[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: PPP & Compression: Useful or Useless?



"Jens B. Jorgensen" <jjorgens@bdsinc.com> writes:

First off, it was my impression that bsd_comp only affects ppp (TCP?)
headers.  It doesn't do any general data compression (so it wouldn't
try to compress the jpg data in a packet, just the headers).  I had
heard that it is essentially always a win.

> Note that compression protocols, while improving bandwidth, will
> always introduce some amount of latency since encoding/decoding
> takes compute time.

This really depends on how you define latency.  Let's say I want to
send a 1K TCP packet.  Presumably, how fast the first byte gets to the
other side is essentially irrelevant since the application waiting on
the data can't do anything until it gets the entire packet (given the
way TCP works).  In that case, if you can spend a trivial amount of
time compressing the packet (and you can with today's mostly idle
CPUs) to some fraction of its original size, the whole packet will get
transferred faster.  Although the "first byte through" latency is
higher, the packet latency can be *much* lower.

Another more abstract example where the *effective* latency is
dramatically reduced when compression is used is in the case of gzip
compressed ssh connections and X.  In a test letting ssh use gzip to
compress all it's transfers cut a remote launch of xview across a ppp
link to 1/3 of the non-compressed launch time.

-- 
Rob


--
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-user-request@lists.debian.org . 
Trouble?  e-mail to templin@bucknell.edu .


Reply to: