COMMUNICATION CONTROL SYSTEM, SWITCH NODE AND COMMUNICATION CONTROL METHOD

• US 20140016648A1
• Filed: 03/21/2012
• Published: 01/16/2014
• Est. Priority Date: 03/23/2011
• Status: Active Grant

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Hidaka, Youichi
• Granted Patent: US 9,407,577 B2
• Time in Patent Office: Days
• US Class Current: 370/400
• CPC Class Codes: H04L 49/15 Interconnection of switching modules

Abstract

In a switch node connected with an external control server, a high functional service protocol processing can be realized by utilizing a multi-route compatible switch and a network interface (NW I/F), which are prescribed by the PCI express. Specifically, in a system which is provided with a switch node and a control server, a plurality of CPUs having a great deal of memories and a plurality of extended NW I/Fs are connected by a multi-route compatible PCI express switch, to configure a switch port of the plurality of extended NW I/Fs. Load distribution transfer processing to the plurality of CPUs from the network interfaces is made possible.

High-speed packet processing is realized through the multiple processing by using the plurality of CPUs. A high-speed switch node is provided in which a large-capacity flow table is configured with the software-based switch node by using a large-capacity memory space of the CPU.

Claims

1. A communication control system comprising:

a switch node configured to carry out processing of each of received packets based on a flow entry defining a rule and an action to uniformly control the packets; and

a control server configured to set the flow entry into a flow table of the switch node,

wherein the switch node comprises;

means for connecting a plurality of processors having large-capacity memories and a plurality of extended network interfaces by a multi-route compatible PCI express switch, to configure switch ports composed of the plurality of extended network interfaces;

means for carrying out load distribution transfer processing from the plurality of extended network interfaces to the plurality of processors and carrying out high-speed packet processing through multiple processing by using the plurality of processors; and

means for configuring a large-capacity flow table in the switch node which is software-based, by using a large-capacity memory space of the plurality of processors.

2. The communication control system according to claim 1, wherein the switch node further comprises:

means for receiving the packets by either of the plurality of extended network interfaces;

means for determining distribution processing to the plurality of processors by using a hash function in one of the plurality of extended network interfaces which receives the packets, and carrying out high-speed processing through the load distribution processing of software packet processing and multiple processing of the packet processing in the plurality of processors; and

means for distributing the packets to the plurality of processors by using a DMA controller of the extended network interface through the PCI express switch, and

wherein each of the plurality of processors comprises;

means for analyzing the received packet with software processing;

means for carrying out table search to the flow table which is configured by using a large-capacity memory provided for its own processor and managing destinations of a great deal of flows to determine a destination output port;

means for controlling the DMA controller of one of the plurality of extended network interfaces which is an output destination after the determination of the destination output port, and transferring the packets through the PCI express switch;

means for transferring the packets to the control server through the extended network interface of the switch node, when the destination output port is not determined as a result of the table search of the flow table, and issues an inquiry of the output destination;

means for storing destination data of a corresponding flow in the flow table when the destination output port is determined as the result of the inquiry; and

means for carrying out synchronization processing of the flow tables among the plurality of processors, such that the flow tables managed by the plurality of processors store the same data to allow the load distribution of the search processing.

3. A switch node comprising:

a plurality of extended network interfaces configured to receive packets;

a plurality of processors having large-capacity memories; and

a multi-route compatible PCI express switch configured to connect the plurality of processors and the plurality of extended network interfaces,

wherein each of the plurality of extended network interfaces comprises;

a LAN interface configured to carry out input and output of the packets;

a packet transferring section configured to carry out at least one transfer processing of transfer processing of the packets to the plurality of processors, transfer processing of the packet between the plurality of processors, and transfer processing to a control server;

a plurality of PF resources configured to carry out transmission and reception of the packets at high speed with the plurality of processors and the DMA transfer; and

a PCI express endpoint connected with the PCI express switch.

4. The switch node according to claim 3, wherein each of the plurality of processors comprises:

a network interface driver connected with the PCI express switch through a PCI express root complex to control one of the plurality of extended network interfaces;

a packet analyzing section configured to carry out analysis processing of the packets supplied from the extended network interface;

a table searching section configured to search a flow table to determine the transfer processing of the packets based on the analysis result of the packets;

a packet buffering section configured to inquire switch processing and processing undetermined packets to the control server and store the packet until a processing method is determined;

a service inquiring section configured to inquire a content of the packet processing to the control server;

an encrypting section configured to carry out encrypted communication with the control server;

an extended network service setting section configured to set the flow table in response to an instruction from the control server and to carry out processing of the extended network service; and

a multiprocessor transfer table synchronizing section configured to communicate the instruction from the control server to other processors and to carry out synchronization processing of the flow tables among the plurality of processors.

