[SunRescue] Sparc 10 cover plates
paul.pries at sonera.com
Tue Dec 21 07:30:25 CST 1999
mvergall at double-barrel.be wrote:
> On Tue, 21 Dec 1999, paul pries wrote:
> > > sadamson at kenan.com wrote:
> > >
> > > } On the other hand, if you load more than 5
> > > Mbit/s
> > > } on an ethernet
> > > } you will definately find it saturated,
> > >
> > > No you won't. Taking TCP as an example, a data packet
> > > containing 8K of data could get ack'ed by a 20 byte
> > > packet. Looking at it this way saturation
> > > appears nearer 9Mbit/s.
> > True. If you have only two units on the physical
> > network.
> > The more workstations you put on the network, the
> > lower
> > the saturation point gets. Under normal
> > conditions, in a
> > normal office LAN, the saturation point is
> > somwhere between
> > 5 and 6 Mbit/s.
> > When you pass 4 to 5 Mbit/s you'll notice an
> > inrease in collisions,
> > which causes retransmissions. Guess what happens
> > then... *grin*
> I don't think so...we have a small network 16 systems here on Two SMC Hubs
> and we get local trafic at 8.8 - 9.1 Mbit/sec on ordenary 10Mb/s network.
As you say, it's a small network. There are also
traffic patterns to take
into the calculation. I base my figures on what
I've seen in practical
life, sitting at customer sites with a sniffer.
The worst case I've ever seen is a saturation
point of 3.8 Mbit/s. But
then again, this is exceptional. Lots of koax
repeaters, some hubs with their TP segments. No
switching at all, just
one logical segment with a lot of traffic...
No one had ever thought of doing a proper design,
no one had thought
of traffic patterns or what protocols they had
It really was a mess.
What I'm saying is that with a good design,
understanding users and a
small enough network, you get good througput.
Otherwise, YMMV. :)
More information about the rescue