JP2021144614A - Interface control unit, information process system, and method of controlling interface control unit - Google Patents

Abstract

To speedily restore a link if one of communication interfaces of two systems connecting a pair of information processing units has a link down.SOLUTION: An interface control unit is included in one of a pair of information processing units connected through communication interfaces of two systems, and controls communication with the other information processing unit through the communication interfaces. The interface control unit executes detection processing for a link down of a communication interface, executes restoration processing on a link in a link down state when detecting a link down of one of the communication interfaces or when receiving notice of a link down occurring to the other communication interface from the other information processing unit through the one communication interface, and transmits a restoration control indication for restoring the link to the other information processing unit through the one communication interface.SELECTED DRAWING: Figure 1

Description

The present invention relates to an interface control device, an information processing system, and a control method for the interface control device.

In a plurality of information processing devices connected by a redundant network, the information processing device on the receiving side that detects an abnormality in one network notifies the information processing device on the transmitting side of the occurrence of an abnormality via the other network. The information processing device on the transmitting side continues to transmit the diagnostic text in a connectionless type at a predetermined cycle to the network in which the abnormality has occurred via the other network. Then, when the information processing device on the transmitting side receives a normal response to the diagnostic text from the information processing device on the receiving side, the network in which the abnormality has occurred is normally restored (see, for example, Patent Document 1).

In a data transmission system including a plurality of data transmission devices connected to a double loop-shaped data transmission line, the loop monitoring device specifies a loopback holding time and indicates a loopback when the data transmission system fails. Is given to the data transmission device before and after the faulty part. Each data transmission device sends a recovery signal to the data transmission device in the loopback state at the time of failure recovery, receives the recovery signal during the loopback, and if the loopback holding time has elapsed, a recoverable response signal. Will be returned. Then, each data transmission device releases the loopback after confirming that the response signal has been normally received (see, for example, Patent Document 2).

In a communication network system having a plurality of terminals including a standard link and a redundant link, when a detection means detects a failure of the standard link, the interface connected to the redundant link is controlled to take over the message transfer. (See, for example, Patent Document 3).

Japanese Unexamined Patent Publication No. 2000-78175 Japanese Unexamined Patent Publication No. 1-2457734 Japanese Patent Publication No. 2008-544678

When each of the pair of information processing devices connected via the two communication interfaces has an interface control unit that controls communication for each communication interface, the two communication interfaces are used in parallel to communicate information. It is possible. For example, when information is being communicated in parallel by two communication interfaces and a failure occurs in one of the communication interfaces, the firmware of the information processing device executes the failure recovery process using the management network. When the failure recovery process is executed by the firmware, the recovery time becomes longer than when the failure recovery process is executed by hardware control, and the longer the recovery time, the lower the performance of the information processing device.

On one side, the present invention aims to quickly restore a link when one of the two communication interfaces connecting a pair of information processing devices is linked down.

According to one viewpoint, the interface control device is included in one of a pair of information processing devices connected via two communication interfaces, and is an interface that controls communication with the other information processing device by the communication interface. When the control device executes a link-down detection process of the communication interface and detects a link-down of any of the communication interfaces, or from the other information processing device via one of the communication interfaces. When the notification of the link down occurring on the other side of the communication interface is received, the recovery process of the downed link is executed, and the recovery control instruction for recovering the link is sent to the other information processing device via one of the communication interfaces. Send.

On one side, the present invention can quickly restore a link if one of the two communication interfaces connecting the pair of information processing devices is linked down.

It is a block diagram which shows an example of the information processing system in one Embodiment. It is a sequence diagram which shows an example of the operation of the information processing system of FIG. It is a sequence diagram which shows another example of the operation of the information processing system of FIG. It is a block diagram which shows an example of the information processing system in another embodiment. It is a block diagram which shows an example of the connection between the cluster of FIG. 4 and SSU. It is a block diagram which shows an example of the internal structure of the cluster and SSU of FIG. It is a block diagram which shows an example of the interface control part provided in the cluster of FIG. It is a block diagram which shows an example of the interface control part provided in the SSU of FIG. It is a flow chart which shows an example of the operation of the information processing system of FIG. It is a sequence diagram which shows an example of the process at the time of link down detection in the information processing system of FIG. It is a sequence diagram which shows an example of the link restoration processing in the information processing system of FIG. It is a sequence diagram which shows an example of the recovery sequence of FIG. It is explanatory drawing which shows an example of the format of the packet sent and received between the cluster of FIG. 11 and SSU. It is a block diagram which shows an example of the information processing system in another embodiment. It is a block diagram which shows an example of the information processing system in another embodiment.

Hereinafter, embodiments will be described with reference to the drawings.

FIG. 1 shows an example of an information processing system according to an embodiment. The information processing system SYS1 shown in FIG. 1 has, for example, a pair of information processing devices 10 and 20 connected to each other via two high-speed serial interfaces SIFa and SIFb (link). For example, the serial interface SIF (SIFa, SIFb) is a 10G interface. The high-speed serial interface SIF is an example of a communication interface, and will be referred to as a serial interface below.

The information processing device 10 includes a core unit 12 that executes calculation processing, and an interface control device 14 that controls communication between the information processing device 20 of the other party. The interface control device 14 has interface control units 14a and 14b connected to the serial interfaces SIFa and SIFb, respectively. The interface control units 14a and 14b are connected to each other via the communication interface RIF for recovery control. The core unit 12 is connected to the interface control units 14a and 14b, respectively, via the communication interface DIF for degeneration control.

The information processing device 20 includes a core unit 22 that controls input / output of information to a storage device (not shown), and an interface control device 24 that controls communication between the information processing device 10 of the other party. The interface control device 24 has interface control units 24a and 24b connected to the serial interfaces SIFa and SIFb, respectively. The interface control units 24a and 24b are connected to each other via the communication interface RIF for recovery control. The core unit 22 is connected to the interface control units 24a and 24b, respectively, via the communication interface DIF for degeneration control.

The interface control units 14a and 24a are connected to each other via a serial interface SIFa having a transmission path and a reception path. The interface control units 14b and 24b are connected to each other via a serial interface SIFb having a transmission path and a reception path. That is, the information processing devices 10 and 20 can transmit and receive information such as data using the two serial interface SIFs.

The information processing devices 10 and 20 transmit and receive information using the two serial interface SIFs in the normal mode in which the two serial interface SIFs can be used. The information processing devices 10 and 20 transmit and receive information using only the other serial interface SIF in the degenerate mode in which one of the serial interface SIFs is linked down.

As shown by the X mark in FIG. 1, the information processing apparatus 10 instructs the information processing apparatus 20 to proactively execute the link-down recovery process when any of the two serial interface SIFs is linked down. Operates as the main information processing device. When any of the two serial interface SIFs is linked down, the information processing device 20 operates as a subordinate information processing device that subordinately executes the recovery process based on the instruction from the information processing device 10. In the example shown in FIG. 1, the serial interface SIFa is linked down.

The linked down serial interface SIFa cannot be used for link recovery processing. Therefore, as shown by a thick arrow in FIG. 1, the interface control unit 14a communicates with the interface control unit 24a via the interface control units 14b and 24b and the serial interface SIFb that operates normally, and the interface control unit 14a communicates with the interface control unit 24a. Execute the recovery process.

FIG. 2 shows an example of the operation of the information processing system SYS1 of FIG. That is, FIG. 2 shows an example of a control method of the interface control devices 14 and 24 of the information processing system SYS1. In FIG. 2, the serial interface SIDa is linked down while the information processing system SYS1 is executing communication in the normal mode using the serial interfaces SIFa and SIFb. The interface control unit 24a of the information processing device 20 executes the link down detection process, and detects the link down before the interface control unit 14a of the information processing device 10.

The interface control unit 24a that has detected the link down instructs the core unit 22 to perform the link degeneration process via the communication interface DIF. Further, the interface control unit 24a notifies the interface control unit 24b of the link down via the communication interface RIF. Upon receiving the link down notification, the interface control unit 24b notifies the interface control unit 14b of the information processing apparatus 10 of the link down via the serial interface SIFb.

