Re: Slightly OT: Compression on a PPP link? (Practical Answer)

From: Howard C. Berkowitz (hcb@gettcomm.com)
Date: Thu Jan 29 2004 - 19:00:57 GMT-3


At 4:47 PM -0500 1/29/04, Kenneth Wygand wrote:
>Hey everyone,
>
>
>
>I have two T1 links in a PPP multilink group, load balancing is working
>PERFECTLY.
>
>
>
>My problem is that the transfer of large image files over this link
>swamps both lines for a long period of time. I'm not sure what the
>format is of these images and if it is highly compressible or not.

Is there a Layer 8 problem in finding out what they are? Without
knowing their format, it's hard to advise.

>However, of the compression options available, are any of them more
>"friendly" to compressing graphical data? Does anyone have any
>real-world experience or practical examples of actual compression
>ratios?

If the images are JPEG, MPEG or a similar image format, you are
unlikely to be able to get any performance increase. The more a
compression algorithm knows about the data being compressed, the more
it can reduce redundancy. Link-level compression can know no more
than byte frequencies in a stream. Two-dimensional compression like
JPEG can, for example, know that it's sending 20 lines of blue sky.
Three-dimensional compression like MPEG know that those 20 lines of
blue sky will stay blue for 30 seconds.

You could try compression, but I doubt it will improve things.

If your concern is with the lines being swamped such that other
services aren't going through, what you may want to do is guarantee
some bandwidth to other services using CBQ, WFQ or CBWFQ, or
rate-limit the graphics stream.

>
>
>The compression algorithms I know of are Compressor, Predictor and MPPC.
>
>
>
>Also, how can I verify the compression ratio I am getting through a show
>command?
>
>
>
>Last question - does anyone know how much these algorithms tax the
>routers? I am using 2610s and 3640s with 8MB Flash and 32MB DRAM
>running 12.0 IOS.
>
>

The numbers I have are too old, but compression historically is
intensive. You could look at compression AIMs, but, as I've
suggested, I doubt you are going to get much performance increase
with link-level compression.



This archive was generated by hypermail 2.1.4 : Mon Feb 02 2004 - 09:07:52 GMT-3