JP2004336498A - Packet switch cluster configuration method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a packet switch cluster configuration method which configures a packet switch cluster which has higher performance than simple blocking switches and is more inexpensive than large-sized non-blocking switches. <P>SOLUTION: In the packet switch cluster configuration method for connecting a plurality of packet switches into a cluster, an entire traffic flow is recognized and is grouped on the basis of its localities. The grouped traffic flow is concentrated on a specific packet switch. An optical cross connect is used to concentrate the grouped traffic flow on the specific packet switch. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

・・・・・・・・・・・・・・・・・ （１）
ここで、あるポートｊ（１≦ｊ≦６）について、下記（２）式のように定義する。
【００１２】
【数２】

・・・・・・・・・・・・・・・・・ （２）
【００１３】
図４に、ｉ番目のノンブロッキングスイッチ２のポート１（ｊ＝１）における、ｔｏ＿ｌｏｃａｌ、ｔｏ＿ｒｅｍｏｔｅ、ｆｒｏｍ＿ｌｏｃａｌ、および、ｆｒｏｍ＿ｒｅｍｏｔｅの位置を示す。
また、トラフィックグルーピングを行うための閾値となる転送量をＴｔｈｒｅｓｈとする。Ｔｔｈｒｅｓｈは、例えば、回線使用率１０％に相当する転送量に設定する。
前述のステップ１３では、以下の処理を実行する。
Ｔｊ，ｔｏｔａｌの多い順にソートし、Ｔｊ，ｔｏｔａｌ＞Ｔｔｈｒｅｓｈを満たすものについて順に以下の判断をしていく。
（ａ）Ｔｊ，ｔｏ＿ｌｏｃａｌ＜Ｔｊ，ｔｏ＿ｒｅｍｏｔｅ、かつ、Ｔｊ，ｆｒｏｍ＿ｌｏｃａｌ＜Ｔｊ，ｆｒｏｍ＿ｒｅｍｏｔｅである場合、ポートｊを移動対象とする。
（ｂ）Ｔｊ，ｔｏ＿ｌｏｃａｌ＞Ｔｊ，ｔｏ＿ｒｅｍｏｔｅ、かつ、Ｔｊ，ｆｒｏｍ＿ｌｏｃａｌ＜Ｔｊ，ｆｒｏｍ＿ｒｅｍｏｔｅである場合、ポートｊを移動対象としない。
（ｃ）Ｔｊ，ｔｏ＿ｌｏｃａｌ＜Ｔｊ，ｔｏ＿ｒｅｍｏｔｅ、かつ、Ｔｊ，ｆｒｏｍ＿ｌｏｃａｌ＞Ｔｊ，ｆｒｏｍ＿ｒｅｍｏｔｅである場合、Ｔｊ，ｆｒｏｍ＿ｌｏｃａｌ＜Ｔｊ，ｔｏ＿ｒｅｍｏｔｅであれば、ポートｊを移動対象とする。
（ｄ）Ｔｊ，ｔｏ＿ｌｏｃａｌ＞Ｔｊ，ｔｏ＿ｒｅｍｏｔｅ、かつ、Ｔｊ，ｆｒｏｍ＿ｌｏｃａｌ＞Ｔｊ，ｆｒｏｍ＿ｒｅｍｏｔｅである場合、ポートｊを移動対象としない。
また、Ｔｊ，ｔｏｔａｌ＜Ｔｔｈｒｅｓｈであるポートには、「移動可能」のフラグを立てておく。
【００１４】
次に、前述のステップ１４では、以下の処理を実行する。
各ノンブロッキングスイッチ２で、「移動対象」となったポートのＴｊ，ｔｏ−ｒｅｍｏｔｅを構成するトラフィックフローの宛先アドレスを調べ、それらがもっとも多く属するノンブロッキングスイッチ２を移動先として選択する。この処理を、全ノンブロッキングスイッチ２に付いて実行する。
次に、前述のステップ１５では、以下の処理を実行する。
トラフィック切替え装置１を操作して順にポートの移動を実行する。
即ち、ｉ番目のノンブロッキングスイッチ２の「移動対象」ポートｊを、移動先のｎ番目のノンブロッキングスイッチ２の「移動可能」ポートの中で最も通信量の少ないポートｍと交換する。
この際、「移動対象」ポートｊに関連付けられていたＭＡＣアドレスと、「移動可能」ポートｍに関連付けられていたＭＡＣアドレスのエントリを全スイッチから消去する。
なお、前述の説明では、一度に一組ずつ交換する場合について説明したが、ｋれに限らず、例えば、Ｔｊ，ｔｏｔａｌの一番多いものを２ポート、一番少ないものを２ポートずつ取り出して「移動対象」、「移動可能」の組みを２組づつとしてもよい。
最後に、ステップ１６で、次の制御ループに入るまで指定時間待機する。
【００１５】
このように、本実施の形態では、各ノンブロッキングスイッチ２が、モジュール間リンク回線３の使用率を下げるようにポート収容の交換を行う。
これにより、全体のトラフィックのローカリティに対して適応的に組み替えが行われるため、各ノンブロッキングスイッチの使用率を高め、設備を効果的に使用することが可能となる。
特に、インターネットのように、トラフィックのローカリティが大きく、そのローカリティが予測不能な場合、従来のように単一の大型ノンブロッキングスイッチを設置することなく、小型のモジュール型スイッチを必要に応じて増減して使用することができるので、効率的な設備計画が可能となる。
以上、本発明者によってなされた発明を、前記実施の形態に基づき具体的に説明したが、本発明は、前記実施の形態に限定されるものではなく、その要旨を逸脱しない範囲において種々変更可能であることは勿論である。
【００１６】
【発明の効果】
本願において開示される発明のうち代表的なものによって得られる効果を簡単に説明すれば、下記の通りである。