When the interface control unit 14b receives the link down notification from the interface control unit 24b, the interface control unit 14b notifies the interface control unit 14a of the link down via the communication interface RIF. Upon receiving the link down notification, the interface control unit 14a instructs the core unit 12 to perform the link degeneration process via the communication interface DIF. As a result, the information processing system SYS1 transitions from the normal mode to the degenerate mode, and executes communication using the communication interface RIFb until the link down of the communication interface RIFa is restored.

Further, the interface control unit 14a notifies the interface control unit 14b of the recovery control instruction via the communication interface RIF, and starts the recovery process of the downed link. The interface control unit 14b transmits a recovery control instruction to the interface control unit 24b via the normal serial interface SIFb. The interface control unit 24b notifies the interface control unit 24a of the received recovery control instruction via the communication interface RIF. The interface control unit 24a starts the recovery process based on the reception of the recovery control instruction, and notifies the interface control unit 24b of the recovery response corresponding to the recovery control instruction via the communication interface RIF.

The interface control unit 24b transmits the recovery response received from the interface control unit 24a to the interface control unit 14b via the serial interface SIFb. The interface control unit 14b notifies the interface control unit 14a of the received recovery response via the communication interface RIF. Upon receiving the recovery response, the interface control unit 14a transmits the next recovery control instruction to the interface control unit 24a via the interface control units 14b and 24b. Then, the recovery control instruction and the recovery response are repeated a predetermined number of times to complete the recovery of the downed link.

When the restoration of the link is completed, the interface control unit 14a issues a restoration completion notification to the core unit 12. Upon receiving the restoration completion notification, the core unit 12 executes a process of releasing the degeneracy. Similarly, when the restoration of the link is completed, the interface control unit 24a issues a restoration completion notification to the core unit 22. Upon receiving the restoration completion notification, the core unit 22 executes a process of releasing the degeneracy. Then, the information processing system SYS1 returns from the degenerate mode to the normal mode, and the information processing devices 10 and 20 resume transmission / reception of information using the serial interfaces SIFa and SIFb.

FIG. 3 shows another example of the operation of the information processing system SYS1 of FIG. That is, FIG. 3 shows an example of a control method of the interface control devices 14 and 24 of the information processing system SYS1. A detailed description of the same operation as in FIG. 2 will be omitted. In the example shown in FIG. 3, the serial interface SIDa is linked down while the information processing system SYS1 is operating in the normal mode, as in FIG. However, the interface control unit 14a of the information processing device 10 executes the link down detection process, and detects the link down before the interface control unit 24a of the information processing device 20.

The interface control unit 14a that has detected the link down instructs the core unit 12 to perform the link degeneration process via the communication interface DIF. Further, the interface control unit 14a notifies the interface control unit 14b of the recovery control instruction via the communication interface RIF, and starts the recovery process of the downed link. The interface control unit 14b transmits a recovery control instruction to the interface control unit 24b via the normal serial interface SIFb.

The interface control unit 24b notifies the interface control unit 24a of the received recovery control instruction via the communication interface RIF. Upon receiving the recovery control instruction, the interface control unit 24a instructs the core unit 22 to perform the link degeneracy process via the communication interface DIF. Further, the interface control unit 24a starts the recovery process based on the reception of the recovery control instruction, and notifies the interface control unit 24b of the recovery response corresponding to the recovery control instruction via the communication interface RIF.

The interface control unit 24b transmits the recovery response received from the interface control unit 24a to the interface control unit 14b via the serial interface SIFb. The interface control unit 14b notifies the interface control unit 14a of the received recovery response via the communication interface RIF. Upon receiving the recovery response, the interface control unit 14a transmits the next recovery control instruction to the interface control unit 24a via the interface control units 14b and 24b. Then, as in FIG. 2, the recovery control instruction and the recovery response are repeated a predetermined number of times to complete the recovery of the downed link.

In this embodiment, when one of the serial interfaces SIFa and SIFb is linked down, the recovery control instruction and the recovery response are transmitted and received using the other of the degenerate normal serial interfaces SIFa and SIFb. For example, the interface control units 14a and 24a corresponding to the linked down serial interface SIFa bypass the interface control units 14b and 24b and transmit and receive the recovery control instruction and the recovery response using the normal serial interface SIFb.

At this time, the interface control units 14a and 14b can send and receive recovery control instructions and recovery responses via the communication interface RIF. Recovery control instructions and recovery responses can be transmitted and received between the interface control units 24a and 24b via the communication interface RIF. Therefore, the recovery of the downed link is executed by the interface control units 14a, 14b, 24a, 24b (that is, hardware) without using the firmware or the like executed by the service processor mounted on the information processing devices 10 and 20. can do.

As a result, in this embodiment, the link can be restored more quickly, the link down period can be minimized, and the performance deterioration period of the information processing system SYS1 can be minimized, as compared with the case where the link is restored by the firmware. Can be minimized. That is, when one of the two communication interfaces connecting the pair of information processing devices 10 and 20 is linked down, the link can be quickly restored. In addition, since the link down period can be minimized, the period for transmitting and receiving information with only one system can be minimized, and the risk of system down due to the link down of the remaining one system can be reduced. can.

FIG. 4 shows an example of an information processing system according to another embodiment. The same elements as those in FIG. 1 are designated by the same reference numerals, and detailed description thereof will be omitted. The information processing system SYS2 shown in FIG. 4 has, for example, eight clusters 100 (# 0 to # 7) and four SSUs (System Storage Units) 500 (# 0 to # 3). Each cluster 100 has a CPU (Central Processing Unit) 110 and a management device 120. Each SSU500 has an ASIC (Application Specific Integrated Circuit) 510, a management device 520 and a storage device 530.

Each CPU 110 controls the entire cluster 100, and each ASIC 510 controls the entire SSU 500. Further, the management device 120 of each cluster 100 and the management device 520 of each SSU 500 are connected to each other via a management bus MBUS such as a LAN (Local Area Network) interface. For example, the management devices 120 and 520 are service processors such as a BMC (Baseboard Management Controller) or an iRMC (integrated Remote Management Controller), and can communicate with each other by executing firmware. The storage device 530 is, for example, a memory module such as a DIMM (Dual Inline Memory Module), but the type of memory is not limited to SDRAM (Synchronous Dynamic Random Access Memory).

Each cluster 100 is connected to four SSU500s via a serial interface SIF. The one serial interface SIF shown in FIG. 1 corresponds to the four serial interfaces SIFa and SIFb shown in FIG. The number of clusters 100 and the number of SSUs 500 mounted on the information processing system SYS2 are not limited to FIG. For example, the serial interface SIF is a 10G interface. In the following, clusters # 0 to cluster # 7 and SSU # 0 to SSU # 3 may be referred to in order to identify the eight clusters 100 and the three SSUs 500, respectively.

Each cluster 100 is an example of a main information processing apparatus that instructes the corresponding SSU 500 to proactively execute a link-down recovery process when any of the serial interfaces SIFa and SIFb is linked down. Each SSU500 is an example of a subordinate information processing device that subordinately executes a recovery process based on an instruction from the corresponding cluster 100 when any of the serial interfaces SIFa and SIFb is linked down.

FIG. 5 shows an example of the connection between the cluster 100 of FIG. 4 and the SSU500. FIG. 5 illustrates the connection between clusters # 0 and SSU # 0, but the same applies to the connections between other clusters # 1 to # 7 and other SSUs # 1 to # 3. Each cluster # 0 to # 7 has four sets of interface control units 300a and 300b connected to each of SSU # 0 to # 3. Each SSU # 0 to # 3 has seven sets of interface control units 700a and 700b connected to each of the clusters # 0 to # 7. The pair of interface control units 300 (300a, 300b) is an example of an interface control device, and the pair of interface control units 700 (700a, 700b) is an example of an interface control device.

