RU2574350C2 - Computer system and method for communication in computer system - Google Patents

Abstract

FIELD: radio engineering, communication.

SUBSTANCE: computer system comprises: a controller; a plurality of switches, each performing a relay procedure which is defined in a stream element specified by the controller with respect to a packet coordinated with said stream element, and a plurality of nodes performing communication through any of the plurality of switches, wherein the controller receives a first MAC address of a first node from the plurality of nodes from the first node and specifies the first MAC address as a destination address in a stream element rule for each of the plurality of switches and specifies transfer processing for the destination node as the action of the stream element for each of the plurality of switches; each of the plurality switches transfers a packet containing a destination address to the destination node based on the stream element specified for said switch, irrespective of the transmission source address of said packet.

EFFECT: reducing consumption of computer system resources by implementing optimal routing.

10 cl, 13 dwg

Description

FIELD OF THE INVENTION

The present invention relates to a computer system and a method for communicating in a computer system, and in particular, relates to a computer system using the OpenFlow method.

State of the art

When communicating using Ethernet (registered trademark), the flexibility of a physical communication line that can be used on a network is lost due to the Spanning Tree Protocol (STP) and, accordingly, the ability to communicate over multiple routes is gradually lost.

To solve the problem, route management using OpenFlow is proposed (see Non-Patent Source 1). A computer system using the OpenFlow method is disclosed, for example, in JP 2003-229913A (Patent Source 1). The network switch corresponding to the method (hereinafter referred to as the programmable flow switch (PFS), saves detailed information, for example, protocol type and port number, in the flow table and can control the flow. It should be noted that PFS is also called the OpenFlow switch.

1 is a diagram showing an example configuration of a computer system using the OpenFlow method. 1, a programmable stream controller 100 (PFC, referred to as an open stream controller) defines a stream element for PFS 200 and 300 in a single subnet (P-stream network) for controlling flows in a subnet.

Each of the PFS 200 and 300 refers to its own stream table to perform an action (for example, relaying and rejecting a data packet) specified in the stream element and the corresponding header information of the received packet. In particular, in the case of receiving a packet transferred between hosts (HOST) 400, each of the PFS 200 and 300 performs the action specified in the stream element if the header information of the received packet matches the (matches) (rule) stream element specified in its own table streams. On the other hand, when the header information of the received packet is not consistent with (matches) the (rule) stream element specified in its own stream table, each of the PFS 200 and 300 recognizes the received packet as the first packet, informs the PFC 100 of the receipt of the first packet and transmits packet header information to the PFC 100. The PFC 100 defines a stream element (stream + action) corresponding to the reported header information for the PFS, which is the source of information of the first packet.

As described above, in the conventional OpenFlow method, after any of the PFS 200 and 300 receives a packet being transferred between the hosts 400, the transfer control is performed on the packet transmitted and received between the hosts 400 by the PFC 100.

Bibliography

Patent Source 1: JP 2003-229913A.

Non-Patent Source 1: OpenFlow Switch Specification Version 1.0.0 (Wire Protocol 0x01) December 31, 2009.

SUMMARY OF THE INVENTION

The PFC in the traditional OpenFlow method sets the route of the packet carried between the source terminal and the destination terminal, and sets the flow element for switches on the route. In addition, even if there is one and the same destination, the flow element and the route between the source terminal and the destination terminal must be set each time another packet is generated at the source terminal. Thus, when using the OpenFlow method, there is a danger that the resources of the system as a whole (the number of flow elements) will be spent in large volumes.

The computer system of the present invention includes a controller; a plurality of switches, each of which carries out a relay operation defined in a flow element specified by the controller, with respect to a packet consistent with the flow element; and a plurality of nodes communicating through any of a plurality of switches. The controller sets the destination address as a rule of the stream element and sets the transfer processing for the destination node as the action of the stream element. Each of the many switches transfers a packet containing the destination address to the destination node based on the flow element specified for the switch, regardless of the source address of the receive packet.