本発明のパケットスイッチクラスタ構成方法によれば、単純なブロッキングスイッチよりも性能が良く、大型のノンブロッキングスイッチに比べ安価なシステムを構成することが可能となる。
【図面の簡単な説明】
【図１】本発明に基づくスイッチクラスタの最も基本的なタイプ（直列型）を示す模式図である。
【図２】本発明に基づくスイッチクラスタの最も基本的なタイプ（ツリー型）を示す模式図である。
【図３】図１、図２に示す切替え装置コントローラの処理手順を示すフローチャートである。
【図４】図３で生成されるトラフィックフロー行列の一例を示す図である。
【符号の説明】
１…トラフィック切替え装置、２…ノンブロッキングスイッチ、３…モジュール間リンク回線、４…ユーザ収容リンク回線、５…制御インターフェース、６…替え装置コントローラ。[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a configuration method of a switch cluster for increasing the capacity by combining a plurality of packet switches (Ethernet switches, FDDI switches, etc.) used for constructing a packet network represented by the Internet. The present invention relates to a method for configuring a switch cluster combining optical switches.
[0002]
[Prior art]
In a packet network, a packet is transferred to a destination by a packet switch. As the packet switch, an electric switch using a semiconductor and an optical switch using an optical device have been developed, and the electric switch is generally widely used. A typical example is an Ethernet switch, and products with various internal exchange capacities and external transmission speeds are commercially available.
There are two types of packet switches, blocking switches and non-blocking switches. In the former case, the switching capacity is not sufficient for the sum of the port accommodation bands, and there is a possibility that packets may be lost in the switching operation.
The latter has a sufficient switching capacity exceeding the sum of the port accommodating bands, and does not cause packet loss in switching.
A large-sized packet switch having a large number of accommodation ports is provided with a high-speed and complicated internal switching network in order to maintain the non-blocking property, but the unit cost of the port increases accordingly.
On the other hand, in a small packet switch in which the number of accommodated ports is about 8 to 48 or less, a non-blocking switch can be configured at a relatively low cost by applying an ASIC (Application Specific Integrated Circuit) technology.
[0003]
Prior art documents related to the present invention include the following.
[Non-patent document 1]
Cisco Systems, Cisco IOS Netflow Technology Dataset,
http: // www. cisco. com / warp / public / cc / pd / iosw / prodlit / iosnf_ds. htm
[Non-patent document 2]
S. Panchen, et. Al. InMon Corporation's sFlow: A Method for Monitoring Traffic in Switched and Routed Networks, IETF, RFC3176.