In the following, the interface control unit 300 provided in each cluster # 0 to # 7 is indicated by the reference numeral MMij, and the interface control unit 700 provided in each SSU # 0 to # 3 is indicated by the reference numeral SMmn. Reference numeral i indicates a number of the SSU500 to be connected, and reference numeral j indicates an identification code (a or b) of the serial interface SIF. Reference numeral m indicates a number of the cluster 100 to be connected, and reference numeral n indicates an identification code (a or b) of the serial interface SIF.

For example, the interface control units MM0a and MM0b of cluster # 0 are connected to the interface control units SM0a and SM0b of SSU # 0. The interface control units MM3a and MM3b of cluster # 0 are connected to the interface control units SM0a and SM0b of SSU # 3 (not shown). The interface control units SM7a and SM7b of SSU # 0 are connected to the interface control units MM0a and MM0b of cluster # 7 (not shown).

FIG. 6 shows an example of the internal configuration of clusters # 0 and SSU # 0 in FIG. The same elements as those in FIG. 1 are designated by the same reference numerals as those in FIG. 1, and detailed description thereof will be omitted. Hereinafter, the internal configurations of clusters # 0 and SSU # 0 will be described, but the internal configurations of other clusters # 1 to # 7 and other SSUs # 0 to # 3 are the same as those in FIG.

The CPU 110 of cluster # 0 has a core unit 200 that executes calculation processing, and four sets of interface control units MMij connected to each of SSU # 0 to # 3. The core unit 200 and each interface control unit MMij are connected to each other via a communication interface DIF for degeneracy control. The interface control units MMia and MMib are connected to each other via the communication interface RIF for recovery control. The two lines of the communication interface RIF surrounded by the broken line frame indicate the communication interfaces RIFa and RIFb, respectively.

The ASIC 510 of SSU # 0 has a core unit 600 that controls reading and writing of data to and from the storage device 530, and eight sets of interface control units SMmn connected to each of clusters # 0 to # 7. The core unit 600 and each interface control unit MMmn are connected to each other via a communication interface DIF for degeneracy control. The interface control units MMma and MMmb are connected to each other via a communication interface RIF for recovery control.

FIG. 7 shows an example of each interface control unit 300 provided in the cluster 100 of FIG. In FIG. 7, reference numeral FF indicates a flip-flop circuit, and reference numeral SYNC indicates a synchronization circuit for synchronizing a clock with an interface control unit 300 connected via a communication interface RIF. Examples of the operation of the interface control unit 300 provided in the cluster 100 are shown in FIGS. 10 to 12.

The interface control unit 300 includes a recovery packet transmission unit 402, a recovery packet generation unit 404, a normal packet transmission unit 406, a packet insertion unit 408, and a communication interface unit 410 (PHY). The interface control unit 300 includes a recovery sequence control unit 412, an enable unit 414, an error detection unit 416, a packet determination unit 418, a recovery packet reception unit 420, and a normal packet reception unit 422. The interface control unit 300 includes a degeneracy instruction unit 426, a core notification unit 428, an inter-module control unit 430, and an or circuit 432.

The recovery packet transmission unit 402 sends a link down notification or a link recovery notification received from the adjacent interface control unit 300 via the communication interface RIF to the opposite SSU 500 (destination), so that the recovery packet generation unit 404 Output to. For example, the link down notification and the link recovery notification are recovery control instructions (*** _ req_send) described later. As shown in the legend at the lower right of FIG. 6, the symbol *** is any of interface_disable, reset, lock_detect, interface_enable or link_down.

interface_disable indicates suppression of the corresponding serial interface SIF (disable PHY410 (link)). reset indicates a reset of a circuit such as PHY410 that controls the serial interface SIF. lock_detect indicates confirmation of link-up of serial interface SIF (detection of lock status of PHY410). interface_enable indicates the activation of the serial interface SIF (enable PHY410 (link)). link_down indicates the link down of the serial interface SIF.

The recovery packet generation unit 404 generates a recovery packet using the recovery control instruction from the recovery packet transmission unit 402, and outputs the generated recovery packet to the packet insertion unit 408. The recovery packet transmission unit 402 and the recovery packet generation unit 404 are examples of the recovery control transmission unit.

The normal packet transmission unit 406 has a RAM (SYNC) that holds a normal packet (for example, 8-byte wide data) received from the core unit 200 via the communication interface DIF. The normal packet transmission unit 406 sequentially outputs the normal packets held in the RAM (SYNC) to the packet insertion unit 408.

The packet insertion unit 408 sequentially outputs normal packets received from the normal packet transmission unit 406 to the PHY 410. When the packet insertion unit 408 receives the recovery packet from the recovery packet generation unit 404, the packet insertion unit 408 inserts the recovery packet between the outputs of the normal packet and outputs the recovery packet to the PHY 410. As a result, the recovery packet can be transmitted to the SSU500 while transmitting and receiving the normal packet using the degenerate link. The recovery packet will be described with reference to FIG.

The communication interface unit 410 (PHY) has a function as a so-called physical layer. The PHY 410 converts the packet (parallel data) transferred from the packet insertion unit 408 into serial data and transmits it to the destination SSU 500 via the serial interface SIF. Further, the PHY 410 converts the serial data received from the SSU 500 via the serial interface SIF into parallel data and outputs the serial data to the enable unit 414.

When the PHY410 receives the reset signal reset (tx / rx) from the recovery sequence control unit 412, the PHY410 resets and initializes the internal circuit. The reset of the internal circuit is executed by one or both of the transmitting unit (tx) and the receiving unit (rx) depending on the content of the reset signal reset. Further, the PHY 410 outputs the lock state of the link as lock_status to the recovery sequence control unit 412.

When the link down of the serial interface SIF connected to the PHY 410 is detected, the recovery sequence control unit 412 executes the recovery sequence of the link. The recovery sequence is executed when a link down is detected (link_down_det) or when a link down notification (link_down_req_send) is received from the opposing SSU 500 via the inter-module control unit 430. The recovery sequence control unit 412 is an example of a recovery control unit that controls the recovery process of the link.

When the interface enable signal (interface_enable) output from the recovery sequence control unit 412 indicates an effective level, the enable unit 414 outputs a packet output from the PHY 410 to the error detection unit 416. When the interface enable signal output from the recovery sequence control unit 412 indicates an invalid level, the enable unit 414 suppresses the output of the packet output from the PHY 410 to the error detection unit 416.

The error detection unit 416 detects an error included in the packet received from the opposite SSU500. When the error detection unit 416 detects a link down, the error detection unit 416 outputs a link down detection signal link_down_det to the recovery sequence control unit 412 and the or circuit 432. Further, the error detection unit 416 outputs the packet received from the SSU 500 to the packet determination unit 418.

When the packet from the error detection unit 416 shows a recovery response (*** _ end_recv), the packet determination unit 418 outputs the packet to the recovery packet reception unit 420. When the packet from the error detection unit 416 indicates a link down notification (link_down_req_recv), the packet determination unit 418 outputs the packet to the recovery sequence control unit 412.

The recovery packet receiving unit 420 outputs the recovery response (*** _ end_recv) determined by the packet determination unit 418 to the adjacent interface control unit 300 via the communication interface RIF. The recovery packet receiving unit 420 of the interface control unit 300 is an example of a reception notification unit.

The normal packet receiving unit 422 has a RAM (SYNC) for holding normal packets sequentially received from the SSU 500 via the PHY 410. The normal packet receiving unit 422 outputs a packet (for example, 8-byte wide data) held in the RAM (SYNC) to the core unit 200 via the communication interface DIF.

The or circuit 432 outputs the link-down detection signal link_down_det from the error detection unit 416 or the link-down reception signal link_down_recv from the recovery sequence control unit 412 to the degeneration instruction unit 426.

