[SPAM] Re: SCP speed question

[Bill Campbell]

On Wed, Jan 10, 2007, Alma J Wetzker wrote:
>Rick Bowers wrote:
>> At 1/6/2007 12:08 AM, you wrote:
>>> Are you sure you're not seeing between 4 and 11MB/s?  Most measurements of
>>> transfer speed are done in BYTES not bits, as the network gear is rated.
>>> Approx divide by 10, particularly over wan links (analog: 8 bits data, 1
>>> start, 1 stop).  Or if you want to think perfect world switching, /8 is more
>>> your cup o' tea.
>> 
>> Perhaps I'm wrong, but I thought start/stop bits applied only to 
>> synch/asynch communications. TCP/IP has no start/stop bits as far as 
>> I know. So, don't we (mostly) all divide by 8 these days?
>
>The only time I am aware of that you can divide by eight is on a pure
>synchronous connection with out of band signaling.  That ain't Ethernet.
> At a physical level, you need some mechanism to make sure the send and
>receive clocks are the same speed, at the very least.  That overhead
>will occur well below the TCP layers.  It still is a divide by ten rule.
> The exception is ISDN (and some types of DSL) that are pure synchronous.

This whole argument is about as silly as NFL referees measuring
things to inches that are accurate to a yard or so at best.  It's
like trying to get 10 significant digits of accuracy when using a
slide rule for the calculations.

Whether one divides by 8 or 10 here doesn't make a whole lot of
difference when one is doing rough performance checks, and really
are relevant only in evaluating whether a change makes it better
or worse.  It's the relative changes that are meaningful.

Bill
--
INTERNET:   bill at Celestial.COM  Bill Campbell; Celestial Software LLC
URL: http://www.celestial.com/  PO Box 820; 6641 E. Mercer Way
FAX:            (206) 232-9186  Mercer Island, WA 98040-0820; (206) 236-1676

``Virtually everything is under federal control nowadays except the
federal budget.''
    -- Herman E. Talmadge, 1975