JP2002044178A - Communication control system for i/o channel - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a communication control system and a device for an I/O channel having a redundant path to realize a Fail-over and Load Balancing. SOLUTION: Using information with respect to redundant paths registered in advance in the communication control device controls communication so as to attain Load Balancing and Fail-over without explicit notification of a communication software program even when different addresses are added to each port. Furthermore, the redundant paths are grouped and designating a transmission ratio in respective groups can designate in details how to use the redundant path resources.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an I / O for connecting a plurality of computers and a computer / I / O device.
Regarding the communication control method of the channel, particularly in the case of an I / O channel having a redundant path,
The present invention relates to a communication control method and device for implementing balancing.

[0002]

2. Description of the Related Art A connection between computers is generally composed of a network such as Ethernet (registered trademark) and a network adapter device serving as a connection port on the computer side. Usually, the network adapter device is a PCI (Peripheral Component).
ent Interconnect) between the I / O bus and the network to control communication. In addition, an apparatus for connecting a computer and an I / O device is often constituted by a PCI bus, a dedicated I / O channel, a controller thereof, and the like.

In such a connection between computers and a connection between computers and I / O devices, a dual system is formed when high reliability, high availability, etc. are required, and a down rate in the event of a failure is suppressed.

Normally, in a cluster system, a computer main body, a disk device, and the like are duplicated, and the computer and the disk are connected so that switching operation is possible. In addition, a network between computers is provided to exchange information such as heartbeats.

In such a system, each connection itself is duplicated. In this method, the same address is designated when passing through either of the two paths by a unique protocol, and packets are distributed to the two paths according to the operation status of each path.

[0006]

However, such a conventional method has the following problems.

That is, this method cannot be applied to a case where a separate address is added to each port, such as a network constituted by Ethernet or IP, or a fiber channel.

It is also possible to realize duplexing by, for example, changing the path used by the software when a failure is detected. However, in such a method, it is necessary to provide a path change function for each software. There is a problem that cannot be used as it is.

As described above, according to the conventional method, when the connection between the nodes is duplicated in a network in which different addresses are added to the respective ports, the Failover and the Load Balancing can be performed without changing the software.
There is a problem that the uncing function cannot be realized.

[0010] It is an object of the present invention to provide a method for controlling a plurality of computers
When an I / O channel connecting a computer and an I / O device has two or more redundant paths, a hardware-side process such as Failover or Load Balance is performed.
An object of the present invention is to provide a communication control method for implementing ing.

[0011]

SUMMARY OF THE INVENTION The present invention provides an I / O device for connecting a plurality of computers and a computer / I / O device.
Failov when the / O channel has a redundant path
The present invention relates to a communication control method and an apparatus for realizing er and Load Balancing. In the present communication control method, when two nodes (equivalent to a computer and an I / O device) have redundant paths, information on a plurality of paths is registered in advance in a communication control device (equivalent to a channel adapter) of each node. The registration information includes information for identifying a partner node, information for identifying ports at both ends of a path, information indicating an operation state of each path, L
The information includes information indicating which ratio is used for load balancing between the paths at the time of load balancing, information indicating the use order for determining which path a packet of the failed path is to be sent to at the time of failover, and the like.

[0012] Then, when the information is transmitted from the communication control device by using the information, the port identifier in the header is replaced as necessary, so that the load balancing is performed.
And Failover.

When a packet is received by each port, information in the received packet is allocated by identifying a reception request (or a reception queue) by an identifier of a header above a port identifier. At this time, when the information in the header is checked, information on the registered redundant path may be referred to in order to cope with a case where the port identifier is changed on the transmission side.

As described above, by performing communication control using information on a plurality of redundant paths registered in the communication control device in advance, even when a different address is added to each port, software for performing communication can be provided. Load Balancing, Failover, etc. can be performed without being explicitly aware.

[0015]

Embodiments of the present invention will be described below in detail.

An object of the present embodiment is to provide a method for controlling load balancing or communication without explicitly considering communication software when a different address is added to each port in an I / O channel having a plurality of redundant paths. It is an object of the present invention to provide a communication control method capable of Failover and the like.

FIG. 1 shows a communication control apparatus 10 according to the present invention.
FIG. 3 is a block diagram showing a configuration of a 0. FIG. 2 is a block diagram including where the communication control device 100 is located in the computer 200, and FIG. 3 is a flowchart showing a processing procedure of the redundant path control unit 150 of FIG.
FIG. 5 shows an example of a table placed in the table holding area 152 of FIG. FIG. 6 shows a format of an FC-VI (standard for realizing a virtual interface on a fiber channel) packet, and FIG. 7 shows a case where two nodes including the communication control device of the present invention are connected by a plurality of redundant paths. An example in the case of connection is shown.

