KR100283104B1 - Selective Host Interrupt Method of Communication Equipment - Google Patents

Abstract

The present invention relates to a selective host interrupt method of communication equipment, and more particularly to a selective host interrupt method of communication equipment for minimizing the interrupt burden when transmitting and receiving large packets.

The present invention uses a packet header equipped with a flag to inform the receiver network control program whether an interrupt has occurred, and then informs the network control program of the last packet information after splitting a packet in a user process. It is characterized by transmitting only the flag of the packet header and transmitting. And, when sending and receiving a packet larger than the maximum transmission unit of the network card, the interrupt can be generated as the sender intends. Since it can be eliminated, it is to provide a selective host interrupt method of the communication equipment that can greatly improve the overall communication performance.

Description

Selective Host Interrupt Method of Communication Equipment

The present invention relates to a selective host interrupt method of communication equipment, and more particularly to a selective host interrupt method of communication equipment for minimizing the interrupt burden when transmitting and receiving large packets. The present invention provides an environment in which a card control program can be directly implemented by a developer among network cards developed for a high-speed closed network disconnected from the outside, such as a clustering system, that is, only on a card that supports programmable firmware. Applicable

Conventional communication software for workstation clustering systems prefer polling. This is due to the performance burden caused by interrupts.

However, in a system where many processes work at the same time, there is a burden of scheduling several processes, which greatly reduces the performance of polling-based communication processes. In order to overcome this disadvantage, the interrupt method is inevitable.

1 is a general packet structure diagram of a conventional closed network card for high speed communication, and FIG. 3 is a flowchart of a large packet in a conventional closed network.

In the packet header shown in FIG. 1, there is a pad 101 used to align routing information, a packet type, a port number of a user to which a packet is to be transmitted, a packet length, and a header size.

In Fig. 3, reference numeral 301 denotes a packet owned by the user process on the sender side, 302 denotes logical transmission at the user level, 303 denotes packets divided according to the maximum transmission unit of the network card at the user level, and 304 DMA is the packet header of FIG. 1, 305 is the physical header at the card level, 307 is the host interrupt, 308 is the arrived packets before reassembly, and 309 is the recipient's user Represents a packet received by a process.

However, in the conventional interrupt method, an interrupt is generated every time a packet arrives, as in Ethernet. The disadvantage of this case arises when a user transmits a much larger packet than the maximum transmission unit (MTU) of the transmission equipment. In other words, at the user level, only one packet is sent or received, while at least one interrupt is generated. For systems aimed at high speed communications this is a heavy burden.

Accordingly, an object of the present invention is to provide a selective host interrupt method of a communication device capable of delivering a packet to a user with a single interrupt even when transmitting a packet much larger than a maximum transmission unit (MTU).

The present invention for achieving the above object is characterized by using a packet header equipped with a flag indicating to the receiver network control program whether an interrupt has occurred.

In addition, the present invention is characterized by notifying the network control program the last packet information after the packet is divided in the user process, and transmits the flag set only in the header of the last packet based on this information.

1 is a general packet structure diagram of a closed network card for a conventional high speed communication.

2 is a packet structure diagram of a closed network card according to the present invention;

3 is a flowchart of a large packet in a conventional closed network.

4 is a large packet flow diagram in a closed network in accordance with the present invention.

<Description of Signs of Major Parts of Drawings>

101: Pad area used to align header sizes in a packet structure of a conventional closed network card

201: Flag area added to determine whether an interrupt occurs in the present invention

301: Packet owned by sender's user process

302: logical transfer at the user level

303: Packets split according to the maximum transmission unit of the network card at the user level

304: DMA of a packet to be transmitted 305: packet header of FIG.

306: Physical transfer at card level 307: Host interrupt

308: Arrived packets before being reassembled

309: Packet received by the user process on the receiver side

401: packet header of FIG.

402: information about a packet divided by a user for transmission

Hereinafter, a method of generating an interrupt only once when receiving a packet according to the present invention will be described with reference to the accompanying drawings.

2 is a packet structure diagram of a closed network card according to the present invention, showing a header of a packet actually used in the present invention.

The header of the packet includes information for routing, the type of the packet, the port number of the user to which the packet is to be delivered, the length of the packet, and a flag 201 indicating whether the packet is to be interrupted or not. have.

4 is a flowchart illustrating a large packet size in a closed network according to the present invention, and illustrates a process of transferring a packet from a sender process to a receiver process in the present invention.

The sender process segments the packet to be sent according to the MTU of the network card and stores the information 402 where the control program of the network card can access. The control program of the network card reads the packet information and generates a header 401 for each packet. At this time, a flag is inserted into the header of the last packet and transmitted. The receiver's network card control program looks at the packet's header when it arrives, determines whether or not it will generate an interrupt, and directs the packet to the receiving user process for direct memory access (DMA).

The present invention described above is capable of various substitutions, modifications, and changes without departing from the spirit of the present invention for those skilled in the art to which the present invention pertains. It is not limited.

As described above, the present invention can deliver the packet to the receiver with only one interrupt burden regardless of the size of the packet transmitted by the user process.

In addition, when a large packet such as a multimedia data is frequently transmitted and received within the clustering system, when the conventional method is used, there is a possibility that a lot of interrupts are generated and the performance is greatly reduced. However, the method of the present invention may be applied. In this case, it is expected to improve performance.

Claims (2)

An optional host interrupt method of a communication device, characterized by using a packet header equipped with a flag indicating whether an interrupt has occurred. After the packet is split in the user process, the last packet information is informed to the network control program, and the host interrupt method of the communication device is characterized by transmitting a flag set only in the header of the last packet based on the information.