KR101344360B1 - Adaptive packet coalescing methed - Google Patents

Abstract

The present invention relates to an adaptive packet merging method, when a certain amount of data is received for a time required for switching from an inactive mode to an active mode and a time required for switching from an active mode to an inactive mode, the present invention does not switch from the active mode to the inactive mode. Energy efficiency is increased because the active mode can be maintained.
The present invention for this purpose is disclosed. In the adaptive packet merging method, a packet received in an inactive state is stored in a storage module, and the average arrival rate of the packet is calculated using the packet stored in the storage module, and the average arrival rate of the packet and the maximum transmission of the packet are possible. Calculating a maximum storage rate of the storage module using a congestion window indicating a packet amount, and switching a mode of Ethernet by comparing the maximum storage rate with the number of packets stored in the storage module.

Description

Adaptive Packet Merging Method {ADAPTIVE PACKET COALESCING METHED}

The present invention relates to an adaptive packet merging method, and more particularly, to an adaptive packet merging method for switching an Ethernet mode according to traffic characteristics of a transport layer.

Ethernet is the most commonly used networking technology in local area networks (LANs) and is a networking technology that operates on a network connecting multiple PCs, such as laptops, desktops, and server computers. Ethernet is widely used in homes and commercial districts and is the most widely deployed subscriber network networking technology in the world.

This network environment continuously transmits idle signals to keep the transmitter and reception time aligned even when there is no data to transmit, so most components of the network card are always active and consume unnecessary power. It is observed to consume about 6 TWh (Tera Watt hours) power.

For this reason, the IEEE 802.3az Working Group has standardized on energy efficient solutions to address the corresponding costs of using IT equipment such as PCs, display devices, printers, servers, and network equipment. As a result, the IEEE 832.3az Working Group adopted the standard based on the Low Power Idle (LPI) mode and specified the detailed operation of the physical layer to support it.

LPI operates in active mode only when there is data to be transmitted, allowing low-use access networks to save significant energy. However, in order to show the optimum efficiency, the transport packet should be transmitted in the form of burst back-to-back. However, speed and reliability should be applied differently according to the importance of the packet. For packets that require high speed, packet coalescing to optimize energy efficiency will adversely affect network performance.

An object of the present invention for solving the above problems is, when a certain amount of data is received for a period required to switch from the inactive mode to the activation mode and a period required to switch from the activation mode to the deactivation mode, from the activation mode to the deactivation mode An adaptive packet merging method for maintaining an active mode without switching is provided.

Another object of the present invention for solving the above problems is, when the inactive mode holding time in the inactive mode ends, the number of packets stored in the storage module and the QoS (Quality of The present invention provides an adaptive packet merging method for adaptively maintaining an inactive state according to a service level.

In order to achieve the above object of the present invention, an adaptive packet merging method according to an embodiment of the present invention stores a packet received in an inactive state in a storage module, and uses the packet stored in the storage module to store the packet. Calculating an average arrival rate, calculating a maximum storage rate of the storage module using a congestion window indicating an average arrival rate of the packet and a maximum amount of packets that can be transmitted, and the number of packets stored in the storage module and the Comparing the maximum storage rate to switch the mode of Ethernet.

The switching of the Ethernet mode may include: switching from the deactivated state to an activated state when the number of the stored packets is equal to or greater than a maximum storage rate of the storage module as a result of the comparison; And maintaining the deactivation state when the number of stored packets is less than the maximum storage rate of the storage module as a result of the comparison.

In the calculating of the maximum storage rate of the storage module, when the packet is the first packet stored in the storage module, the deactivation mode holding time for maintaining the deactivation state may be determined with reference to the header of the packet.

Comparing the number of packets stored in the storage module with a threshold storage rate corresponding to half of the maximum storage rate of the storage module and switching the Ethernet mode according to the comparison result when the deactivation mode maintenance time is determined. It may further include.

The switching of the Ethernet mode may include: switching from an inactive state to an activated state when the number of packets stored in the storage module is greater than or equal to the threshold storage rate; And maintaining the deactivation state when the number of packets stored in the storage module is less than the threshold storage rate.

The maintaining of the deactivation state may include determining the importance of the packet using a Quality of Service (QoS) level included in the header of the packet stored in the storage module, and then, if the importance is greater than or equal to a switching threshold, the deactivation state may be activated. In the case of switching, if the importance is less than the switching threshold, the deactivation mode holding time may be switched to an initial state and the deactivation state may be maintained.

In the adaptive packet merging method according to another embodiment of the present invention, transmitting a packet stored in a storage module in an activation mode, and when the transmission of the packet stored in the storage module is terminated, Calculating a packet reception amount receivable in mode switching by using a storage record; and switching the mode of Ethernet by comparing the calculated packet reception amount with a maximum storage rate of packets that can be stored in the storage module.

The switching of the Ethernet mode may include: maintaining the deactivated state when the calculated packet reception amount is equal to or greater than a maximum storage rate of the storage module; And switching from the deactivation mode to the activation mode when the calculated packet reception amount is less than the maximum storage rate of the storage module.

In addition, the adaptive packet merging method according to another embodiment of the present invention, receiving the packet in the activation mode; Calculating an average arrival rate of the received packet by using a storage record of the storage module in which the received packet is stored; And updating the calculated average arrival rate.

According to the adaptive packet merging method according to an embodiment of the present invention, if a certain amount of data is received for a period required for switching from inactive mode to active mode and for a period required for switching from active mode to inactive mode, the active mode is deactivated. Energy efficiency is increased because the active mode can be maintained without switching to the mode.

In addition, when the inactive mode maintenance time ends in the inactive mode, the present invention adaptively deactivates according to the number of packets stored in the storage module and the quality of service (QoS) level of the packet without switching from the inactive mode to the active mode. Energy efficiency increases by maintaining

1 is a sequence chart illustrating an operation procedure of a low power idle mode (LPI).
2 is a graph for explaining the change in the congestion window size of slow start and congestion avoidance.
3 is a diagram illustrating an interface between an LPI client and a TCP layer.
4 is a flowchart illustrating an adaptive packet merging method according to an embodiment of the present invention.
5 is a flowchart illustrating an adaptive packet merging method according to another embodiment of the present invention.
6 is a flowchart illustrating an adaptive packet merging method according to another embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings. It is to be noted that the same components of the drawings are denoted by the same reference numerals and symbols as possible even if they are shown in different drawings. In the following description of the present invention, if it is determined that a detailed description of a related known function or configuration may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted.

Also, when an element is referred to as "comprising ", it means that it can include other elements as well, without departing from the other elements unless specifically stated otherwise.

Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the accompanying drawings.

An ethernet mode switching method for performing energy efficient ethernet (EEE) will be substantially shown or described in connection with at least one of the figures and will be more fully developed in the claims.

Ethernet is being applied in a variety of environments (eg, twisted pairs, backplanes, etc.) and is becoming an increasingly compelling technology. IEEE 802.3az Energy Efficient Ethernet evaluates various methods to reduce the energy used during the time of low link utilization. In this process, a protocol may be defined that facilitates the transition from / to low power consumption modes according to changes in network requirements.

Low Power Idle (LPI) relies on switching from active mode to inactive mode when not transmitting. Energy is thereby saved when the link is off. Refresh signals may be sent periodically to switch from inactive mode to active mode.

In one embodiment, a synchronization signal on the interfaces (i.e. medium dependent interface (MDI) and PHY / MAC interface) to quickly switch from inactive mode to active mode and maintain a frequency lock. Can be used.

1 is a sequence chart illustrating an operation procedure of a low power idle mode (LPI).

Referring to FIG. 1, the LPI mode may include an activation mode, an inactivation mode, and a refresh mode. In this case, when there is a packet to be transmitted from the transmitting end to the receiving end, the transmitting end may transmit the packet to the receiving end in the activation mode.

In addition, in the activating mode, when the transmitting end completes packet transmission and there is no data to transmit to the receiving end, the transmitting end is switched from the activating mode to the deactivating mode. In order to maintain synchronization between the transmitting end and the receiving end, the transmitting end may transmit the refresh signal to the receiving end in the refresh mode.

In FIG. 1, the holding time of the activation mode, the deactivation mode, and the refresh mode is arbitrarily defined, respectively, but the retention time of the activation mode, the deactivation mode, and the refresh mode depends on the state of the network and the reception rate of data received in the buffer in the deactivation mode. It should be noted that this may be changed, so that FIG. 1 may be changed accordingly.

When the transmitting end receives the packet from the upper layer in the inactive mode, the LPI mode is switched from the inactive mode to the active mode at the time T1 after the Tw time. The transmitting end may transmit the packet to the receiving end from the time T1 at which the activation mode begins to the time T2 at which the activation mode ends.

In the active mode, when the transmitting end completes the transmission of the packet and there is no data to transmit to the receiving end, the LPI mode is changed from the activation mode to the inactive mode after the Ts time from the time point T2 to maintain the inactive mode.

In this case, the transmitting end may transmit the packet to the receiving end in an interval from the time T1 at which the activation mode is started to the time T2 at which the deactivation mode is switched, and thus may be called an active period.

In addition, in the inactive mode, the transmitting end does not transmit data to the receiving end, but the transmitting end must maintain synchronization with the receiving end. To this end, the LPI mode maintains the refresh mode by switching from the inactive mode to the refresh mode at a time T3 after the Ts time, for example, T4 after the Tq time.

In the refresh mode, the transmitting end transmits a refresh signal to the receiving end to maintain synchronization with the receiving end. When the transmitting end completes the transmission of the refresh signal in the refresh mode, the LPI mode is switched from the refresh mode to the inactive mode.

When the packet is received from the upper layer in the deactivation mode, the LPI mode is switched from the deactivation mode to the activation mode at the time T5 after the Tw time. In the inactive mode, the transmitting end may transmit the packet to the receiving end.

In the activating mode, when the transmitting end completes the transmission of the packet and there is no packet to transmit to the receiving end, the LPI mode is changed from the activating mode to the deactivating mode after the Ts time from the time T6 to maintain the inactive mode.

In this activation mode, the transmittable packets may be merged and transmitted in burst form. This packet merging function means a function of merging or buffering packets flowing into one First In First Out (FIFO) queue in an Ethernet interface in order to transmit as much packets as possible in burst form.

This packet merging function is still used to reduce the overhead of CPU processing on the receiving side of the Fast Ethernet interface. The packet merging function is a function of merging packets stored in a transmission buffer by receiving a predetermined number of packets from a higher layer and transmitting them in burst form or merging packets received from a higher layer for a predetermined time in burst form.

However, packet merging improves the energy efficiency of EEE's LPI mode. Especially in the case of low traffic load, the energy efficiency is further improved by previous studies. However, the packet merging mechanism has also been shown to cause an increase in packet delivery delay.

This is because, in the packet merging mechanism, the performance of the energy efficiency and the propagation delay largely depend on the packet reception timer setting value (Tcoalesce) and the maximum number of packets to be merged (Qmax).

For example, if the maximum number of packets to be merged and the packet reception timer settings are large, energy efficiency is improved, but packets received after receiving packets until the set packet reception timer expires or until the maximum number of packets to be merged are reached. Because of the merging, packet delays increase.

For example, if the maximum number of packets to be merged and the packet reception timer setting value are smaller, the delay of packets is reduced, but the energy efficiency is relatively low due to frequent packet merging.

The packet merging function of the LPI results in an increase in packet delay due to buffering, and this increase in packet delay results in an increase in a round trip time (RTT) value of a packet. Conventionally, such a delay represents an increase in the order of hundreds or hundreds of microseconds, so that the RTT value of a typical TCP flow does not have a significant effect on performance in light of hundreds of milliseconds (msec).

However, since most TCP connections can be made through multiple networks, the accumulation of delays on multiple interfaces can affect TCP application services. In addition, since various services such as a real-time multimedia application service are recently provided through the Internet, a mechanism for providing differentiated QoS for each application service is required. Therefore, a function to support differentiated quality of service (QoS) for each application service is necessary in the LPI packet merging function.

Therefore, there is a need for an improved packet merging mechanism that maximizes energy efficiency by adaptively adjusting the number of packets and timer settings according to the traffic characteristics of the transport layer, and guarantees differential delays according to application service requirements.

On the other hand, the characteristics of congestion control of TCP is that the data transmission period is determined by the RTT value, but the actual data transmission is not transmitted evenly during the RTT time, but data is transmitted in burst form in a very short time at the beginning of the period. . The size of the burst data is also determined by the congestion window (cwnd) size of TCP.

Therefore, in the present invention, the current packet arrival rate is measured each time a packet is received, and the maximum storage rate of the transmission buffer is calculated by using the measured packet arrival rate and a congestion window indicating the maximum transmittable packet amount of the packet received from the upper layer. After that, the number of packets stored in the transmission buffer and the maximum storage rate may be compared, and the packets stored in the transmission buffer may be merged and transmitted according to the comparison result.

2 is a graph for explaining the change in the congestion window size of slow start and congestion avoidance.

Referring to FIG. 2, when receiving a plurality of packets having the same ACK sequence number, the transmitter may retransmit the lost packet assuming that the packet is lost during transmission. If the retransmission packet is lost or damaged during transmission, the transmitting end does not receive an ACK packet indicating that the retransmitted packet has been received from the receiving end, and thus cannot transmit for a predetermined time until the retransmission timer expires.

As a result, when the transmission time increases and the retransmission timer expires, the transmitting end may perform a slow start. In the slow start state, the window may be exponentially increased by the ACK packet received from the receiving end.

Therefore, if we can measure the time and congestion window starting from the first packet received during Ts until the arrival of n packets, we can predict how many packets arrive during Tmax and use this to determine the average packet arrival rate. Can be calculated

However, the congestion window value is often larger than the MAC buffer size. Therefore, if the ssthresh value is provided from TCP and can be obtained by applying a ratio to the value, the mode conversion can be efficiently performed. However, when a certain amount of the TCP buffer is maintained, the active mode is always maintained. Therefore, the result value must be limited to the MAC buffer size so that an appropriate mode switch can be performed.

In other words, TCP congestion avoidance begins when the congestion window increases exponentially at the slow start to equal the ssthresh value. In the TCP congestion avoidance state, the value of the congestion window is increased by 1 for one ACK received from the receiver.

The value of ssthresh is set to the size / 2 of the congestion window. Here, the value of ssthresh is determined as half of the current window, which means that the maximum size of the approximate window where packet loss does not occur is 1.

The congestion window is also an element of the TCP congestion control algorithm that affects the size of the segment to be transmitted in SLOW START, CONGESTION AVOIDANCE, and FAST RECOVERY.

3 is a diagram illustrating an interface between an LPI client and a TCP layer.

In the embodiment of FIG. 3, detailed descriptions of other layers except for the TCP layer among the OSI 7 layers will be apparent to those skilled in the art, and thus a detailed description thereof will be omitted, and information transmission between the TCP layer and the LPI client focused on the present invention will be omitted. Let's explain.

Represents a framework for the interface between the LPI client and the TCP congestion control and flow control mechanisms. Information transfer between LPI Client and TCP is performed by applying cross-layering technique.

That is, the LPI client stores the received packet in the inactive state in the storage module, calculates the average arrival rate of the packet using the packet stored in the storage module, and then uses the average packet arrival rate and cross-layering technique of the TCP layer. The maximum storage rate of the storage module may be calculated using a congestion window indicating the maximum amount of packets that can be received from the packet, and the Ethernet mode may be switched by comparing the calculated maximum storage rate with the number of packets stored in the storage module. .

4 is a flowchart illustrating an adaptive packet merging method according to an embodiment of the present invention.

Referring to FIG. 4, when a transmitting end receives a packet to be transmitted from a higher layer to a receiving end in an inactive state (S401) (S402), the transmitting end stores the received packet in a storage module and uses the packet stored in the storage module to average the arrival rate of the packet. To calculate (S403).

The transmitting end determines whether the packet received from the upper layer is the first packet to be stored in the storage module (S404), and when the determination result is that the packet received from the upper layer is the first packet to be stored in the storage module, the calculated average arrival rate and the maximum number of packets The maximum storage rate of the storage module may be calculated using a congestion window indicating the amount of packets that can be transmitted (S405), and the deactivation mode holding time for maintaining the deactivation state may be calculated with reference to the header of the packet (S406). Here, the congestion window may be received from the TCP layer through cross-layering as described with reference to FIG. 3.

If it is determined that the packet received from the upper layer is not the first packet to be stored in the storage module (S404), the transmitter may compare the number of packets stored in the storage module with the maximum storage rate that can be stored in the storage module (S407). If it is determined that the number of packets stored in the storage module is greater than or equal to the maximum storage rate that can be stored in the storage module, the transmitter switches from the inactive mode to the activation mode (S408), and the number of packets stored in the storage module is less than the maximum storage rate that can be stored in the storage module. In operation S409, it may be determined whether the inactive mode maintenance time for maintaining the inactive mode has ended.

If it is determined that the deactivation mode holding time for maintaining the deactivation state has not ended, the transmitting end may maintain the deactivation mode (S401). On the other hand, if it is determined that the deactivation mode maintenance time for maintaining the deactivation state has ended, the importance of the packet is determined using the quality of service (QoS) level included in the header of the packet stored in the storage module, and the determined importance is a switching threshold. You can determine if it is greater than the value.

Here, the QoS level is transmitted to the TCP layer when the initial connection is established. In the TCP layer, the QoS level is marked in the header part of the packet and delivered to the MAC layer. In the MAC layer, the packet is referred to by referring to the QoS level marked in the header part of the packet. You can judge the importance of

If it is determined that the determined importance is equal to or greater than the switching threshold, the transmitter switches from the deactivation mode to the activation mode (S411). If it is determined that the determined importance is less than the switching threshold, the transmitter initializes the mode holding time (S412). Can be maintained (S401).

5 is a flowchart illustrating an adaptive packet merging method according to another embodiment of the present invention.

Referring to FIG. 5, the transmitting end may transmit a packet stored in a storage module in an inactive mode in an active mode (S501). Thereafter, the transmitting end may determine whether all packets stored in the storage module have been transmitted (502). If it is determined that all packets stored in the storage module have been transmitted (S502), the transmitting end may calculate an amount of packets that can be received in mode switching (S503). Here, all the conversions mean the time taken to switch from the inactive mode to the activated mode and the time taken to switch from the activated mode to the inactive mode.

The transmitter compares the amount of packets that can be received in the mode switch with the maximum storage rate of the packets that can be stored in the storage module (S504), and as a result of the comparison, the amount of packets that can be received in the mode switch is greater than or equal to the maximum storage rate of the packets that can be stored in the storage module. If it is determined that the switch to the inactive mode from the active mode (S505). On the other hand, if it is determined that the amount of packets that can be received in the mode switching is less than the maximum storage rate of the packets that can be stored in the storage module, the transmitting end maintains the activation mode (S506).

6 is a flowchart illustrating an adaptive packet merging method according to another embodiment of the present invention.

Referring to FIG. 6, when a transmitting end receives a packet in an activation mode (S601), the transmitting end performs buffering to store the received packet in a storage module (S602). Thereafter, the transmitting end updates the average arrival rate of the received packet by using the reception record of the packet stored in the storage module (S603).

The present invention has been described with reference to preferred embodiments. Although specific terms have been employed herein, they are used for purposes of illustration only and are not intended to limit the scope of the invention as defined in the claims or the claims.

Therefore, it will be understood by those skilled in the art that the present invention may be implemented in a modified form without departing from the essential features of the present invention. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The scope of the present invention is shown in the claims rather than the stated description, and all differences within the scope will be construed as being included in the present invention.

Claims (9)

delete Storing the received packet in a deactivated state in a storage module and calculating an average arrival rate of the packet using the packet stored in the storage module;
Calculating a maximum storage rate of the storage module by using a congestion window indicating an average arrival rate of the packet and a maximum amount of transmittable packets of the packet; And
Switching the mode of Ethernet by comparing the number of packets stored in the storage module with the maximum storage rate,
The switching of the Ethernet mode may include: switching from the inactive state to an activated state when the number of the stored packets is equal to or greater than a maximum storage rate of the storage module as a result of the comparison; And
And if the number of stored packets is less than the maximum storage rate of the storage module as a result of the comparison, maintaining the deactivated state. 3. The method of claim 2,
The calculating of the maximum storage rate of the storage module may include determining an inactive mode holding time for maintaining the inactive state with reference to a header of the packet when the packet is the first packet stored in the storage module. Type packet merging method. The method of claim 3, wherein
Comparing the number of packets stored in the storage module with a threshold storage rate corresponding to half of the maximum storage rate of the storage module and switching the Ethernet mode according to the comparison result when the deactivation mode maintenance time is determined. Adaptive packet merging method further comprising. 5. The method of claim 4,
The switching of the Ethernet mode may include: switching from an inactive state to an activated state when the number of packets stored in the storage module is greater than or equal to the threshold storage rate; And
And if the number of packets stored in the storage module is less than the threshold storage rate, maintaining the deactivation state. The method of claim 5, wherein
The maintaining of the deactivation state may include determining the importance of the packet using a Quality of Service (QoS) level included in the header of the packet stored in the storage module, and then, if the importance is greater than or equal to a switching threshold, the deactivation state may be activated. And switching the deactivation mode holding time to an initial state and maintaining the deactivation state when the importance is less than a switching threshold. delete Transmitting a packet stored in the storage module in an activation mode;
When the transmission of the packet stored in the storage module is finished, calculating a packet reception amount receivable in mode switching by using a storage record of the packet stored in the storage module; And
Comprising the step of switching the mode of Ethernet by comparing the calculated packet reception amount and the maximum storage rate of the packets that can be stored in the storage module,
The switching of the Ethernet mode may include: switching from the activation mode to the deactivation mode when the calculated packet reception amount is equal to or greater than a maximum storage rate of the storage module; And
And maintaining the activation mode when the calculated packet reception amount is less than the maximum storage rate of the storage module. The method of claim 8,
Receiving a packet in the activation mode;
Calculating an average arrival rate of the received packet by using a storage record of the storage module in which the received packet is stored; And
Adaptive packet merging method further comprising the step of updating the calculated average arrival rate.

Images