Method, system, and apparatus for reliable data packet recovery in a link layer of a data center ethernet network

A system for recovering dropped data packets in a Data Center Ethernet (DCE) network, comprising:

• a transmitter for storing data packets for transmission in a replay buffer, the replay buffer accessible at a link layer of the network, and transmitting the data packets across a link layer in the DCE network, wherein each data packet includes a header having a field that indicates if the data packet is formatted for reliable recovery at the link layer;
• and
• a receiver for receiving the data packets transmitted across the link layer, and determining, at the link layer, whether a data packet has been dropped, wherein if a data packet is determined to have been dropped, the receiver sends a non-acknowledgement signal from the receiver to the transmitter in another data packet across the link layer indicating that the data packet has been dropped, and if a data packet is determined not to have been dropped, the receiver sends an acknowledgement signal to the transmitter in the other data packet across the link layer indicating that the data packet has been received the acknowledgement signal sent only after the last packet of the data packets is received at the receiver;
• wherein the field in the header of the data packet is set while the packet is in transit between the transmitter and the receiver, the field enabling reliable recovery at the link layer, wherein setting the field to a first value causes links in the network to apply reliable recovery at the link layer to the data packet, and setting the field to a second value causes the data packet to pass through the links in the network without applying reliable recovery at the link layer to the data packet;
• wherein if no acknowledgement signal is received after a timeout period, it is assumed that the data packet or the acknowledgement signal has been dropped, and
• wherein, in response to receipt of the non-acknowledgement signal at the transmitter or in response to no receipt of an acknowledgement signal at the transmitter after the timeout period, the transmitter retrieves the data packet from the replay buffer, at the link layer, and resends the data packet to the receiver across the link layer;
• wherein the acknowledgement signal is received only after the last packet has been received by the recipient, responsive to the recipient being busy with other tasks.

An apparatus for recovering dropped packets in a Data Center Ethernet (DCE) network, comprising:

• a replay buffer for storing data packets, the replay buffer accessible at a link layer of the network, wherein each data packet includes a header having a field that indicates if the data packet is formatted for reliable recovery at the link layer;
• and
• a transmitter for transmitting the data packets to a receiver across a link layer in the DCE network, wherein
• the field in the header of the data packet is set while the packet is in transit between the transmitter and the receiver, the field enabling reliable recovery at the link layer, wherein setting the field to a first value causes links in the network to apply reliable recovery at the link layer to the data packet, and setting the field to a second value causes the data packet to pass through the links in the network without applying reliable recovery at the link layer to the data packet;
• the link layer, in response to receipt of a non-acknowledgement signal from the receiver transmitted in another data packet across the link layer indicating that a data packet sent to the receiver has been dropped, or in response to no receipt of an acknowledgement signal transmitted in the other data packet across the link layer to the receiver indicating that a data packet has been received after a timeout period, from which it is assumed that the data packet or the acknowledgment signal has been dropped, the transmitter retrieves the packet from the replay buffer, at the link layer, and resends the data packet to the receiver across the link layer;
• wherein the acknowledgement signal is received only after the last packet has been received by the recipient, responsive to the recipient being busy with other tasks.