In addition, it is desirable that the controller defines a flow element for each of the plurality of switches before transferring the packet among the plurality of nodes.

In addition, it is desirable that the controller obtains the first MAC (media access control) address of the first node from the many nodes in response to the first ARP (address resolution protocol) request from the first node and sets the first MAC address for each of the many switches as a rule of the element flow.

In addition, it is desirable that the controller transmit to the first node an ARP response having the MAC address of another node from the plurality of nodes as a transmission source, as a response to the first ARP request from the first node to another node.

In addition, the controller receives the first MAC (medium access control) MAC address of the first node (VM1) based on the first ARP (address resolution protocol) request from the first node from the many nodes and sets the first MAC address for each of the many switches as a rule of the flow element . In addition, it is desirable that the controller issue a second ARP request and specify the second MAC address of the second node obtained based on the response to the second ARP request, for each of the many switches as a rule, a flow element.

In addition, the controller transmits to the first node an ARP response having the MAC address of the other node as the source address in response to the first ARP request addressed to the other node from the first node. In addition, it is desirable that the controller transmit to another node an ARP response to a third ARP request addressed to the first node and transmitted from another node.

In addition, it is desirable that the plurality of switches include a plurality of first switches directly connected to the plurality of nodes. In this case, it is desirable for the controller to specify a flow element for arbitrarily selected switches from the plurality of first switches, without setting it for the remaining switches.

In addition, it is desirable that the controller sets a flow element for each of the multiple switches to implement ECMP routing (selecting multiple routes of the same cost) on the receive packet.

The communication method of the present invention includes the step of setting a flow element for each of the plurality of switches by the controller; a step for each of the plurality of switches to perform a relay operation specified in the flow element with respect to a receive packet consistent with the flow element specified by the controller; and a step for communicating with each of the plurality of nodes through each of the plurality of switches. Defining a flow element includes the step of setting the controller to the destination address as a rule of the flow element; and a step of setting transfer processing for the destination node as an action of the flow element. The implementation of the communication includes the transfer by each of the plurality of switches of the receive packet containing the destination address to the destination node regardless of the transmit address of the source of the receive packet.

In addition, it is desirable that the task of the flow element was carried out before the transfer of the packet among many nodes.

According to the present invention, it is possible to reduce the resource consumption of a computer system as a whole using the OpenFlow method.

Brief Description of the Drawings

Other objects, effects, and features of the aforementioned invention will become more apparent on the basis of the description of exemplary embodiments given in conjunction with the accompanying drawings:

figure 1 is a diagram showing an example configuration of a computer system using the OpenFlow method;

figure 2 is a diagram showing an example configuration of a computer system according to the present invention;

3A is a diagram illustrating an example of a method for defining a stream and a method for communicating in a computer system according to the present invention;

3B is a diagram showing an example of a method for defining a stream and a method for communicating in a computer system according to the present invention;

3C is a diagram showing an example of a method for defining a stream and a method for communicating in a computer system according to the present invention;

3D is a diagram showing an example of a method for defining a stream and a method for communicating in a computer system according to the present invention;

3E is a diagram showing an example of a method for defining a stream and a method for communicating in a computer system according to the present invention;

3F is a diagram showing an example of a method for defining a stream and a method for communicating in a computer system according to the present invention;

3G is a diagram showing an example of a method for defining a stream and a method for communicating in a computer system according to the present invention;

FIG. 3H is a diagram illustrating an example of a method for defining a stream and a method for communicating in a computer system according to the present invention;

FIG. 3I is a diagram illustrating an example of a method for defining a stream and a method for communicating in a computer system according to the present invention;

FIG. 3J is a diagram showing an example of a method for defining a stream and a method for communicating in a computer system according to the present invention; and

4 is a diagram illustrating a configuration of a logical network divided into multiple networks due to flow control according to the present invention.

Description of Exemplary Embodiments

Next, exemplary embodiments of the present invention will be described with reference to the accompanying drawings. In the drawings, the same or similar reference numerals are assigned to the same or similar components.