[Non-Patent Document 3]
P. Amsden, et. Al. Cabletron's Light-weight Flow Admission Protocol Specification Version 1.0, IETF, RFC2124. JP-A-4-000508
[Problems to be solved by the invention]
However, in the prior art, when accommodating a large number of ports of hundreds to thousands, there is a trade-off between cost and performance.
That is, if a non-blocking switch is configured with emphasis on performance, the cost increases. On the other hand, if a blocking switch is configured with emphasis on cost, the performance is not sufficient.
SUMMARY OF THE INVENTION The present invention has been made to solve the problems of the prior art, and an object of the present invention is to provide a packet switch cluster configuration method which has better performance than a simple blocking switch and is less expensive than a large non-blocking switch. Is to provide.
The above and other objects and novel features of the present invention will become apparent from the description of the present specification and the accompanying drawings.
[0005]
[Means for Solving the Problems]
The following is a brief description of an outline of typical inventions disclosed in the present application.
The present invention is characterized in that small-sized packet switches having non-blocking properties at low cost are combined in a cluster, traffic between ports is grouped, and connection is switched so as to fit in one housing of these small-sized packet switches.
In other words, in the present invention, by combining an inexpensive non-blocking switch and adjusting the flow of traffic so as to make the most of the non-blocking property, it operates as a non-blocking switch when the traffic has locality even though it is an inexpensive blocking switch. Let it.
[0006]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
In all the drawings for describing the embodiments, components having the same function are denoted by the same reference numerals, and repeated description thereof will be omitted.
[Basic concept of the present invention]
FIG. 1 is a schematic diagram showing the most basic type (series type) of a switch cluster according to the present invention. The configuration of the module is not limited to the serial type, and may be a tree structure as shown in FIG. 2 described later.
The operation of the switch cluster according to the present embodiment will be described with reference to FIG.
The components of the switch cluster shown in FIG. 1 include a traffic switching device 1, a plurality of non-blocking switches 2 interconnected in series, and a switching device controller that collects bandwidth usage information from each switch and controls the traffic switching device. 6; the inter-module link line 3 for connecting them; the user accommodation link line 4;
The user device is connected to the user accommodation link, and the internal connection of the traffic switching device 1 in the initial state is one-to-one with the port of the non-blocking switch as shown in FIG.
The capacity of the inter-module link 3 is small compared to the maximum value of the traffic that can occur.
For this reason, a cheaper system can be constructed as compared with a single non-blocking switch. However, when this device is viewed from the user accommodating link side, non-blocking properties are not guaranteed, and data may be lost due to congestion of the inter-module link 3. There is. That is, in the initial state, it is equivalent to the blocking switch according to the related art.
[0007]
Here, the switching device controller 6 operates as follows.
(1) The traffic flow is grasped based on the information obtained from each non-blocking switch 2.
(2) The flows in each non-blocking switch are grouped according to the combination of the source port / destination port, and totalized for each port. This process is performed for all the non-blocking switches 2.
(3) Based on the information obtained in the above (2), a port having a larger communication volume with the outside of the switch than the inside of the switch and a port having the least communication volume are selected.
In addition, for a port having a large traffic, it is determined which non-blocking switch has a large traffic based on the destination address. This process is performed for all the non-blocking switches 2.
(4) The traffic switching device 1 is operated to exchange the port with the highest traffic obtained in the above (3) with the port with the lowest traffic among the ports belonging to the non-blocking switch 2 that is the destination. I do. This process is performed for all the non-blocking switches 2.
[0008]
The above-described processing is performed periodically according to the traffic fluctuation. The control loop is generally suitable in units of several minutes, but is not particularly limited.
As described above, by grouping the traffic, the same performance as that of the conventional non-blocking switch can be obtained as long as the traffic flow can be accommodated in the non-blocking switch 2.
Further, when the locality (locality) of the traffic is not sufficient and the flow cannot be accommodated in the non-blocking switch 2, allocation is performed in order from the flow having the highest bandwidth usage rate. As a result, the usage rate of the inter-module link 3 is reduced, so that performance higher than that of the conventional blocking switch can be obtained.