5. The switch node according to claim 3, wherein the packet processing section further comprises:

means for checking whether or not a destination MAC address of each of packets is a MAC address used by any of the processors in the switch node, when receiving the packets;

means for transmitting, when the destination MAC address is the MAC address used by any of the processors in the switch node, the packets to the PF resource by turning around without outputting the packet outside; and

means for transmitting, when the destination MAC address is not the MAC addresses used by the processors, the packets to the control server through the LAN interface and inquiring a content of the packet processing to the control server, and transmitting a response packet from the control server to the PF resource.

6. A communication control method in a switch node which carries out processing of a received packet based on a flow entry which defines a rule and an action to uniformly control packets as a flow and which is set in its own flow table from a control server, the communication control method comprising:

connecting a plurality of processors having large-capacity memories and a plurality of extended network interfaces by a multi-route compatible PCI express switch to configure switch ports composed of the plurality of extended network interfaces;

carrying out load distribution transfer processing to the plurality of processors from the plurality of extended network interfaces, and carrying out high-speed packet processing through multiple processing by using the plurality of processors; and

configuring a large-capacity flow table in the switch node which is software-based, by using large-capacity memory spaces of the plurality of processors. 25

7. The communication control method according to claim 6, further comprising:

receiving the packet by any of the plurality of extended network interfaces;

determining distribution processing to the plurality of processors by using a hash function in the extended network interface which has received the packet, and speeding up processing through the multiple processing of the packet processing and load distribution processing of software packet processing in the processors;

distributing the packets to the plurality of processors by using DMA controllers of the extended network interfaces through the PCI express switch;

in each of the plurality of processors;

analyzing the received packets in software processing;

carrying out table search of the flow table which manages destinations of a great deal of flows and which is configured by using the large-capacity memories provided to the processors, to determine a destination output port;

transferring the packets through the PCI express switch by controlling the DMA controller of the extended network interface as an output destination, after the determining of the destination output port;

inquiring the output destination by transferring the packets to the control server through the extended network interface of the switch node, when the destination output port is not determined as a result of the table search of the flow table;

storing destination data of the flow in the flow table when the destination output port is determined as a result of the inquiring; and

carrying out synchronization processing of the flow table among the plurality of processors and storing same data in the flow tables managed by the plurality of processors to achieve the load distribution of search processing.

8. A recording medium which stores a program executed by a switch node in which a plurality of extended network interfaces which receive packets and a plurality of processors having large-capacity memories are connected through a multi-route compatible PCI express switch, and which carries out processing of a received packet based on a flow entry which defines a rule and an action to uniformly control packets as a flow and which is set in its own flow table from a control server,

wherein the program comprises:

extracting a header section of one of the packets when any of the plurality of extended network interfaces receives the packets from a terminal;

carrying out hash processing in a flow unit by using at least one of a MAC address, a VLAN address, and an IP address, of data of the extracted header section;

determining one of the processors as a distribution destination through the hash processing;

transmitting the packets to a packet queue of a PF resource corresponding to the distribution destination processor; and

carrying out DMA transfer of the packets to the distribution destination processor based on a control of the distribution destination processor.

9. The recording medium according to claim 8, wherein the program comprises:

checking whether or not a destination MAC address of each of the packets is same as any of MAC addresses used by the processors in the switch node, when receiving the packets;

transmitting the packets to the PF resource by turning around without outputting the packets outside, when the destination MAC address is the MAC address used by any of the processors in the switch node; and

transmitting the packets to the control server, when the destination MAC address is not any of the MAC addresses used by the processors in the switch node, and inquiring a processing content of the packets to the control server, to transmit a response packet from the control server to the PF resource.

10. The recording medium according to claim 9, wherein the program comprises:

setting a flow table stored in a large-capacity memory provided for its own CPU in response to an entry addition request which is based on a service processing method which is determined by the control server; and

writing the flow table into another CPU based on the setting of the flow table and carrying out synchronization processing of the flow tables among the plurality of processors.