In FIG. 1, a communication control unit 100 includes a transmission control unit 120, a redundant path control unit 150, port transmission / reception control units 160, 162, 164, 166, and a transmission / reception port 1.
70, 172, 174, and 176. Although FIG. 1 shows a configuration having four ports, the number of ports may be any number of two or more. There are as many port transmission / reception control units as the number of ports.

The transmission / reception control unit 120 passes a transmission / reception request from a node (corresponding to a computer or an I / O device) to the redundant path control unit 150, and when receiving a packet, the reception header processing unit 122 checks the header. For example, it transfers received data to a reception request (or reception queue).

The redundant path control unit 150 includes the transmission / reception control unit 12
The port information in the transmission request passed from 0 is compared with the information on the redundant path registered in the table holding area 152 in advance, and when the conversion of the port identifier or the like is necessary, the transmission port conversion unit 154 determines the transmission / reception port identifier. It converts the packet and sends out the transmission packet to the port transmission / reception control units 160, 162, 164, 166. The received packet is passed to the transmission / reception control unit 120 as it is.

Port transmission / reception control units 160, 162, 16
4 and 166 send out the packet passed through the redundant path control unit 150 to the I / O channel via the port 1, port 2, port 3 and port 4. Also, it plays a role of transferring a packet received from an I / O channel via the port 1, port 2, port 3, or port 4 to the redundant path control 150.

In FIG. 2, computer 200 is a CP
U214, main memory 230, input / output bridge 220,
22, a connection unit 250 for connecting them, a channel adapter 224, a display device, a keyboard, a disk, and the like. In the figure, the communication control device 100 is connected to an interconnect 25 like an input / output bridge 220.
0 or the input / output bridge 222 via an I / O bus such as PCI as in the channel adapter 224.
Take a form that leads to

The nodes terminating the I / O channels include computers and I / O channels such as disk devices.
Although there is an I / O device, the communication control device is connected to some part of the I / O device as in the case of the computer.

FIG. 3 is a flowchart showing the operation of the transmission process of the redundant path control unit 150 of FIG. First, a transmission request is received from the transmission / reception control unit 120 (step 30).
0). Destinati in the table holding area 152
The on identifier and the Source identifier are compared with the Destination identifier and the Source identifier in the transmission request (step 305), and it is checked whether there is a matching entry. If there is no matching entry, the entry is transmitted to the port transmission / reception control unit of the port requesting transmission (step 350). If there is a match,
There are redundant paths. Therefore, it is determined from which port to transmit (step 310). If the transmission port determined at this time is the same as the one requested for transmission, the transmission is performed to that port (step 350). If they are different, the identifier needs to be converted (step 330), the identifier is converted (step 340), and then sent to the port transmission / reception controller of the converted port (step 350).

FIGS. 4 and 5 show the table holding area 15 of FIG.
2 shows an example of a table placed in 2. This example is shown in FIG.
3 shows examples of settings in nodes 1 and 2 when three redundant paths exist between two nodes in the configuration of FIG.

The redundant path numbers 400 and 500 are numbers that take different values for each of the partner nodes when there are redundant paths with a plurality of partner nodes, and indicate which of the redundant paths with the same partner node. Use to judge. In the example of FIG. 7, since only the node 1 and the node 2 are connected, the redundant path number has the same value (1 in this case) in both nodes.

Destination port identifier 41
0 and 510 are partner ports, and Source port identifiers 415 and 515 are identifiers for identifying their own ports, and are identifiers corresponding to D_ID and S_ID of fiber channel, MAC address and IP address of Ethernet, and the like.

The operating states 420 and 520 are information indicating whether each path is operable or has a fault. In this example, the operating state is 1 when the path is 1, and the stopped state when 0.

The group identifiers 425 and 525 divide a plurality of redundant paths into groups, and load each group.
It is an identifier used when performing Balancing,
The transmission ratios 430 and 530 indicate the ratio at which a transmission request (or transmission queue) is allocated to each path in the group.
In this example, the top two paths belong to the same group,
The transmission ratio is 2: 1. If the transmission ratio is set to 0, this path becomes a standby path and is not normally used, but is used when a failure occurs in another path belonging to the same group.