The degeneracy instruction unit 426 instructs the core unit 200 to degenerate the link via the communication interface DIF based on the link down detection signal link_down_det or the link down reception signal link_down_recv from the or circuit 432. That is, the degeneracy instruction unit 426 cores the degeneracy of the link when the error detection unit 416 detects the link down or based on the notification of the link down from the opposite interface control unit 700 connected via the serial interface SIF. Instruct unit 200.

The core notification unit 428 outputs the restart signal re_start output by the recovery sequence control unit 412 after the link restoration process to the core unit 200 via the communication interface DIF. Further, the core notification unit 428 outputs the link blockage signal link_quiet, which is output when the recovery sequence control unit 412 fails the link recovery process, to the core unit 200 via the communication interface DIF.

The inter-module control unit 430 outputs a recovery control instruction ***_req_send output by the recovery sequence control unit 412 at the time of link recovery processing to the adjacent interface control unit 300 via the communication interface RIF. The inter-module control unit 430 outputs a recovery response (*** _ end_recv) received from the adjacent interface control unit 300 to the recovery sequence control unit 412 via the communication interface RIF. The inter-module control unit 430 outputs a link down notification (link_down_req_recv) received from the adjacent interface control unit 300 to the recovery sequence control unit 412 via the communication interface RIF.

FIG. 8 shows an example of the interface control unit 700 provided in the SSU 500 of FIG. The same elements as those in FIG. 7 are designated by the same reference numerals, and detailed description thereof will be omitted. Examples of the operation of the interface control unit 700 provided in the SSU 500 are shown in FIGS. 10 to 12.

The interface control unit 700 has a recovery processing unit 413 instead of the recovery sequence control unit 412 of the interface control unit 300 of FIG. 7, and has a PHY710 instead of the PHY410 of the interface control unit 300 of FIG. The function of PHY710 is the same as that of PHY410.

Further, the interface control unit 700 has a link-down notification unit 424 which is not included in the interface control unit 300. The other block configurations of the interface control unit 700 are the same as the block configurations of the interface control unit 300. Hereinafter, a block having a function different from that of the interface control unit 300 will be described.

The recovery packet transmission unit 402 is a recovery packet generation unit for transmitting a link down notification or a link recovery response received from the adjacent interface control unit 700 via the communication interface RIF to the opposite cluster 100 (destination). Output to 404. Here, the link down notification is link_down_req_send, and the recovery response is ***_end_send.

When the packet insertion unit 408 receives the link down notification packet or the link recovery response packet from the recovery packet generation unit 404, the packet insertion unit 408 inserts the link down notification packet or the link recovery response packet between the outputs of the normal packets. Output to PHY410. This makes it possible to send a link down notification packet or a link recovery response packet to the SSU 500 while sending and receiving normal packets using the degenerate link.

The link-down notification unit 424 receives the link-down detection signal link_down_det output by the error detection unit 416 that has detected the link-down as the link-down notification link_down_req_send. The link-down notification unit 424 outputs the received link-down notification link_down_req_send to the adjacent interface control unit 700 via the communication interface RIF.

The link-down notification unit 424 can notify the interface control unit 300 of the opposing cluster 100 via the serial interface SIF that operates normally, the link-down detected by the interface control unit 700. The interface control unit 300 notified of the link down notifies the adjacent interface control unit 300 in which the link down has occurred, and causes the link restoration process to be executed.

The interface control unit 700 does not receive a recovery response (*** _ end_recv) from the interface control unit 300 of the opposing cluster 100. Therefore, the packet determination unit 418 does not determine the recovery response (*** _ end_recv), but determines the recovery control instruction (*** _ req_recv) and outputs it to the recovery packet receiving unit 420. The recovery packet receiving unit 420 outputs the received recovery control instruction (*** _ req_recv) to the adjacent interface control unit 700 via the communication interface RIF.

The inter-module control unit 430 outputs a link down notification (link_down_req_send) output from the recovery processing unit 413 when a link down is detected to the adjacent interface control unit 700 via the communication interface RIF. The inter-module control unit 430 outputs a recovery control instruction (*** _ req_recv) output from the adjacent interface control unit 700 to the recovery processing unit 413 via the communication interface RIF. The inter-module control unit 430 outputs a recovery response (*** _ end_send) output by the recovery processing unit 413 at the time of link recovery processing to the adjacent interface control unit 700 via the communication interface RIF.

When the error detection unit 416 detects a link down, the recovery processing unit 413 executes control to notify the opposing cluster 100 of the link down (link_down_req_send). Further, the recovery processing unit 413 executes the link recovery processing based on the recovery control instruction (*** _ req_recv) output from the recovery sequence control unit 412 of the opposing cluster 100. Then, the recovery processing unit 413 outputs a recovery response (*** _ end_send) corresponding to the recovery control instruction (*** _ req_recv). Other functions of the recovery processing unit 413 are the same as those of the recovery sequence control unit 412 of the interface control unit 300, except that the link recovery processing is not voluntarily executed.

FIG. 9 shows an example of the operation of the information processing system SYS2 of FIG. That is, FIG. 9 shows an example of a control method of the interface control device (300a, 300b, 700a, 700b) mounted on the information processing system SYS2. In the initial state of FIG. 2, the information processing devices 10 and 20 execute packet transmission / reception using both the serial interfaces SIFa and SIFb (normal mode). FIG. 2 describes an example in which the serial interface SIFa is linked down.

First, in step S10, either the error detection unit 416 of the interface control unit 300a (cluster 100) or the interface control unit 700a (SSU500) connected to the serial interface SIFa detects the link down. Next, in step S12, the cluster 100 or SSU500 that detects the link down retracts the linked down serial interface SIF.

Next, in step S14, if the cluster 100 detects a link down, the process proceeds to step S18, and if the SSU500 detects a link down, the process proceeds to step S16. In step S16, the SSU 500 notifies the cluster 100 of the link down and proceeds to step S18.

In step S18, the recovery sequence control unit 412 of the cluster 100 starts the recovery process, and causes the recovery packet generation unit 404 to generate a recovery control instruction packet. When the link down notification is received from the SSU 500, the cluster 100 degenerates the linked down serial interface SIF.

When either the cluster 100 or the SSU500 detects a link down, the cluster takes the lead in executing the link recovery process. As a result, the cluster 100 and the SSU 500 can start the link-up process independently of each other, and can prevent the link-up protocol from being broken and being unable to link up. Here, the link-up protocol indicates that the serial interface is controlled in the order of interface disable, initialization, setting, and link establishment.

The cluster 100 transmits a link down notification to the SSU 500 as the first recovery control instruction packet of the recovery process. The recovery control instruction packet is inserted between normal packets by using the non-degenerate serial interface SIF.

Next, in step S20, the recovery processing unit 413 of the SSU 500 executes recovery processing based on the recovery control instruction packet received from the cluster 100, and transmits a recovery response to the cluster 100. The recovery processing unit 413 degenerates the linked down serial interface at the beginning of the recovery processing.

Next, in step S22, the cluster 100 shifts to step S24 when all the recovery processes are completed, and returns the process to step S18 when there is an uncompleted recovery process. In steps S18 and S20, the cluster 100 and the SSU500 proceed with the recovery process one by one by handshaking, so that the recovery process of the cluster 100 and the SSU500 can be executed in synchronization with each other.

In step S24, the cluster 100 releases the degeneracy of the serial interface SIF that has been linked down, and completes the recovery process. The SSU500 releases the degeneracy of the serial interface SIF that has been linked down based on the reception of the last recovery control instruction packet (interface_enable_req_send).

FIG. 10 shows an example of processing at the time of link down detection in the information processing system SYS2 of FIG. That is, FIG. 10 shows an example of a control method of the interface control device (300a, 300b, 700a, 700b) mounted on the information processing system SYS2. In the example shown in FIG. 10, in the packet transfer between the cluster # 0 and the SSU # 0, either the interface control unit MM0a of the cluster # 0 or the interface control unit SM0a of the SSU # 0 detects the link down.