Computer system configuration

2, a configuration of a computer system according to the present invention will be described. 2 is a diagram illustrating an example configuration of a computer system according to the present invention. A computer system according to the present invention includes a programmable flow controller 10 (hereinafter referred to as PFC 10), a plurality of programmable switches 20-1 to 20-3 and 30-1 to 30-3 of a stream (hereinafter referred to as PFS 20-1 to 20-3 and 30-1 - 30-3), physical servers 40-1 - 40-5 (hereinafter referred to as SV 40-1 - 40-5), and storage 50, which are connected through a communication network. Moreover, when PFS 20-1 - 20-3 and 30-1 - 30-3 are described without distinguishing them from each other, each of PFS 20-1 - 20-3 and each of PFS 30-1 - 30-3 are called PFS 20 and PFS 30, respectively. And when SV 40-1 - 40-5 are described without distinguishing them from each other, each of SV 40-1-40-5 is referred to as SV 40.

SV 40 and storage 50 are computer units, each of which has a central processing unit (CPU), a main storage unit and external storage device, which are not shown in the figure, and communicate with other SV 40s by executing a program stored in an external storage device. Communication with SV 40 is carried out through PFS 20 and 30. In accordance with the program, SV 40 performs the function represented by a web server, file server, application server, client terminal, etc. For example, when SV 40 acts as a web server, SV 40 transfers an HTML document and image data from a storage unit (not shown) to another SV 40 (e.g., a client terminal) in accordance with a request from a client terminal that is not shown.

SV 40 includes a virtual machine VM, implemented by logical or physical separation of the CPU (not shown) and the storage area of the storage unit (not shown). In the example shown in FIG. 2, virtual machines VM1 and VM2 are implemented in SV 40-1, virtual machines VM3 and VM4 are implemented in SV 40-2, virtual machines VM5 and VM6 are implemented in SV 40-3 and virtual machines VM7 and VM8 implemented in SV 40-4. VM1-VM8 virtual machines can be implemented by the guest operating system (GOS) emulated on the main operating system (HOS) on each server, or by software running on the GOS.

A virtual machine VM transmits and receives data to and from other devices (for example, a computer unit on an external network and a virtual machine VM on another physical server 40) through virtual switches (not shown) controlled by a virtual machine monitor or physical NIC (not shown). In the present exemplary embodiment, packet transmission is carried out, by way of example, in accordance with TCP / IP (Transmission Control Protocol / Internet Protocol).

In addition, the virtual switch (not shown) according to the present invention can be controlled based on the OpenFlow method described below, and the virtual switch can perform the traditional switching operation (level 2). In addition, each of the VM1-VM8 virtual machines and the external environment of the physical server are connected to each other by the principle of a bridge. Thus, direct transmission from the external environment can be based on the MAC addresses and IP addresses of the VM1-VM8 virtual machines.

PFC 10 controls communication in the system based on the OpenFlow method. The OpenFlow method demonstrates the method according to which, in accordance with the routing policy (flow element: flow and action), the controller (in this case, PFC 10) defines a multilayer structure and routes data in flow units to PFS 20 and 30 for route management and control nodes. Thus, the route management function is separate from the router and the switch, and optimal routing and traffic control can be implemented through a centralized management controller. PFS 20 and 30, to which the OpenFlow method is applied, are manipulated not by communication in units of transit segments, as in a traditional router and switch, but by communication as an end-to-end flow (END2END).

PFC 10 is implemented by a computer having a CPU and a storage unit (not shown). Flow control processing in PFC 10 is carried out by executing a program stored in a storage unit (not shown), and controls the operations of PFS 20 and 30 (for example, the relay operation of data packets) by setting the flow element (stream and action) for each of the PFS 20 and 30.

In addition, the MAC addresses of the host terminal (SV 40 and storage 50) and the virtual machine VM are set for the PFC 10 according to the present invention, before transferring packets between terminals (for example, between virtual machines VM). For example, PFC 10 pre-receives the MAC addresses of the host terminal and VM virtual machine in response to an ARP (Address Resolution Protocol) request.

