Path: utzoo!mnetor!uunet!lll-winken!lll-tis!ames!telebit!rls
From: rls@telebit.UUCP ( Sr. Systems Engineer)
Newsgroups: comp.dcom.modems
Subject: Re: hayes 9600 vs. trailblazer
Message-ID: <280@telebit.UUCP>
Date: 13 Apr 88 16:48:10 GMT
References: <8804110136.AA16978@ucbvax.Berkeley.EDU> <15612@onfcanim.UUCP> <494@edsews.EDS.COM>
Organization: Telebit Corporation, Cupertine, CA
Lines: 71
Summary: Data Compression in the TrailBlazer vs. news compression

In article <494@edsews.EDS.COM>, roberts@edsews.EDS.COM (Ted Roberts) writes:
> I realize that news is normally compressed by the system before transmission
> partly due to the modems that don't do data compression in the hardware.
> However, I'm curious if in a this is actually the best method for the
> Trailblazer.  Is the modified Lempel-Ziv data compression algorithm used
> by the news better than the type used by the Trailblazer?  What algorithm
> is used by the TB?

Just to set the record straight, here's a brief summary about compression.

1.  The Telebit TrailBlazer uses Lempel-Ziv data compression, when the S110
    register is set appropriately. It is an option that can (and should) be
    disabled in certain cases.

2.  With compression algorithms in general, there are a couple of rules that
    should be considered to determine when compression is desired, and when it
    should be turned off.

2a. If you have data that is regular, non-random structure, then there is a
    chance that it will be compressible to some degree. Most compression 
    algorithms search for some form of pattern in the data, and use various
    methods to reduce the amount of data sent. 

2b. When a compression algorithm tries to operate on a random (such as binary,
    or executable) file, it will lose in two ways: first, there will be a TIME
    penalty involved with the actual time required to execute the compression
    algorithm. Second, there will be a SIZE penalty, because when uncompressible
    data is run through a compression algorithm, it will GROW in size from 10 to
    30 percent (variable depending on data and the algorithm used).

2c. With compressible data, different data has different levels of 
    compressiblilty. Obviously, data with very regular patterns, or lots of
    white space is much more compressible than random text, etc. The point is,
    the compression ratio can vary quite widely. A common number that we at
    Telebit use is about 1.5:1. This means that you can transmit 15K worth
    of data using only 10K of actual transmitted information. Thus it would
    take only 2/3rds the time to send the compressed version relative to
    the time it would take to send the original 15K. 

3.  With regard to compressing data in the modem versus in the computer, let's
    think about a hypothetical example. Suppose you have a modem link that 
    runs at 19.2K (perhaps using a Telebit TrailBlazer). So the maximum speed
    that the link could run would be limited by the RS-232 connection between
    the modems and the computers. OK? 

    Now, if we compress in the computer, we can send the COMPRESSED output out
    at 19.2K. This means the ACTUAL THROUGHPUT would be the 19.2K link speed
    times the compression ratio. If we use a compression ratio of 1.5:1, then
    the actual throughput would be 28.6K. But if we did compression in the
    modem, we'd always be limited by the link speed to 19.2K. OK?

    You can achieve throughput rates above the link speed by pre-compressing
    the data in the computer. You are limited by link speeds if you compress
    in the modem.

4.  So the general rule of thumb is, only compress ONCE in the data stream, 
    either in the hardware (modem) or in the software (computer). Usually, it 
    will be desireable to compress in the computer, unless the computer cycles
    are scarce or expensive.
    
I hope that this helps clear up some of the confusion about compression. Let
me know if there are any other questions/comments about modem technology.

Regards,

================================================================================
Richard Siegel                           Phone:                   (415) 969-3800
Senior Systems Engineer                  UUCP:  {uunet,ames,hoptoad}!telebit!rls
Telebit Corporation                      ARPA:  telebit!rls@ames.ARPA
                                
            "When the going gets tough, the weird turn pro"...HST
================================================================================