Very Computer

by **David Golpi** » Mon, 21 Aug 1995 04:00:00

Which provides better throughput, CSLIP or PPP?
Thanks.

by **Joel Ka** » Mon, 21 Aug 1995 04:00:00

Quote:>Which provides better throughput, CSLIP or PPP?
>Thanks.

The short answer is it depends. The reality is that their
throughputs are so comparable that throughput is the _last_ consideration
when deciding which to use.

--
Joel Katz Corporate Web Pages - http://WWW.*Plus.COM
Information on Objectivism, Atheism, the 8031/8051 Microcontrollers, Linux
and more is available at "http://www.veryComputer.com/~stimpson/".

by **John Phillip** » Mon, 21 Aug 1995 04:00:00

>Which provides better throughput, CSLIP or PPP?

My provider has both CSLIP and PPP access, and I have measured ftp
download speeds using both.

The results are that PPP is on average 1% slower than CSLIP (depending
on file types and on several other circumstances). The acual speeds
achieved with uncompressible data were all within 5% of the theoretical
maximum obtainable (about 3.2 kbyte/s using a V.34 modem).

I estimate the potential experimental error in this test to be more than
the 1% average difference I see (probably closer to 5%).

Conclusion: For me, they provide the same speed within the limits of
experimental error.
--
John Phillips

by **Bernie Doehn** » Mon, 21 Aug 1995 04:00:00

: Conclusion: For me, they provide the same speed within the limits of
: experimental error.
: --

As far as I know the only real benefit of running PPP vs. CSLIP is that
PPP is capable of also passing non-tcp/ip packets like Decnet and
Netware.

Bernie

by **Steve Farre** » Mon, 21 Aug 1995 04:00:00

>>Which provides better throughput, CSLIP or PPP?

>My provider has both CSLIP and PPP access, and I have measured ftp
>download speeds using both.

>The results are that PPP is on average 1% slower than CSLIP (depending
>on file types and on several other circumstances). The acual speeds
>achieved with uncompressible data were all within 5% of the theoretical
>maximum obtainable (about 3.2 kbyte/s using a V.34 modem).

>I estimate the potential experimental error in this test to be more than
>the 1% average difference I see (probably closer to 5%).

>Conclusion: For me, they provide the same speed within the limits of
>experimental error.
>--
>John Phillips

I'm glad to see that people have had these results, but I would suggest
that this is not always the case. I have noticed about 10-20% faster
results with CSLIP. It appears to have less overhead. Also, i thought
any readers of this thread might also do well to be reminded to set
asyncmap 0 in their /etc/ppp/options, as it significantly increases
ppp performance (though i still max at 3.0 w/ppp, and have gotten 3.2
many times with cslip w/ftp.)

--
Stephen Farrell The Ben May Institute
The University of Chicago Image Analysis & Web Admin.

home:312-288-8184 http://student-www.uchicago.edu/users/spfarrel

by **Joel Ka** » Tue, 22 Aug 1995 04:00:00

>: Conclusion: For me, they provide the same speed within the limits of
>: experimental error.
>: --
>As far as I know the only real benefit of running PPP vs. CSLIP is that
>PPP is capable of also passing non-tcp/ip packets like Decnet and
>Netware.

PPP is more configurable and smarter about negotiating parameters
such as compression and IP addresses. I've heard reports that CSLIP is a
smidgeon faster but that may have a lot to do with the particular
hardware, software, parameters, and type of modems/connections used.

--
Joel Katz Corporate Web Pages - http://WWW.*Plus.COM
Information on Objectivism, Atheism, the 8031/8051 Microcontrollers, Linux
and more is available at "http://www.veryComputer.com/~stimpson/".

by **Warner Lo** » Wed, 23 Aug 1995 04:00:00

> PPP is more configurable and smarter about negotiating parameters
>such as compression and IP addresses. I've heard reports that CSLIP is a
>smidgeon faster but that may have a lot to do with the particular
>hardware, software, parameters, and type of modems/connections used.

In theory, CSLIP will always be faster than PPP. However, this
difference is going to be very small. PPP has an overhead of about 4
bytes per packet, while CSLIP has an overhad of 1. For most people,
the differences will be in the statistical noise (that is, you can't
tell from a sample which one is faster with any degree of statistical
certanty, unless you run something like 100 tests or so).

Chose whichever one gives you the most reliablity.

Warner
--

Makers of TIA, The Internet Adapter. http://marketplace.com/