Various methods can be considered for grasping the traffic flow, and details will be described later.
An electrical switch and an optical cross-connect switch can be considered as the traffic switching device 1. However, as the link speed increases, the latter can be configured at a lower cost. In this case, the non-blocking switch 2 used as a module also has an optical interface.
[0009]
[Embodiment]
Hereinafter, an embodiment of the present invention will be described with reference to a tree-type cluster shown in FIG.
FIG. 2 is a schematic diagram showing the most basic type (parallel type) of a switch cluster based on the present invention.
In the configuration of FIG. 2, five 8-port Gigabit Ethernet switches are used as the non-blocking switch 2, and four are used for accommodating user links, of which two ports are used for the inter-module link 3. Therefore, the user accommodation link is (8 -2) × 4 = 24.
In the port numbers, ports 7 and 8 of each non-blocking switch 2 are used as inter-module links 3, and ports 1 to 6 are used as user accommodation links 4.
The ratio between the inter-module link 3 and the user accommodation link 4 in each non-blocking switch 2 is a trade-off between the locality of the accommodated user traffic and the cost that can be applied to the system.
When the expected locality is strong, the bandwidth required for the inter-module link 3 is small, so that a large ratio can be obtained, and as a result, the port unit price of the user accommodation link decreases.
On the other hand, when the expected locality is weak, the bandwidth to be allocated to the inter-module link 3 also increases, and the unit cost of the port increases.
[0010]
Note that, in the present embodiment, a switch having an uplink port faster than the user accommodation port may be employed as the non-blocking switch 2.
For example, if four switches each having a configuration of 48 ports of 100 Mbps Ethernet and two ports of Gigabit Ethernet are applied to the user accommodation switch of FIG. 2, the number of user accommodation links is 192.
As the traffic switching device 1, an optical cross-connect switch is used. This is because the optical cross-connect switch is simpler and cheaper than the electrical switch.
Each non-blocking switch 2 includes a flow information detecting device such as NetFlow (see Non-Patent Document 1 described above), sFlow (see Non-Patent Document 2 described above), or LFAP (see Non-Patent Document 3 described above). It has been incorporated.
FIG. 3 shows a processing procedure of the switching device controller 6 for controlling the traffic switching device 1 (here, the optical cross-connect switch).
First, the non-blocking switch 2 (that is, the optical cross-connect switch) is initialized, and the user accommodation side and the switch side are connected 1: 1 (step 10).
In the main loop, collection of flow information (step 11) and processing (step 12) to (step 16) are main procedures.
If the traffic from the port m to the port n of the i-th (here, 1 ≦ i ≦ 4) non-blocking switch 2 is represented as Tim _{, n} , the traffic matrix generated in the process of step 12 is as follows: It is expressed like a formula.
[0011]
(Equation 1)

・ ・ ・ ・ ・ ・ ・ ・ ・ （１）
Here, a certain port j (1 ≦ j ≦ 6) is defined as in the following equation (2).
[0012]
(Equation 2)

・ ・ ・ ・ ・ ・ ・ ・ ・ （２）
[0013]
FIG. 4 shows positions of to_local, to_remote, from_local, and from_remote in port 1 (j = 1) of the i-th non-blocking switch 2.
In addition, a transfer amount serving as a threshold value for performing traffic grouping is defined as Tthresh. Tthresh is set to, for example, a transfer amount corresponding to a line usage rate of 10%.
In step 13 described above, the following processing is executed.
Sorting is performed in ascending order of Tj and total, and the following determination is made in order of those satisfying Tj, total> Tthresh.
(A) If Tj, to_local <Tj, to_remote and Tj, from_local <Tj, from_remote, port j is set as a movement target.
(B) If Tj, to_local> Tj, to_remote and Tj, from_local <Tj, from_remote, the port j is not set as a movement target.
(C) If Tj, from_local <Tj, to_remote, and Tj, from_local> Tj, from_remote, if Tj, from_local <Tj, to_remote, port j is to be moved.