PFC 10 generates a stream element, applying the obtained MAC address for the rule, and sets the stream element for all PFS 20 and 30 in the network. For example, PFC 10 generates a flow element for each PFS that is used to indicate the packet transfer destination addressed to the MAC address of VM1 and to transfer the packet and sets the flow element for all PFS 20 and 30 on the network. In the present invention, since the flow control is based on only the destination MAC address, a packet transfer destination corresponding to the rule (destination MAC address) specified for the stream item is determined regardless of the transmission source. For this reason, flow control can be carried out regardless of the source of the packet. Thus, according to the present invention, since the multiple route for packet transfer is formed by setting the optimal route for the destination terminal, it is possible to realize the optimal multi-route operation. In addition, since the flow element can be set for PFS without waiting for the first packet to be received, in contrast to the traditional method, network bandwidth can be increased. In addition, in the present invention, since the flow element is generated and set before the packet is transferred between the terminals, that is, before the system starts to operate, the processing load for controlling flows during operation is reduced in comparison with the traditional method.

Additionally, PFC 10 generates a stream element, applying the obtained MAC address for the rule, and sets the stream element for randomly selected from PFS 20 and 30 in the network, and the stream element is not set for the remaining PFS. For example, a flow element that uses the MAC address of VM1 as a rule is set for a selected part of the PFS 30 directly connected to the host terminal (SV 40 and storage 50). In this case, when PFS 30, for which a stream element is not defined, receives a packet addressed to VM1, the packet is discarded and is not transferred anywhere else. Thus, since the packet transfer destination can be logically separated, one physical network can be divided into many logical networks and operated. It should be noted that when a flow element defined for rejecting a packet addressed to a specific MAC address is assigned to a specific PFS, a similar effect can be achieved.

Each of the PFSs 20 and 30 includes a stream table (not shown) for which a stream element is defined, and processes the receive packet (eg, relay process and reject) in accordance with a given stream element. PFS 30 is a first stage switch directly connected to the host terminal (SV 40 and storage 50), and it is preferable to use, for example, an overhead switch (TOR) as PFS 30. In addition, for switch L2 and switch L3 connected to the second stage or following the host terminal, it is preferable to use, for example, a core switch (CORE) in PFS 20.

Each of the PFSs 20 and 30 accesses its own stream table (not shown) and performs the action (for example, relaying and rejecting the data packet) specified in the stream element and corresponding to the header data of the receive packet (in particular, the destination MAC address). In particular, each of the PFS 20 and 30 performs the action specified in the stream element when the header data of the receive packet matches (correspond to) the stream defined by the stream element specified in its own stream table. In addition, each of the PFSs 20 and 30 does not perform any processing on the packet when the header data of the receive packet does not match (correspond to) the stream defined by the stream element specified in its own stream table. In this case, the PFSs 20 and 30 may inform the PFC 10 of the receipt of the packet and may reject the packet.

In a stream element, any combination of addresses and identifiers from level 1 to level 4 of the OSI reference model (open system interactions) is specified as data (hereinafter referred to as a rule) to indicate a stream (data packet), and addresses and identifiers are included, for example, in header data TCP / IP data packet. For example, any of the physical layer 1 combinations; Layer 2 MAC addresses, Layer 3 IP addresses, Layer 4 physical port, and VLAN tag are specified for the flow element as a rule. However, in the present invention, the MAC address and IP address of the transmission source are not set for the stream item, and the destination MAC address is always set for the stream item. At the same time, a predetermined range of identifiers can be set for a stream element, for example, port number, address, etc. For example, the MAC addresses of virtual machines VM1 and VM2 can be set as destination MAC addresses as a rule of a stream element.