In addition, in FIG. 10 and FIGS. 11 and 12 described later, the recovery sequence when a link down occurs between the interface control units MM0b and SM0b is described by replacing the reference numerals MM0a and MM0b and replacing the reference numerals SM0a and SM0b. NS. Also, the link-down recovery sequences that occurred between cluster # x (x is any of 0 to 3) and SSU # y (y is any of 0 to 7) are also shown in FIGS. 10, 11, It is the same as FIG.

In step S30, when the interface control unit SM0a of SSU # 0 detects the link down of the own serial interface SIFa, the interface control unit SM0a performs step S32. When the interface control unit SM0a of SSU # 0 detects a link down, it issues a degeneracy instruction to the core unit 600. In step S32, the interface control unit SM0a notifies the interface control unit SM0b connected via the communication interface RIF of the link down.

Upon receiving the link down notification, the interface control unit SM0b inserts the link down notification packet between the normal packets in step S34. Then, the interface control unit SM0b transmits a link-down notification packet to the interface control unit MM0b of the opposing cluster # 0 via the normal serial interface SIFb.

Upon receiving the link-down notification packet, the interface control unit MM0b transfers the link-down notification packet to the interface control unit MM0a connected via the communication interface RIF in step S36.

On the other hand, in step S38, the interface control unit MM0a executes the link down detection process of the own serial interface SIFa. The interface control unit MM0a executes step S40 when it detects a link down of its own serial interface SIFa or receives a link down notification packet from the interface control unit MM0b. In step S40, the interface control unit MM0a issues a degeneracy instruction to the core unit 200 and starts the restoration process shown in FIG.

FIG. 11 shows an example of the link restoration process in the information processing system SYS2 of FIG. That is, FIG. 11 shows an example of a control method of the interface control device (300a, 300b, 700a, 700b) mounted on the information processing system SYS2. First, in step S42, whether the interface control unit MM0a has an error frequency (error rate) higher than a predetermined frequency in communication using the serial interface SIFa with the interface control unit SM0a of the opposing SSU # 0 in step S42. Judge whether or not. For example, the frequency of errors is determined by the error detection unit 416.

When the frequency of the error is higher than the predetermined frequency, the interface control unit MM0a notifies the core unit 200 of the abnormality, executes the abnormality processing, and ends the link restoration processing. When the frequency of the error is less than or equal to the predetermined frequency, the interface control unit MM0a notifies the interface control unit MM0b of the recovery control instruction in step S44. After that, the interface control unit MM0a executes a recovery sequence for recovering its own link in step S46, and shifts the process to step S60.

The interface control unit MM0b, which has received the recovery control instruction via the communication interface RIF, inserts a dedicated notification packet for the recovery control instruction between the normal packets in step S48. Then, the interface control unit MM0b transmits a recovery control instruction notification packet to the interface control unit SM0b of the opposite SSU # 0 via the normal serial interface SIFb.

Upon receiving the recovery control instruction notification packet, the interface control unit SM0b transfers the recovery control instruction notification packet to the interface control unit SM0a connected via the communication interface RIF in step S50.

Upon receiving the recovery control instruction via the communication interface RIF, the interface control unit SM0a executes a recovery sequence for recovering its own link in step S52, and shifts the process to step S54. In step S54, the interface control unit SM0a notifies the interface control unit SM0b of the completion of the recovery sequence executed based on the recovery control instruction.

Upon receiving the notification of the completion of the recovery sequence, the interface control unit SM0b inserts the recovery sequence completion response packet between the normal packets in step S56. Then, the interface control unit SM0b transmits a completion response packet to the interface control unit MM0b of the opposing cluster # 0 via the normal serial interface SIFb.

In step S58, the interface control unit MM0b that has received the completion response packet of the recovery sequence transfers the completion response notification packet to the interface control unit MM0a connected via the communication interface RIF.

In step S60, the interface control unit MM0a waits for the completion response packet, executes step S62 if the completion response packet is not received, and executes step S64 if the completion response packet is received. In step S62, the interface control unit MM0a determines whether or not a timeout has occurred without receiving the completion response packet. When a timeout occurs, the interface control unit MM0a notifies the core unit 200 of the abnormality, executes the abnormality processing, and ends the link restoration processing. If no timeout has occurred, the interface control unit MM0a returns to step S60.

In step S64, the interface control unit MM0a executes step S66 when all the recovery sequences are completed, returns to step S44 when there is a recovery sequence that has not been executed, and issues the next recovery control instruction to the interface control unit MM0b. Notify to. In this way, the interface control unit MM0a executes handshake control with the interface control unit SM0a for issuing the next recovery control instruction after waiting for the recovery completion response for each processing of the recovery sequence. The link-down link recovery process can be steadily executed between cluster # 0 and SSU # 0 according to the link-up protocol.

In step S66, the interface control unit MM0a instructs the core unit 200 to release the degeneracy of the link, and ends the restoration process.

From the above, when a link down occurs, cluster # 0 takes the lead, generates a dedicated packet for the recovery sequence according to the link-up protocol, interrupts the generated dedicated packet between the normal packets, and SSU # 0. Send to. This allows the normal serial interface SIF to be used to send recovery control instructions for link recovery to SSU # 0 with handshake control, led by cluster # 0 and remotely SSU #. The 0 link can be restored.

In the information processing system SYS2 shown in FIGS. 4 and 5, a maximum of eight clusters 100 are connected to each SSU500. Therefore, a maximum of eight serial interface SIFs may be linked down at the same time. For example, when the link restoration process is executed using the firmware executed by the service processor of the management device 520, the links are sequentially restored for each serial interface SIF.

If it takes a maximum of 3 seconds to restore the link of one serial interface SIF, it takes a maximum of 24 seconds to restore the eight serial interface SIFs. As a result, the degenerate state of the link lasts for up to 24 seconds. During the link recovery process, the data transfer performance is halved due to the degeneracy of the link, resulting in a long-term performance degradation. Furthermore, if the serial interface SIF is degenerated for a long time, the risk of system down increases. This is because if the degenerate interface is linked down, data transfer cannot be continued.

For example, in the serial interface SIF, the required value of the bit error rate (BER), which is the ratio of the number of error bits to the number of received bits, is ^10-12 or less, and an error occurs once every 100 seconds in consideration of the data transfer speed. Suppose it occurs. In this case, in the two serial interface SIFs, an error occurs once every 40 seconds, and in 24 seconds (degenerate period), an error occurs about 0.5 times.

On the other hand, in this embodiment, the link-down recovery process of the serial interface SIF can be executed independently for each connection between the cluster 100 and the SSU500. Further, the link restoration process is executed using the hardware of the cluster 100 and the SSU 500 (interface control units 300, 700). Therefore, the link down dissemination process can be completed in, for example, several microseconds, and the degeneracy period of the link can be set to several microseconds. As a result, there is almost no possibility that the serial interface SIF that sends and receives packets during the degenerate period will link down.

FIG. 12 shows an example of the recovery sequence of FIG. That is, FIG. 12 shows an example of a control method of the interface control device (300a, 300b, 700a, 700b) mounted on the information processing system SYS2. Detailed description of the same processing as in FIGS. 9 to 11 will be omitted. FIG. 12 shows an example in which the serial interface SIFa between cluster # 0 and SSU # 0 is linked down, and the interface control unit SM0a of SSU # 0 detects the link down. In FIG. 12, the packet transmitted / received between cluster # 0 and SSU # 0 is a recovery packet for recovering the link.

First, the interface control unit SM0a of SSU # 0 outputs the link down detection signal link_down_det to the interface control unit SM0b based on the detection of the link down. The interface control unit SM0b transmits a link down notification packet (link_down_req_send) to the opposite interface control unit MM0b based on the reception of the link down detection signal link_down_det. The interface control unit MM0b transmits a link down notification (link_down_req_recv) to the interface control unit MM0a based on the link down notification packet (link_down_req_send).