(D) If Tj, to_local> Tj, to_remote, and Tj, from_local> Tj, from_remote, port j is not set as a movement target.
In addition, a flag of “movable” is set for a port where Tj, total <Tthresh.
[0014]
Next, in step 14 described above, the following processing is executed.
Each non-blocking switch 2 examines the destination address of the traffic flow forming the Tj, to-remote of the port that has become the “moving target”, and selects the non-blocking switch 2 to which they most belong to as the destination. This process is executed for all the non-blocking switches 2.
Next, in step 15 described above, the following processing is executed.
The ports are sequentially moved by operating the traffic switching device 1.
That is, the “move target” port j of the i-th non-blocking switch 2 is replaced with the port m having the smallest communication amount among the “movable” ports of the n-th non-blocking switch 2 at the destination.
At this time, the entry of the MAC address associated with the “movable” port j and the entry of the MAC address associated with the “movable” port m are deleted from all the switches.
In the above description, the case of exchanging one set at a time has been described. However, the present invention is not limited to this case. For example, two ports with the largest Tj and total and two ports with the smallest Tj and total are taken out. Two sets of “movable object” and “movable” may be used.
Finally, in step 16, the process waits for a specified time before entering the next control loop.
[0015]
As described above, in the present embodiment, each non-blocking switch 2 exchanges the port accommodation so as to lower the usage rate of the inter-module link line 3.
Accordingly, the rearrangement is performed adaptively with respect to the locality of the entire traffic, so that the usage rate of each non-blocking switch can be increased and the equipment can be used effectively.
In particular, when the locality of traffic is large and the locality is unpredictable, as in the Internet, the number of small modular switches can be increased or decreased as necessary without installing a single large non-blocking switch as in the past. Since it can be used, efficient equipment planning becomes possible.
As described above, the invention made by the inventor has been specifically described based on the embodiment. However, the present invention is not limited to the embodiment, and can be variously modified without departing from the gist of the invention. Needless to say,
[0016]
【The invention's effect】
The following is a brief description of an effect obtained by a representative one of the inventions disclosed in the present application.
According to the packet switch cluster configuration method of the present invention, it is possible to configure a system that has better performance than a simple blocking switch and is less expensive than a large non-blocking switch.
[Brief description of the drawings]
FIG. 1 is a schematic diagram showing the most basic type (series type) of a switch cluster according to the present invention.
FIG. 2 is a schematic diagram showing the most basic type (tree type) of a switch cluster according to the present invention.
FIG. 3 is a flowchart showing a processing procedure of the switching device controller shown in FIGS. 1 and 2;
FIG. 4 is a diagram illustrating an example of a traffic flow matrix generated in FIG. 3;
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Traffic switching apparatus, 2 ... Non-blocking switch, 3 ... Inter-module link line, 4 ... User accommodation link line, 5 ... Control interface, 6 ... Switching device controller.

Claims (4)

In a packet switch cluster configuration method for connecting a plurality of packet switches in a cluster,
A packet switch cluster configuration method characterized by grasping the entire traffic flow and grouping traffic based on its locality. The method according to claim 1, wherein the grouped traffic flows are concentrated on a specific packet switch. 3. The packet switch cluster configuration method according to claim 2, wherein the grouped traffic flows are concentrated on a specific packet switch using an optical cross-connect switch. In a packet switch cluster configuration method for connecting a plurality of packet switches in a cluster,
Collecting traffic flow information from each of the packet switches;
A step 2 of grouping the intra-switch flows based on the combination of the source port and the destination port based on the traffic flow information collected in the step 1, and summing up each port of each packet switch;
A step of selecting, for each port unit of each of the packet switches, a port having a larger traffic volume from the inside of the packet switch to the outside of the packet switch and a port having the smallest traffic volume based on the information obtained in the step 2; ,
A step 4 of determining which packet switch receives a large amount of communication based on the destination address for the port having a large communication volume obtained in the step 3;
For each of the packet switches, the port with the largest traffic volume determined in step 3 and the port with the largest traffic volume among the ports belonging to the packet switch that is the communication destination of the port with the large traffic volume determined in step 4 Step 5 of exchanging a small number of ports.

Images