The action of the flow element determines, for example, the method of processing the TCP / IP data packet. For example, information is set indicating whether the received data packet is relayed and the destination of the data packet in the event of relaying the data packet. Additionally, when performing an action, you can specify data that requires the copying or rejection of a data packet.

The method of setting the stream and the method of communicating in a computer system

3A-3J, a method for defining a stream and a method for communicating in a computer system according to the present invention will be described in detail below. The flow setting for the VM1 virtual machine and the flow setting for the VM5 virtual machine will be described below by way of example. In addition, when the VM1-VM8 virtual machines, the physical servers 40-1 to 40-5, and the storage 50 are relatively the same, they are collectively referred to as nodes.

At the end of the system configuration (or system configuration change), the PFC 10 learns the system topology in the same way as a traditional flow controller does. The topology data known at this time include data related to the connection status of PFS 20 and 30, nodes (VM1-VM8 virtual machines, physical servers 40-1 - 40-5, and storage), an external network that is not shown ( for example, the Internet), etc. In particular, as the topology data, the number of device ports and destination port data are associated with a device identifier for indicating PFS 20 and 30 and nodes, and thus, the device identifier is recorded in the PFC storage unit 10. Destination port data includes the type of connection (comm ator / node / external network) to indicate the opposite side of the connection, and data to indicate the destination of the connection (ID switch in the case of the switch, MAC address in the case of the node and ID of the external network in the case of the external network).

3A, PFC 10 blocks an ARP request from a node to obtain (examine) the location (MAC address) of the requesting node. For example, an ARP request addressed to VM5 from VM1 arrives at PFC 10. PFC 10 extracts the MAC address of VM1 as the source node from the received ARP request. PFC 10 defines a rule for specifying a MAC address for a destination to generate a stream item. In this case, a flow element is generated for all PFS 20 and 30 in the system. It should be noted that the flow element to the MAC address can be predefined for the storage unit PFC 10.

3B, PFC 10, having examined the location (MAC address) of a node, registers a route to the node. For example, PFC 10 defines for all PFS 20 and 30 a flow element that defines the transfer of a packet addressed to the MAC address of VM1 and the transfer destination device. In this case, it is preferable to set the stream element for PFS 30-1 to determine the physical port connected to the virtual machine VM1 as the output destination, and to set the stream element for PFS 30 at the first stage, other than PFS 30-1, for alignment loads for PFS 20 at the second or next stage. For example, it is preferable to specify a flow element for PFS 30 to implement ECMP routing (selecting multiple routes of the same cost) for PFS 30.

In a normal level 2 study (L2 study), there may be cases where a loop (LOOP) is formed due to overflow (FLOODING) and when the intended study cannot be carried out due to load balancing. However, the present invention uses the OpenFlow method, and accordingly, these problems do not arise.

3C, PFC 10, for which a flow element is specified, transmits an ARP request for the destination requested by the node to all nodes except the node when receiving (learning) the MAC address. For example, PFC 10 transmits an ARP request addressed to VM5 as the destination of the ARP request shown in FIG. 3A to all nodes (virtual machines VM2-VM8, SV 40-5 and storage 50) except the requesting virtual machine VM1.

3D, PFC 10 obtains (learns) the location (MAC address) of the destination node based on the response (ARP response) to the ARP request shown in FIG. 3C. In the present example, the ARP response is transmitted from the VM5 virtual machine, and the PFC 10 obtains the location (MAC address) of the VM5 virtual machine by blocking the ARP response.

3E, PFC 10, having obtained (learned) the location (MAC address) of a node, registers a route to the node. In this case, PFC 10 sets for all PFS 20 and 30 a stream element that defines the transfer of a packet addressed to the MAC address of VM5 and the destination device. In this case, as described above, it is preferable to set the flow element for PFS 30 on the first stage from the host terminal to balance the load for PFS 20 on the second or next stage.

3F, the PFC 10 responds to an ARP request from the node shown in FIG. 3A through an intermediary. In this case, PFC 10 uses the MAC address of VM5 as a transmission source and issues an ARP response whose destination is VM1. Virtual machine VM1 receives an ARP response to an ARP request sent by itself, and receives the requested MAC address of VM5.