Upon receiving the link down notification (link_down_req_recv), the interface control unit MM0a recognizes the link down in SSU # 0 and starts the sequence control of the link restoration process in step S70. The interface control unit MM0a outputs a link down reception signal link_down_det (degeneration instruction) to the core unit 200.

The core unit 200 starts the degeneracy process of the link, and in order to guarantee the data lost due to the link down, the data transfer when the link down occurs is restarted from the beginning. This makes it possible to prevent data loss due to link down. When the interface control unit MM0a detects the link down by itself, the interface control unit MM0a starts step S70 without receiving the link down notification (link_down_req_recv).

Next, in step S72, the interface control unit MM0a gives an instruction (link_quiet) to forcibly block the link and prohibit the use of the link when the error frequency (error rate) is higher than the predetermined frequency. Output to 200 and end the recovery sequence. When the frequency of the error is less than or equal to the predetermined frequency, the interface control unit MM0a executes step S74.

In step S74, the interface control unit MM0a outputs an instruction (interface enable off) for disabling the link to the PHY 410. Further, the interface control unit MM0a instructs the interface control unit MM0b to generate a recovery control packet for disabling the link and transmit it to SSU # 0.

The interface control unit MM0b generates a recovery control packet (interface_disable_req_send), interrupts it between normal packets, and transmits it to the interface control unit MM0b of SSU # 0. The interface control unit SM0b notifies the interface control unit SM0a of a recovery instruction (interface_disable_req_recv) based on the reception of the recovery control packet (interface_disable_req_send).

Upon receiving the recovery instruction (interface_disable_req_recv), the interface control unit SM0a outputs an instruction to disable the link (interface enable off) to the enable unit 414. The interface control unit SM0a notifies the interface control unit SM0b of the completion response notification of link invalidation. The interface control unit SM0b generates a completion response packet (interface_disable_end_send) for disabling the link based on the completion response notification, and transmits the completion response packet (interface_disable_end_send) to the interface control unit SM0b of the cluster # 0.

The interface control unit SM0b transmits a completion response notification (interface_disable_end_recv) to the interface control unit SM0a based on the reception of the completion response packet (interface_disable_end_send). If the interface control unit MM0a receives the completion response notification (interface_disable_end_recv) in step S76 before the timeout indicating the elapse of a predetermined time occurs, the interface control unit MM0a executes the reset process in step S78. When a timeout occurs, the interface control unit MM0a forcibly shuts down the link, outputs an instruction (link_quiet) prohibiting the use of the link to the core unit 200, and ends the recovery sequence.

In this way, the interface control unit MM0a monitors the timeout in each sequence of the recovery process by the handshake, and when the timeout occurs, forcibly closes the link and prohibits the use of the link. As a result, in the unlikely event that the link restoration process does not proceed due to a defect in the interface control units SM0a, SM0b, MM0b or a defect in the serial interface SIFb after degeneration, the restoration process can be interrupted.

In step S78, the interface control unit MM0a outputs a link reset instruction (reset (tx / rx)) to the PHY 410. In this example, the interface control unit MM0a instructs to reset both the transmission system (tx) and the reception system (rx). However, the interface control unit MM0a may instruct to reset only the abnormal system among the transmission system (tx) and the reception system (rx). Further, the interface control unit MM0a instructs the interface control unit MM0b to generate a recovery control packet for resetting the link and transmit it to SSU # 0.

The interface control unit MM0b generates a recovery control packet (reset_req_send), interrupts it between normal packets, and transmits it to the interface control unit SM0b of SSU # 0. The interface control unit SM0b notifies the interface control unit SM0a of a recovery instruction (reset_req_recv) based on the reception of the recovery control packet (reset_req_send).

Upon receiving the restoration instruction (reset_req_recv), the interface control unit SM0a outputs the reset instruction (reset (tx / rx)) to the PHY710. The interface control unit SM0a notifies the interface control unit SM0b of the completion response notification of the reset of the PHY710. The interface control unit SM0b generates a reset completion response packet (reset_end_send) of the PHY710 based on the completion response notification, and transmits the completion response packet (reset_end_send) to the interface control unit MM0b of the cluster # 0.

The interface control unit MM0b transmits a completion response notification (reset_end_recv) to the interface control unit MM0a based on the reception of the completion response packet (reset_end_send). If the interface control unit MM0a receives the completion response notification (reset_end_recv) in step S80 before the timeout occurs, the interface control unit MM0a executes the lock status confirmation process in step S82. When a timeout occurs, the interface control unit MM0a forcibly shuts down the link, outputs an instruction (link_quiet) prohibiting the use of the link to the core unit 200, and ends the recovery sequence.

In step S82, the interface control unit MM0a makes, for example, an inquiry (reset (rx) on / off lock status) of the lock status for determining whether the serial interface SIFa is normally linked up to the PHY 410. A lock status inquiry for determining whether the transmission system (tx) has been reset normally may be executed.

The lock status query is repeated until a lock is detected. Then, the interface control unit MM0a instructs the interface control unit MM0b to generate a recovery control packet for detecting the lock status and transmit it to SSU # 0.

The interface control unit MM0b generates a recovery control packet (lock_detect_req_send), interrupts it between normal packets, and transmits it to the interface control unit SM0b of SSU # 0. The interface control unit SM0b notifies the interface control unit SM0a of a recovery instruction (lock_detect_req_recv) based on the reception of the recovery control packet (lock_detect_req_send).

Upon receiving the recovery instruction (lock_detect_req_recv), the interface control unit SM0a makes a lock status inquiry (reset (rx) on / off lock status) to the PHY710 to determine whether the receiving system (rx) has been reset normally. A lock status inquiry for determining whether the transmission system (tx) has been reset normally may be executed.

When the interface control unit SM0a confirms that the link has been reset normally (lock detection), the interface control unit SM0a notifies the interface control unit SM0b of the completion response notification of the lock detection. The interface control unit SM0b generates a lock detection completion response packet (lock_detect_end_send) based on the completion response notification, and transmits the lock detection completion response packet (lock_detect_end_send) to the interface control unit MM0b of cluster # 0.

The interface control unit MM0b transmits a completion response notification (lock_detect_end_recv) to the interface control unit MM0a based on the reception of the completion response packet (lock_detect_end_send). If the interface control unit MM0a receives the completion response notification (lock_detect_end_recv) in step S84 before the timeout occurs, the interface control unit MM0a executes the lock detection determination process in step S86. When a timeout occurs, the interface control unit MM0a forcibly blocks the link, outputs an instruction (link_quiet) prohibiting the use of the link to the core unit 200, and ends the recovery sequence.

In step S86, the interface control unit MM0a executes step S88 when the link-up of the serial interface SIFa by PHY410 is successful and the link-up of the serial interface SIFa by PHY710 is successful. On the other hand, when the link-up is not successful in step S82, the interface control unit MM0a outputs an instruction (link_quiet) for forcibly blocking the link and prohibiting the use of the link to the core unit 200, and outputs a recovery sequence to the core unit 200. finish.

In step S88, the interface control unit MM0a outputs an instruction (interface enable on) for enabling the serial interface SIFa to the enable unit 414. Further, the interface control unit MM0a instructs the interface control unit MM0b to generate a recovery control packet for enabling the serial interface SIFa on the PHY710 side and transmit the recovery control packet to SSU # 0.

The interface control unit MM0b generates a recovery control packet (interface_enable_req_send), interrupts it between normal packets, and transmits it to the interface control unit MM0b of SSU # 0. The interface control unit SM0b notifies the interface control unit SM0a of a recovery instruction (interface_enable_req_recv) based on the reception of the recovery control packet (interface_enable_req_send).

Upon receiving the recovery instruction (interface_enable_req_recv), the interface control unit SM0a outputs an instruction (interface enable on) for enabling the serial interface SIFA to the PHY710. The interface control unit SM0a notifies the interface control unit SM0b of the completion response notification of link enable. The interface control unit SM0b generates a completion response packet (interface_enable_end_send) for enabling the link based on the completion response notification, and transmits the completion response packet (interface_enable_end_send) to the interface control unit MM0b of the cluster # 0.

