CISCO’S DATA COMPRESSION SOLUTIONS

The Cisco IOS currently provides these data compression solutions[i]:

• Van Jacobson Header Compression for TCP/IP, which conforms to RFC 1144.
• Per-interface compression (also called link compression), which has one set of dictionaries per hardware link.
• Per-virtual circuit compression (also called payload compression), which has one set of dictionaries per compressed virtual circuit.
• Microsoft Point to Point Compression—MPPC which is RFC 2118 and is a per PPP connection compression algorithm
• Frame Relay Payload Compression—FRF.9 is per DLCI

TCP/IP Header Compression

RFC 1144 defines the Van Jacobson algorithm. In doing so, it also defines the algorithm for TCP/IP header compression.

This algorithm is protocol specific and effective on TCP/IP traffic, it lowers the overhead created by disproportionately large headers, thus reducing the amount of data to be transmitted. This compression technique is most beneficial when transmitting small packets such as Telnet packets.

Cisco's header compression product supports X.25, Frame Relay, and dial-on-demand WAN link protocols. Transaction-oriented applications such as DEC LAT, Telnet, rlogin, XWindows, and acknowledgment packets typically can take best advantage of this type of compression

Due to the processing overhead Header compression is works best at low speed lines generally at 64 kbps

Link Compression

Link compression is also known as per-interface compression, it compresses the entire frame, which means that both the header and the data are compressed. Because the header is compressed it must be decompressed as it reaches the next link, so that the header information can be extracted and the frame can be forwarded to its destination.

Link compression is protocol independent, and every frame is reduced in size.Link compression uses STAC or Predictor to compress traffic, and then encapsulates the compressed data in another link layer such as PPP or LAPB to ensure error correction.

It is also important to remember that if compression is enabled in one side of a link, it must also be enabled at the other side, in both cases using the same algorithm. To find out more about the STAC and Predictor algorithms look at the algorithms section

Payload Compression

Payload compression is also called Per-virtual circuit compression. In payload compression only the data part of the frame is compressed the layer 2 headers must be left intact.

Not all the WAN links are dedicated point to point connections such as: PPP or HDLC. For these circuit Link compression should be used since switches do not examine the data part of the frame.

In cases where switches need to examine the data being transferred payload compression is used. Frame relay for example uses payload compression since it requires that the layer 2 headers are left untouched so that forwarding decisions can be made faster.

The ability to preserve the header makes per-virtual circuit compression a good choice for routers that provide WAN packet services and use a single interface with multiple virtual circuits destined for different locations.[ii]

Payload compression cannot rely on continuous mode compression algorithms, as it would have many virtual circuits each requiring its own dictionary, thus requiring more memory. In these cases packet mode compression algorithms are the preferred option.

Microsoft Point-to-Point compression

This compression method was developed by Microsoft as a means of representing Point-to-Point Protocol packets in a compressed form, it provides a standard method for transporting multi-protocol datagrams over point-to-point links.

The Microsoft Point-to-Point Compression (MPPC) algorithm is designed to optimize processor utilisation and bandwidth utilisation in order to support large number of simultaneous connections.

MPPC is a layer 3 compression method and should not be used with modem compression enabled on a PC. This is because when compressing data that has already being compressed it actually expands.

“MPPC is based on the LZ algorithm, it uses a sliding window history buffer. The MPPC algorithm keeps a continuous history so that after 8192 bytes of data has been transmitted compressed there is always 8192 bytes of history to use for compressing, except when the history is flushed”. [iii]

Frame Relay Payload Compression

The Frame Relay Forum (FRF) defines a set of rules and agreements to ensure vendor interoperability of frame relay networks. FRF9 is a standard for compression for Switched Virtual Circuits (SVC) and Permanent Virtual Circuits (PVC), and is negotiated when DLCI is initialised.

“The compressed payload is transported through the Frame relay network and is decompressed at the termination point.”[iv]