Originally Posted by hks85
how is that not header compression? EVERY packet needs some sort of address because at one point or another equipment needs to know where to send it. Regular IPv4/TCP = 40 Byte address, header compression = 4 Bytes address. If there is something I''m missing, please leave layman's terms out of it.
WAN optimization does its magic almost entirely by reducing bits on the wire. The key is optimizing TCP traffic on the WAN, primarily through adjusting receive-side windows and more aggressively responding to congestion in the sender side. Enterprise applications and protocols, such as Exchange MAPI and SMB/CIFS file transfer, are optimized primarily by object caching and reducing the ping-pong effect. The biggest gains are in data reduction. This approach works very much like compression, but with larger, customized dictionaries. Compression appliances replace a fixed block of bits with a smaller set of bits that represents the larger block. Typically, compression alone gives two to four times reduction in bits on the wire, depending on file type. Text files and bitmap images are highly compressible, where as audio and video are not due to lack of repetitiveness. Data reduction relies on the fact that in many organizations, data passing over the WAN is repetitive--more than one person will read or write to the same file, or maybe someone will read a file, make a few changes, and send it back over the WAN largely unchanged. So all in all no it is not really header compression, it reverts back to my original statement that the payload is compressed. There are specialized caching appliances (BlueCoat, Cisco, and Riverbed are some examples) are placed at both ends of the wire. The solution is not cheap and requires heavy design and administration.