a network system that comprises a client cluster and a server cluster, comprising of the steps of: detecting a load condition of at least one server in the server cluster to determine a load level of the at least one server; transmitting a multicast message including the load level to the client cluster by multicast at the server; and determining a server to which a service request is to be sent at a client of the client cluster based on the multicast message.

2. The method for controlling load balance according to claim 1, wherein the load level comprises light, moderate, high, extremely high, blocked or down.

3. The method for controlling load balance according to claim 1 or 2, wherein the multicast message further comprises a server attribute and/or a timer duration for broadcast.

4. The method for controlling load balance according to any one of the preceding claims, wherein the step of detecting the load condition comprises detecting at least one of the following: the load condition of the server periodically and based on a service request.

5. The method for controlling load balance according to claim 3, wherein the step of determining a server to which a service request is to be sent comprises the step of: determining a server based on the load level and/or server attribute contained in the multicast message.

6. The method for controlling load balance according to claim 1, wherein the method further comprises the step of: maintaining a load list of candidate servers at the client.

7. The method for controlling load balance according to claim 6, wherein the step of determining a server to which a service request is to be sent comprises the step of: determining a server by means of the load list of candidate servers.

8. The method for controlling load balance according to claim 6, wherein the step of maintaining the load list of candidate servers comprises the step of: updating the load list of candidate servers at the client based on the received multicast message.

9. The method for controlling load balance according to claim 8, wherein the step of updating the load list of candidate servers comprises the step of: adding a new server option, deleting an existing server option or updating an existing server option in the load list of candidate servers.

10. The method for controlling load balance according to any one of claims 6 to 9, wherein the load list of candidate servers comprises a channel map.

11. The method for controlling load balance according to any one of claims 6 to 9, wherein the step of transmitting multicast message comprises transmitting at least one of the follow: the multicast message at the server of the server cluster as soon as possible if the load of the server becomes high and only after a certain interval when the server detects the load is lightened.

12. A client configured to receive multicast messages from a plurality of servers, comprising: determining means, for determining a server to which a service request is to be sent based on the multicast messages, wherein the multicast messages include corresponding load levels of the servers.

13. The client according to claim 12, wherein the multicast messages further comprise server attributes and/or a timer duration for broadcast.

14. The client according to claim 12, wherein the client further comprises: maintaining means, for maintaining a load list of candidate servers.

15. The client according to claim 14, wherein the determining means is configured to determine a server by means of the load list of candidate servers.

16. The client according to claim 14, wherein the maintaining means is configured to update the load list of candidate servers according to the multicast messages.

17. The client according to claim 16, wherein the maintaining means is configured to update the load list of candidate servers by adding a new server option, deleting an existing server option or updating an existing server option in the load list of candidate servers.

18. The client according to any one of claims 14 to 17, wherein the load list of candidate servers comprises a channel map.

19. A server configured to send a multicast message to a multicast group of a client cluster, comprising: detecting means, for detecting a load condition of the server to determine a corresponding load level; and multicast means, for transmitting a multicast message including the load level to the client cluster.

20. The server according to claim 19, wherein the load level comprises light, moderate, high, extremely high, blocked or down.

21. The server according to claim 19 or 20, wherein the multicast message further comprises a server attribute and timer duration for broadcast.

22. The server according to any one of claims 19 to 20, wherein the detecting means is configured to detect at least one of the follow: the load condition of the server periodically and based on a service request.

23. The server according to any one of claims 19 to 20, wherein the multicast means is configured to send at least one of the following: the multicast message as soon as possible if the load of the server becomes high and only after a certain interval when the server detects the load is lightened.

24. A network system, comprising: the client of any one of claims 12 to 18 and the server of any one of claim 19 to 23.

25. A program product, comprising computer executable instructions which, when being executed on a computer device, makes the device to be configured to execute the method according to any of claims 1 to 11.