What is the difference between a toe (TCP/IP Offload Engine) Nic and a general Nic?

TCP/IP
The Protocol has long been the standard language of the network. With the advent of Internet SCSI and remote diret memory access, TCP/IP has become a storage protocol.




We know that using the TCP/IP protocol to process network traffic requires a large amount of server resources. To reduce the pressure on the server, a TCP offload
Engine
: Toe) technology came into being. The TCP burden reduction engine is generally composed of two components: hardware and software. It extends the functions of the traditional tip/IP protocol stack and forwards all the processing of network data traffic to the network adapter.
The server only processes TCP/IP control information. This technology reduces the burden on servers, which is recognized by most vendors.

The common Nic performs a series of TCP/IP-related operations in software mode. Therefore, the server load is increased in three aspects: Data replication, protocol processing, and interrupt processing.

each application on the Network Program causes a large number of network I/O interruptions when sending and receiving a large number of packets, responding to these I/O interrupt signals has become a heavy burden on the server. For example, when a typical
64 Kbps application sends data to the network, in order to assemble the data into Ethernet data packets and respond to the network receiving confirmation signals, to trigger more than 60 interrupted items between the server and the network card
, such a high interruption rate and protocol analysis workload is already considerable. Although some network operating systems have the interrupt bundle function, which can effectively reduce the generation of Interrupt signals, it cannot reduce the total amount of Event Response Processing between servers and NICs
.
the working principle of the TCP burden reduction engine Nic is different from that of the common Nic. An interruption is triggered when a common Nic processes each data packet, while a TCP
minus engine Nic triggers an interruption only after each application completes a complete data processing process, this significantly reduces the server's response load to interruptions. Take a 64 Kbps application as an example. After the application
sends all data to the network in sequence, it sends a data channel to the server to reduce the negative event interruption, the processing of data packets is done by the TCP minus engine Nic, rather than by the server, so as to eliminate the excessive interference of frequent interruption events on the server. When network applications send and receive data, they often need to copy multiple copies of the same data. In this case, the benefits of the TCP minus engine NIC are the most obvious.
by using hardware and software that supports verification, a common Nic can reduce the amount of data copies sent, however, it cannot reduce the amount of data copies received. Copying a large amount of received data
usually occupies a large number of machine work cycles. The common network adapter first copies the received data in the server's buffer zone and assigns it to one of the TCP connections after processing by the system. Then, the system then associates the data with the application that uses it, and copies the data from the system buffer to the application buffer. When receiving data, the TCP burden reduction engine Nic performs protocol processing in the NIC. Therefore, it does not necessarily copy data to the Server Buffer
, but directly to the application buffer, this "zero copy" method avoids unnecessary data reciprocating copies between the network card and the server.
the TCP minus engine Nic can significantly reduce the server overload burden caused by massive data movement. Test results show that for file servers and content-based server application environments, if you use a TCP minus engine Nic to replace a common Nic, it is equivalent to adding a CPU to the server