In the above operation, the processing (stream element) applied to packets respectively addressed to both the destination node and the requesting source ARP request node is specified for all PFS 20 and 30 in the system. In the example shown in FIG. 3G, by the above operation, a flow element applied to packets respectively addressed to the virtual machines VM1 and VM5 is set for all PFS 20 and 30. Thus, the transmission addressed to the virtual machine VM1 and the transmission addressed VM5 virtual machine are implemented normally. In this case, a packet addressed to each of the destinations is transmitted along a route consistent with the flow element determined by the destination MAC address, regardless of the source of transmission.

Additionally, to configure a single tree structure in the spanning tree protocol according to traditional Ethernet (registered trademark), a physical communication line is generated that is not used. For this reason, it is not possible to set many routes between specific nodes in Ethernet (registered trademark). However, in the present invention, a packet transfer destination is set for each of the PFS according to the destination, which allows the formation of multiple routes for implementing load balancing. For example, in the case of the above example, multiple routes are formed according to the flow element in each of the transfers for the virtual machine VM1 and the transfers for the virtual machine VM5, and load balancing is implemented.

In the above example, load balancing is applied by the ECMP specified in the flow element. However, the present invention is not limited to this, and link aggregation or load balancing can be applied to each flow element.

On the other hand, in order to be able to transmit an ARP request and make two-way communication between the requesting source node and the destination node, the destination node receives (examines) the location (MAC address) of the requesting source node from PFC 10. In particular, according to FIG. the ARP request addressed to the virtual machine VM1 from the virtual machine VM5 is transmitted to the PFC 10. According to FIG. 3 PFC 10, where the location (MAC address) of the virtual machine VM1 is stored, an ARP response is transmitted having the MAC address of the virtual machine VM1 as a transmission source to the VM5 virtual machine. The VM5 virtual machine blocks it to obtain the location (MAC address) of VM1. Thus, as shown in FIG. 3J, the virtual machine VM5 can transmit a data packet addressed to the virtual machine VM1. It should be noted that, since the flow element addressed to virtual machine VM1 and the flow element addressed to virtual machine VM5 are set independently, the communication path from virtual machine V1 to virtual machine V5 and the communication path from virtual machine V5 to virtual machine V1 not always the same.

Through the above operation, both virtual machines VM1 and VM5 obtain (learn) each other's locations (MAC addresses), and the transfer destination for the packet addressed to each of the virtual machines VM1 and VM5 is set for all PFS 20 and 30. Thus, it is ensured two-way communication between the VM1 virtual machine and the VM5 virtual machine.

In the present invention, since the flow element is determined based on the MAC address of the destination, the location of the source transmission node is not always required when defining the flow element. For this reason, the flow element can be set before the start of communication between nodes. In addition, it is not necessary to set the stream element for the communication route between nodes, unlike the traditional method, and it is enough to set the stream element of the destination MAC address for each of the PFS. Accordingly, it is possible to reduce the consumption of resources in the computer system as a whole.

Further according to FIG. 4, an example application of a computer system according to the present invention will be described. In the above example, a flow element for a packet addressed to a node is defined for all PFSs 20 and 30. However, the present invention is not limited to this, and nodes for which a stream element is specified may be limited to a portion of the PFS 30 directly connected to the node.

The computer system shown in FIG. 4 includes higher level switches (PFS 20-1 and 20-2) connected to a network 500, PFS 30-1, 30-2, and 30-3 directly connected to the host a terminal (not shown), for example, SV40, and nodes S and A. In this case, node A is connected to the system via PFS 30-2, and node S is connected to the system through PFS 30-3.