Although not shown, the interface control unit SM0a gives a restart instruction (re_start on) to the core unit 600 in order to restore the link from the degenerate operation state to the normal operation state after the serial interface SIFa is enabled. Is output. Upon receiving the restart instruction, the core unit 600 restores the link from the degenerate state.

The interface control unit MM0b transmits a completion response notification (interface enable_end_recv) to the interface control unit MM0a based on the reception of the completion response packet (interface_enable_end_send). If the interface control unit MM0a receives the completion response notification (interface enable_end_recv) in step S76 before the timeout occurs, the interface control unit MM0a executes the recovery process from the degeneration in step S92. When a timeout occurs, the interface control unit MM0a forcibly shuts down the link, outputs an instruction (link_quiet) prohibiting the use of the link to the core unit 200, and ends the recovery sequence.

In step S92, the interface control unit MM0a outputs a restart instruction (re_start on) to the core unit 200 in order to restore the link from the degenerate state, and completes the link restoration process shown in FIG. The core unit 200 restores the link from the degenerate operation state to the normal operation state based on the restart instruction (re_start on). As a result, packet communication using both the serial interfaces SIFa and SIFb between cluster # 0 and SSU # 0 is restarted.

As shown in FIG. 12, in the link recovery process, cluster # 0 executes the process for each of a plurality of sequences included in the recovery sequence, and issues a recovery control instruction to SSU # 0. Then, cluster # 0 executes the next sequence of processing based on the completion of its own processing and the recovery response from SSU # 0. As a result, the recovery sequence can be steadily executed one by one under the initiative of cluster # 0, and the possibility of link recovery failure can be reduced.

FIG. 13 shows an example of the format of packets sent and received between the cluster 100 and the SSU 500 of FIG. The data packet (normal packet) transmitted / received by the system operation of the information processing system SYS2 has a header unit, a data unit, a CRC (Cyclic Redundancy Check) unit, and a termination unit. The recovery packet used in the recovery sequence described with reference to FIG. 12 has a header part, a CRC part, and a termination part.

The header portion of the data packet has 8 bytes of bytes 0 to 7. Byte 0 is set to 0xFB, bytes 1 to 5 and byte 7 are set to arbitrary values, and byte 6 is set to an operation code (other than 0x02). Here, the code 0x indicates that the value is a hexadecimal number.

The header portion of the recovery packet has 8 bytes of bytes 0 to 7. Byte 0 is 0xFB, Byte 1, Byte 4, Byte 5 and Byte 7 are arbitrary values, Byte 2 is the control bit for control instruction, Byte 3 is the control bit for response notification, and Byte 6 is the operation code (0x02). Set. Byte 2 in the header portion of the recovery packet is used in the recovery control packet transmitted from the cluster 100 to the SSU500. Byte 3 in the header portion of the recovery packet is used in the completion response packet transmitted from the SSU 500 to the cluster 100.

In byte 2 (for control instruction) of the header part of the recovery packet, bit 0 is set when link down is detected, bit 2 is set when link is disabled, and bit 3 is set when resetting PHY710. Is set to. Further, in byte 2, bit 4 is set when confirming the lock status, and bit 5 is set when enabling the link. Arbitrary values are set for bit 1, bit 6 and bit 7 of byte 2.

In byte 3 (for response notification) of the header portion of the recovery packet, bit 0 is set to indicate that it is a completion response packet. In byte 3, bit 2 is set when the lock status is on, and bit 3 is set when the lock status is in the fail state (unlocked state). Arbitrary values are set for bit 1 of byte 3 and bits 4 to 7.

As described above, even in the embodiments shown in FIGS. 4 to 13, the same effects as those in the embodiments shown in FIGS. 1 to 3 can be obtained. For example, the link can be restored more quickly and the link down period can be minimized as compared with the case where the link is restored by the firmware executed by the service processor mounted on the cluster 100 and the SSU 500. Further, the link down recovery process of the serial interface SIF can be independently executed for each connection unit between the cluster 100 and the SSU 500. Further, the link restoration process can be executed by the hardware of the cluster 100 and the SSU 500 (interface control units 300, 700). As a result, the period of deterioration in the performance of the information processing system SYS2 can be minimized.

For example, the link recovery time (several microseconds) when the present embodiment is applied can be significantly shortened as compared with the link recovery time (maximum 24 seconds) by the firmware. Therefore, when the serial interface SIF is degenerated, the risk of system down can be significantly reduced.

Further, in the embodiments shown in FIGS. 4 to 13, the following effects can be obtained. For example, when a link down occurs, the cluster 100 takes the lead in generating a dedicated packet for the recovery sequence according to the link-up protocol, interrupting the generated dedicated packet between the normal packets, and transmitting the generated dedicated packet to the SSU500. As a result, the recovery control instruction for recovering the link can be transmitted to the SSU500 by handshake control using the normal serial interface SIF, and the cluster 100 takes the initiative to remotely transfer the link on the SSU500 side. It can be restored.

As described with reference to FIG. 12, the cluster 100 and the SSU500 advance the recovery processing sequence one by one by handshaking, so that the recovery processing of the cluster 100 and the SSU500 can be executed in synchronization with each other. As a result, the link restoration process can be steadily executed between the cluster 100 and the SSU 500. By steadily executing the recovery sequence one by one under the initiative of the cluster 100, the possibility of link recovery failure can be reduced.

The interface control unit 300 monitors the timeout in each sequence of the recovery process by the handshake, and when the timeout occurs, forcibly closes the link and prohibits the use of the link. As a result, in the unlikely event that the link restoration process does not proceed due to a defect in the interface control unit 700 or a defect in the serial interface SIF after the degeneration, the restoration process can be interrupted.

The packet insertion unit 408 of the interface control unit 300 can insert a recovery packet between the outputs of the normal packet and transmit it to the SSU 500. Similarly, the packet insertion unit 408 of the interface control unit 700 can insert a link down notification packet or a link recovery response packet between the outputs of the normal packet and transmit it to the cluster 100. Therefore, in the cluster 100 and SSU500 connected via the two serial interfaces SIFa and SIFb, even if the link is degenerated, the recovery packet, the recovery response packet, and the like are transmitted and received using the live link. Can be done. As a result, when a link down occurs, it is possible to execute a link recovery process using the degenerate link while continuing to send and receive normal data.

FIG. 14 shows an example of an information processing system according to another embodiment. The same elements as those in the above-described embodiment are designated by the same reference numerals, and detailed description thereof will be omitted. In the information processing system SYS3 shown in FIG. 14, the core unit 200 of each cluster 100 and the core unit 600 of each SSU500 are connected via a dedicated network (for example, LAN) different from the management bus MBUS. Then, the recovery control instruction packet and the completion response packet are transmitted and received via the dedicated network. Other configurations and functions of the information processing system SYS3 are the same as those of the information processing system SYS2 shown in FIGS. 4 to 13.

In this embodiment, the interface control unit 300 (300a or 300b) that has detected the link down in the cluster 100 transmits a recovery control instruction packet to the SSU 500 and receives a completion response packet from the SSU 500 via the core unit 200. .. Further, the interface control unit 700 (700a or 700b) that has detected the link down in the SSU 500 transmits a link down notification packet to the cluster 100 via the core unit 600.

In this embodiment, the transmission / reception of the recovery packet (FIG. 13) using the normal serial interface SIF that is not linked down is not executed. Therefore, it is possible to eliminate the process of transmitting and receiving recovery packets to the interface control units 300 and 700 connected to the normal serial interface SIF. As a result, the recovery sequence for recovering the link can be easily executed as compared with the recovery sequence shown in FIGS. 10 to 12, and the time required for the link recovery process can be shortened.

As described above, even in the embodiment shown in FIG. 14, the same effect as that of the embodiment shown in FIGS. 1 to 13 can be obtained. Further, in the embodiment shown in FIG. 14, the recovery packet can be easily transmitted and received between the cluster 100 and the SSU 500, and the time required for the link recovery process can be shortened.

