On Mon, 2009-05-04 at 09:47 -0700, William "Chops" Westfield wrote:
> On May 4, 2009, at 8:04 AM, Herbert Graf wrote:
> 
> > hubs were horrible for performance at 10Mbps speeds (I remember
> > that little "collision" light on my old 10Mbps hub being lit almost
> > continuously when more then two machines were using the network), I
> > can't image how saturated a network would be at 100Mbps speeds if you
> > used a hub. Fine for debug, but hell for normal use.
> 
> Huh?  Hubs are (should be) equivalent to the bus cable.  Collisions  
> happen, and it's not nearly as big a deal as some people (token ring  
> zealots) would have you believe.  A simple TCP connection between two  
> hosts on a 10Mbps ethernet bus gets a collision every couple of  
> packets, and still sees 8+Mbps throughput (Biggs, et al.)

I used to run a network in a, let's just say, "cash constrained"
environment. These were places where each run of CAT3 actually had two
Ethernet plugs on each end. 

We used hubs, and things worked swimmingly until windows file sharing
started to get big. Once more then just a few machines started
transferring files, overall network performance completely tanked. More
important then bandwidth was the latency, the insane flood of collision
backoffs resulted in some machines taking SECONDS before they could get
their packet through (not good when most of the machines were doing
things over telnet).

Obviously better network planning would have lessened the problems, in
the end though I finally convinced them to move over to switches, the
network never experienced those issues again.

-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist