Hola Gilberto, puede ser que hagas referencia a Large Send Offload?
Te incluyo aquí una descripción que he encontrado por Internet:
The TCP large send offload option allows the AIX® TCP layer to build a TCP message up to 64 KB long and send it in one call down the stack through IP and the Ethernet device driver.
The adapter then re-segments the message into multiple TCP frames to transmit on the wire. The TCP packets sent on the wire are either 1500 byte frames for a MTU of 1500 or up to 9000 byte frames for a MTU of 9000 (jumbo frames).
With the TCP large send offload option, you can see that the adapter does a large amount of kernel processing. For example, to send 64 KB of data, it takes 44 calls down the stack, using 1500 byte packets. With the TCP large send option, you can do this with one call down the stack, which reduces host processing and results in lower CPU utilization on the host CPU. The savings varies depending on the average TCP large send size. For example, you can see a reduction of host CPU by 60 to 75% with the PCI-X GigE adapters with a MTU size of 1500. For jumbo frames, the savings are less because the system already sends larger frames. For example, you can see a reduction of host CPU by 40% with jumbo frames.
The first generation Gigabit Ethernet PCI adapters (FC2969 and FC 2975) support the TCP large send offload option. If your primary concern is lower host CPU utilization, use this option. However, for best raw throughput, you should not enable this option because the data rate on the wire is slower with this option enabled.
With the PCI-X GigE adapters, using the TCP large send offload option results in best raw throughput and in lowest host CPU utilization and is therefore enabled by default.
Note: The socket buffer size is increased to 64K at the connection time when the TCP large send offload is enabled and the buffer is originally set smaller than 64K.
Espero que te sirva.... coméntanos, ok?