FIG. 15 shows an example of an information processing system according to another embodiment. The same elements as those in the above-described embodiment are designated by the same reference numerals, and detailed description thereof will be omitted. In the information processing system SYS4 shown in FIG. 15, each cluster 100 and each SSU500 are connected via four interface control units (300a to 300d and 700a to 700d in FIG. 15).

Then, in each cluster 100, the recovery packet (link down notification or link recovery notification) is transmitted and received using a non-busy serial interface SIF (any of SIFa, SIFb, SIFc, and SIFd). Therefore, the core unit 200 has a function of a router that determines a transfer route of the recovery packet. Then, the core unit 200 transfers, for example, an instruction from the interface control unit 300 connected to the linked down serial interface SIF to the interface control unit 300 connected to the normal serial interface SIF. Note that each cluster 100 may send and receive recovery packets using a serial interface SIF that is not busy or linked down.

In each SSU500, the recovery response packet (link down notification packet or link recovery response packet) is transmitted and received using a non-busy serial interface SIF (either SIFa, SIFb, SIFc, or SIFd). Here, the non-busy serial interface SIF is the serial interface SIF that has received the recovery packet from the cluster 100. Then, the core unit 600 transfers, for example, an instruction from the interface control unit 700 connected to the linked down serial interface SIF to the interface control unit 700 connected to the normal serial interface SIF. In this way, by selecting the non-busy serial interface SIF to send and receive the recovery packet and the recovery response packet, the serial interface SIF can be used exclusively for the link recovery process.

As described above, even in the embodiment shown in FIG. 15, the same effect as that of the embodiment shown in FIGS. 1 to 14 can be obtained. Further, in the embodiment shown in FIG. 15, a recovery packet and a recovery response packet are transmitted and received using any of a plurality of non-busy serial interface SIFs. Therefore, the transfer efficiency of the recovery packet can be improved and the processing time required for the recovery sequence can be shortened as compared with the case where the recovery packet and the recovery response packet are transmitted and received between the normal packets.

The above detailed description will clarify the features and advantages of the embodiments. It is intended that the claims extend to the features and advantages of the embodiments as described above, without departing from their spirit and scope of rights. Also, anyone with ordinary knowledge in the art should be able to easily come up with any improvements or changes. Therefore, there is no intention to limit the scope of the embodiments having invention to those described above, and it is possible to rely on suitable improvements and equivalents included in the scope disclosed in the embodiments.

10 Information processing device 12 Core part 14 Interface control device 14a, 14b Interface control unit 20 Information processing device 22 Core part 24 Interface control device 24a, 24b Interface control unit 100 Cluster 120 Management device 200 Core part 300 (300a, 300b) Interface control Part 402 Recovery control transmission section 402 Recovery packet transmission section 404 Recovery control transmission section 404 Recovery packet generation section 406 Normal packet transmission section 408 Packet insertion section 410 Communication interface section 421 Recovery sequence control section 413 Recovery processing section 414 Enable section 416 Error detection section 418 Packet judgment unit 420 Recovery packet reception unit 422 Normal packet reception unit 424 Link down notification unit 426 Reduction instruction unit 428 Core notification unit 430 Inter-module control unit 432 Or circuit 500 Information processing device 520 Management device 530 Storage device 600 Core unit 700 ( 700a, 700b) Interface control unit MBUS management bus MM0a to MM3b Interface control unit PHY communication interface unit RIF (RIFa, RIFb) Communication interface SIF (SIFa, SIFb) Serial interface SM0a to SM7b Interface control unit SSU interface control unit SYS1, SYS2, SYS3, SYS4 information processing system

Claims (8)

An interface control device included in one of a pair of information processing devices connected via two communication interfaces and controlling communication with the other information processing device by the communication interface.
The link down detection process of the communication interface is executed, and
When a link down of any of the communication interfaces is detected, or when a link down occurs on the other side of the communication interface is notified from the other information processing device via one of the communication interfaces, the link down occurs. An interface control device that executes a recovery process of the linked link and transmits a recovery control instruction for recovering the link to the other information processing device via one of the communication interfaces. It has a pair of interface control units that are connected to each of the two communication interfaces and control the transmission and reception of packets.
Each of the interface control units
When a link down notification is received from the other information processing device, the link down is notified to the other interface control unit, and when a recovery response corresponding to the recovery control instruction is received from the other information processing device, the other interface control is performed. A reception notification unit that notifies the department of the recovery response,
When a link down of the self-communication interface unit is detected, or when a notification of the link down of the other information processing device is received from the other interface control unit, a degeneration instruction unit for instructing the degeneracy of the link and a degeneration instruction unit.
When a link down of the self-communication interface unit is detected, or when a notification of a link down of the other information processing device is received from the other interface control unit, a recovery process for recovering the downed link is started, and the other A recovery control unit that notifies the interface control unit of a recovery control instruction for recovering a link that has been downloaded to the other information processing device.
The interface control device according to claim 1, further comprising a recovery control transmission unit that transmits the recovery control instruction to the other information processing device when the recovery control instruction is received from the other interface control unit. In each of the plurality of sequence controls included in the recovery process, the recovery control unit instructs the other information processing device to perform the sequence control process via the other interface control unit until the next sequence control disappears. The interface control device according to claim 2, wherein when the completion response of the sequence control process is received from the other information processing device via the other interface control unit, the process of executing the next sequence control is repeated. In the recovery process, the recovery control unit
Degenerate the down link,
Disables the downed link and instructs the other information processing device to disable the downed link via the other interface control unit.
When a link disable completion response is received from the other information processing device via the other interface control unit, the downed link is reset and the downed link to the other information processing device via the other interface control unit. Instructed to reset
When a link reset completion response is received from the other information processing device via the other interface control unit, the link up of the downed link is confirmed, and the link is linked to the other information processing device via the other interface control unit. Instruct to confirm the up,
When a link-up completion response is received from the other information processing device via the other interface control unit, the linked-up link is enabled and linked up to the other information processing device via the other interface control unit. Instructed to activate the linked link
The interface control device according to claim 3, wherein when a link activation completion response is received from the other information processing device via the other interface control unit, the degenerate link is restored. When the recovery control unit does not receive a completion response within a predetermined time from the instruction of sequence control to the other information processing device via the other interface control unit in the recovery process, the recovery control unit uses a down link. The interface control device according to claim 3 or 4. The interface control device according to any one of claims 2 to 5, wherein each of the interface control units transmits a packet including the recovery control instruction between normal packets transmitted to the other information processing device. .. An information processing system having a pair of information processing devices including an interface control device that communicates with a partner information processing device via two communication interfaces.
One of the information processing devices operates as a main information processing device that proactively executes a link-down recovery process of the communication interface, and the other of the information processing devices is based on an instruction from the main information processing device. Operates as a dependent information processing device that performs link-down recovery processing subordinately,
The interface control device of the main information processing device executes a link down detection process of the communication interface and detects any link down of the communication interface, or from the interface control device of the subordinate information processing device. When the link down notification is received, the downed link recovery process is executed, and the link recovery control instruction is notified to the interface control device of the subordinate information processing device via the normal communication interface.
When the interface control device of the subordinate information processing device detects a link down, it notifies the interface control device of the main information processing device of the link down via the normal communication interface, and the main body via the normal communication interface. An information processing system that executes link recovery processing when a recovery control instruction is received from the interface control device of the information processing device. It is a control method of an interface control device included in one of a pair of information processing devices connected via two communication interfaces and controlling communication with the other information processing device by the communication interface.
The link down detection process of the communication interface is executed, and
When a link down of any of the communication interfaces is detected, or when a link down occurs on the other side of the communication interface is notified from the other information processing device via one of the communication interfaces, the link down occurs. A control method of an interface control device that executes a recovery process of the linked link and transmits a recovery control instruction for recovering the link to the other information processing device via one of the communication interfaces.

Images