CN103491019A

CN103491019A - Mixed algorithm for network load balancing

Info

Publication number: CN103491019A
Application number: CN201310429836.9A
Authority: CN
Inventors: 王雷; 赵玉龙; 王宇
Original assignee: Beihang University
Current assignee: Beihang University
Priority date: 2013-09-18
Filing date: 2013-09-18
Publication date: 2014-01-01

Abstract

Network load balancing plays an important role in a computer network, and is built on the basis of an existing network structure. A cheap, effective and transparent method is provided for expanding the bandwidth of network equipment and a server, increasing throughput capacity, enhancing network data processing capacity and improving the flexibility and usability of a network. An existing load balancing algorithm comprises round-robin, Hash dispatching based on streams and FS dispatching based on stream fragmentations. The invention provides a mixed and dispatching algorithm for the network load balancing. Mixed objects are a Hash algorithm and an FS algorithm. According to the core concept, on the basis of the FS algorithm, an FS table with a configurable size is introduced and used for storing five tuples of flow already appearing and the time of the last data package.

Description

A kind of hybrid algorithm of Network Load Balance

Technical field

The present invention relates to computer network, be specially load-balancing algorithm.

Background technology

Network Load Balance tool in computer network plays a very important role, load balancing is based upon on existing network infrastructure, it provide the effectively transparent method extended network equipment of a kind of cheapness and server bandwidth, increase throughput, Strengthens network data-handling capacity, improve flexibility and the availability of network.Load balancing comprises scheduling and forwarding two parts, as shown in Figure 1.Scheduling is the key of load balancing, and forwarding engine is responsible for according to purpose IP Packet Generation to next stage equipment.

Initial dispatching algorithm is simple Round-robin, presses the order of packet, issues successively the forwarding engine of different numberings.Because Round-robin is operated in the aspect of packet, while causing packet to arrive purpose equipment, there will be out of order phenomenon, also destroyed the locality of network flow simultaneously.Afterwards, a kind of based on the flow(data flow) hash algorithm of aspect becomes main flow.Flow refers to have the packet of identical 5 tuples, and 5 tuples refer to source IP, purpose IP, source port, destination interface and protocol type.Key using 5 tuples as the Hash function, the result obtained is exactly the forwarding engine that will send.The drawback of this method is, when the data packet number in a flow is very large, will cause certain forwarding engine load overweight.Tsing-Hua University the time of heap of stone (transliteration), the people such as Liu Bin (transliteration) find by research, in network packet, even if the packet in same flow neither continuously send.If the time interval between two packets in flow has surpassed certain critical value, just think that they belong to different flow slice, at this moment just they should be assigned on different forwarding engines, Here it is based on flow-slice(FS) load-balancing algorithm, as shown in Figure 2.

Summary of the invention

The present invention proposes a kind of load balance scheduling algorithm of mixing, mix here to as if hash algorithm and flow-slice algorithm.Core concept is on the basis of flow-slice algorithm, to introduce a configurable flow-slice table of size (FS table), 5 tuples of the flow occurred in order to storage and the time of a upper packet.

The difference of the present invention and FS is:

1.FS, in scheduling, by the corresponding flow information of hash table storage, for guaranteeing save data as much as possible, this table is very large.FS table in the present invention is an adjustable table of size, and size is generally 500-1000.

2.FS in scheduling, the definite of new flow-slice forwarding engine ID is with overall round-robin, uses the hash algorithm in the present invention.

3.FS in scheduling, the overall round-robin of definite use of the forwarding engine ID of the different flow slice of same flow, the present invention uses round-robin to each flow.

Idiographic flow of the present invention as shown in Figure 3.

The accompanying drawing explanation

Fig. 1 is responsible for the balance dispatching model;

Fig. 2 FS dispatching algorithm flow process;

Fig. 3 mixed scheduling algorithm flow process.

Embodiment

The first step, obtain 5 tuples and time

Read source IP and purpose IP from the IP header of network packet network layer, then read the transport layer protocol type, according to transport layer protocol (tcp or udp) type, read source port and destination interface.So just obtain 5 complete tuples.The time of advent of record data bag.

Second step, the forwarding engine that the specified data bag mails to

This process can be described as:

2.1 judge that 5 tuples whether in FS table, if so, turn 2.2 steps, otherwise turn 2.5 steps.

2.2 the time of recording in calculating current data packet and FS table is poor, whether surpasses the time interval of agreement, if fruit is to turn 2.3, otherwise turn 2.4.

2.3 determine new forwarding engine by round-robin, revise forwarding engine and time.

2.4 send out packet to the forwarding engine recorded in FS table.

2.5 judge that whether FS table is full, if do not have, turn 2.6 steps, otherwise turn 2.7 steps.

2.6 5 tuples are used to Hash, determine the forwarding engine that it initially mails to.By 5 tuples, time and forwarding engine insert in FS table.

There is no at most 2.7 find out the data item a used, calculate time poor of current data packet and a, if, over the designated time interval, turn 2.8, otherwise turn 2.9

2.8 5 tuples are used to Hash, determine the forwarding engine that it initially mails to.Use 5 tuples, time and forwarding engine are inserted and are replaced a data item.

2.9 5 tuples are used to Hash, determine the forwarding engine that it mails to.

Claims

1. the hybrid algorithm of a Network Load Balance, is characterized in that, in conjunction with existing Hash scheduling and FS scheduling.Be specially, on the basis of flow-slice algorithm, introduce a configurable flow-slice table of size (FS table), 5 tuples of the flow occurred in order to storage and the time of a upper packet.

2. the hybrid algorithm of a kind of Network Load Balance according to claim 1, is characterized in that, FS table is an adjustable table of size, and size is generally 500-1000.

3. the hybrid algorithm of a kind of Network Load Balance according to claim 1, is characterized in that, the definite of new flow-slice forwarding engine ID is to use the hash algorithm.

4. the hybrid algorithm of a kind of Network Load Balance according to claim 1, is characterized in that, each flow that determines of the forwarding engine ID of the different flow slice of same flow is used round-robin.