In the present example, using PFC 10 (not shown), the flow element is set for PFS 20-1, 20-2 and 30-3 to control the flow addressed to node S, and the flow element is set for PFS 20-1, 20-2, 30 -1 and 30-2 to control the flow addressed to node A. In this case, the packet addressed to node S reaches node S along the communication path passing through any of the PFS 20-1, 20-2 and 30-3, and the packet, addressed to node A, reaches node A along a communication path passing through any of the PFS 20-1, 20-2, 30-1 and 30-2. Thus, the node S is included in the logical network configured by PFS 20-1, 20-2 and 30-3, and the node A is included in the logical network configured by PFS 20-1, 20-2, 30-1 and 30- 2.

As described above, the computer system shown in FIG. 4 configures one physical network. However, with a selective definition of the flow element, the computer system is divided into two logical networks. Accordingly, a single physical topology can be manipulated as a plurality of VLANs.

As described above, exemplary embodiments of the present invention have been described in detail. However, the specific configuration is not limited to the above-described exemplary embodiments. Various modifications within the scope of the present invention are included in the present invention. 2, a system having a two-stage PFS group is shown in one example. However, the present invention is not limited to this, and the system may have a PFS group with the configuration of more stages. In addition, an external network can be connected to the PFS 20 through a Layer 3 (L3) switch, as in the traditional method.

This application is based on Japanese application No. JP 2010-202468, the disclosure of which is incorporated herein by reference.

Claims (10)

1. A computer system containing:
controller;
a plurality of switches, each of which carries out a relay operation, which is defined in the flow element specified by the controller, with respect to a packet consistent with this flow element, and
many nodes communicating through any of the many switches,
the controller receives the first MAC address (the address of the medium access control level) of the first node from the many nodes from the first node and sets the first MAC address as the destination address in the flow element rule for each of the many switches and sets the transfer processing for the node -address as an action of a flow element for each of the multiple switches; and
each of the many switches transfers a packet containing the destination address to the destination node based on the flow element specified for this switch, regardless of the transmission source address of this packet. 2. The computer system of claim 1, wherein the controller defines a flow element for each of the plurality of switches prior to transferring the packet among the plurality of nodes. 3. The computer system of claim 1, wherein the controller obtains a first MAC address of a first node from a plurality of nodes from a first ARP request (address resolution protocol) transmitted from the first node. 4. The computer system of claim 3, wherein the controller transmits an ARP response to the first node, in which there is a MAC address of another node from the plurality of nodes as a transmission source, as a response to the first ARP request from the first node to said other node. 5. The computer system according to claim 2, in which the controller issues a second ARP request and sets the second MAC address of the second node, which was obtained based on the response to the second ARP request, for each of the many switches as a rule of the flow element. 6. The computer system of claim 4, wherein the controller transmits, to said other node, an ARP response to a third ARP request addressed to the first node and transmitted from said other node. 7. The computer system according to any one of paragraphs. 1-6, wherein said plurality of switches includes a plurality of first switches directly connected to said plurality of nodes, and the controller defines a flow element for arbitrarily selected from said plurality of first switches without specifying a flow element for the remaining switches. 8. The computer system according to any one of paragraphs. 1-6, in which the controller defines a flow element for each of the plurality of switches to perform ECMR routing (selecting multiple routes of the same cost) with respect to the packet. 9. A communication method, comprising the steps of:
using a controller, set a flow element for each of the plurality of switches;
performing, with each of the plurality of switches, a relay operation specified in a flow element with respect to a packet consistent with this flow element; and
communicate between the source node and the destination node from a plurality of nodes through a plurality of switches,
moreover, the aforementioned task of the stream element contains the stages in which:
using the controller, get the MAC address (address of the medium access control level) of the first node from the plurality of nodes from the first node,
using the controller set this MAC address as the destination address in the rule of the flow element of each of the many switches and
specifying the transfer processing addressed to the destination node as an action of the flow element for each of the plurality of switches; and
said communication comprises the step of transferring, using each of the plurality of switches, a packet containing a destination address to a destination node, regardless of the transmission source address of this packet. 10. The communication method of claim 9, wherein said defining a flow element is carried out before packet transfer among a plurality of nodes.

Images