The transmission request (or transmission queue) is divided into, for example, a transmission request identifier and a transmission queue identifier. When the remainder obtained by dividing these identifiers by 3 is 0 or 1, the path of 1 is used. When the remainder is 2, a pass of 2 is used, and so on. The distribution unit may be a transmission packet unit in addition to the transmission request and the transmission queue. However, it is an inappropriate method when the receiving side does not desirably overtake the packet (because the packet is dropped).

The use order 435, 535 indicates the order in which each path is used when a failure occurs. When a failure occurs, firstly, if there is one belonging to the same group, another path in the same group also handles the traffic of the failure path (according to the transmission ratio). If there is no active path in the same group, the other group with the lowest order of use handles the traffic on the failed path.

FIG. 6 shows an FC-VI packet format. In FIG. 6, D_ID 620 is Des
S_ID 630 is S in the termination port identifier.
source port identifier. Also, VI_HA
NDLE 640 corresponds to a queue identifier.

FIG. 7 shows an example in which two nodes including the communication control device of the present invention are connected by a plurality of redundant paths. In this example, the nodes 1 and 2 correspond to either a computer or an I / O device. The communication control devices 712 and 752 have four ports in this example.
Then, (port 11, port 21), (port 12,
Three redundant paths (port 23) and (port 13, port 24) exist between the two nodes. The network 700 is a switch for relaying and exchanging packets,
It is composed of devices such as routers, and communication paths spanned between these devices.

The effect of the embodiment is that communication control is performed using information on a plurality of redundant paths registered in the communication control device in advance, so that even when a different address is added to each port, communication can be performed. Software that performs load balancing and Fai without explicit awareness
Lover and the like become possible.

Further, a plurality of redundant paths are grouped,
By specifying the transmission ratio in each group, it becomes possible to specify in detail how to use a plurality of upper path resources.

[0036]

As described above, according to the present invention, by performing communication control using information on a plurality of redundant paths registered in advance in the communication control device, a separate address is added to each port. In this case, even if the software that performs communication does not explicitly
ng and Failover become possible.

Further, a plurality of redundant paths are grouped,
By specifying the transmission ratio in each group, it becomes possible to specify in detail how to use a plurality of upper path resources.

[Brief description of the drawings]

FIG. 1 is a block diagram of a communication control device according to an embodiment of the present invention.

FIG. 2 is a block diagram of a computer according to the embodiment of the present invention.

FIG. 3 is a flowchart illustrating a processing procedure of a redundant path control unit according to the embodiment of this invention.

FIG. 4 is a diagram showing an example of a table placed in a table holding area according to the embodiment of the present invention.

FIG. 5 is a diagram showing an example of a table placed in a table holding area according to the embodiment of the present invention.

FIG. 6 is a diagram showing a format of an FC-VI packet according to the embodiment of the present invention.

FIG. 7 is a diagram illustrating an example of a case where two nodes including the communication control device according to the embodiment of the present invention are connected by a plurality of redundant paths.

Continued on the front page F term (reference) 5B083 AA04 BB03 CC04 CD11 CE01 DD09 DD13 EE11 GG04 5K030 GA08 JA07 JA08 JT02 KA01 KA02 LC09 LC11 MA04 MB02 MB04 5K034 AA07 AA17 DD03 EE07 HH64 JJ15

Claims (5)

[Claims] 1. A communication control apparatus for transmitting and receiving data from a computer and an I / O device to an I / O channel.
When the / O device is connected via a plurality of network paths via a communication control device having a plurality of ports, the communication control device is provided with a Destina of a plurality of paths in advance.
The destination port identifier and the source port identifier are registered in advance, and the destination identifier and the Sou in the header of the transmission packet are registered using the table information.
A communication control method for an I / O channel, wherein traffic is distributed to another path by changing an rc identifier. 2. The communication control device according to claim 1, wherein information indicating an operation state of each path is registered in a table of the communication control device, and the traffic to the path which is stopped due to a failure or the like is used by using this information. A communication control method characterized by sorting to paths. 3. The method according to claim 1, wherein information indicating a group identifier of each path is registered in a table of the communication control device.
A communication control method wherein a plurality of paths are divided into a plurality of groups by a group identifier, and traffic is distributed within each group. 4. The communication control device according to claim 3, wherein information indicating a transmission ratio of each path is registered in a table of the communication control device, and traffic is distributed to each path in the group grouped by the group identifier at any ratio. Communication control method characterized in that 5. The method according to claim 1, wherein information indicating a use order of each path is registered in a table of the communication control device, and is used when there is no active group in a group to which the failed path belongs when a failure occurs. A communication control method characterized by distributing traffic to active paths in the smallest order.