HP 200 Series Services And Applications page 186

Data Compression for WAN Links
What is Compression?
and receiving devices must have the same dictionary in order to replace the
key with the original text at the receiving end.
Running Dictionary
One method of data compression uses a "running dictionary", meaning that
the mappings of strings to keys is maintained and reused across multiple
packets being transmitted and received. This method has the benefit of
achieving high compression ratios, but it has some restrictions. First,
running dictionaries require large amounts of memory to be maintained.
Generally, the larger the dictionary (i.e., the more memory used), the better
the compression results, because it is more likely that a string will already
have a key defined in the dictionary. Another more serious drawback is that
if the dictionaries at the sending and receiving devices gets out of synchroni-
zation, all packets will be dropped until the dictionaries re-synchronize.
Therefore, if the link between the sending and receiving devices is not
reliable, then it is highly probable that the dictionaries will often be out of
synchronization, which may result in poor performance or complete failure
of the connection. Running dictionaries have difficulty with WAN technolo-
gies that use datagram oriented-protocols (e.g., X.25 or frame relay), in
which packets may arrive at the receiving device out of order or corrupted.
In other words, running dictionaries require that packets be reliably
delivered across the link with no data loss or corruption. Additionally, if the
WAN link uses multiple virtual circuits, then a separate dictionary is used for
each virtual circuit, thus requiring more memory.
Packet-by-Packet Dictionary
To overcome the memory and link quality limitations of running dictionaries,
another method of data compression was developed in which a very small
dictionary is used, and it is reset for each packet. Therefore, very little
memory is required to store the dictionary, and there is no way for the
sending and receiving devices to get out of synchronization, even when
using unreliable WAN link protocols. Packet-by-packet dictionary methods
do not achieve compression ratios as high as running dictionary methods,
because there is no chance of reusing a key from matching strings in
previous packets. However, the higher ratios of running dictionary methods
are diminished by the fact that reliable link protocols use more link band-
width for acknowledgment packets. Therefore, the overall throughput may
not be as high as expected when using running dictionaries. Additionally,
the synchronization problems associated with running dictionaries makes
the overall performance of packet-by-packet methods more effective for use
on wide area links.
2-134