Will Hill wrote:
> 256k?  They doubled the speed from a dismal 30 kilo-bytes/second to 60 
> kilo-bytes/second.  Yes, I just tested it again to verify.  

60KBps = 480kbps .  Still a bit off from 512kbps but round trip 
latencies and various other potential limiters come into play.  Traffic 
shaping and possible MTU changes may get you a bit closer though I'd 
think you'd never exceed 500kbps (62KBps) on any consistent basis due to 
overhead.

Look into wondershaper (http://lartc.org/wondershaper/ ) if your router 
is a linux box.  Many of the linux router distributions (smoothwall, 
shorewall) incorporate wondershaper.  Some of the netgear/linksys 
consumer broadband routers out now support limited traffic shaping 
abilities as well.  Primarily those boxes prioritize VoIP over other 
traffic and don't do the kind of real, granular IP traffic shaping that 
would best benefit you.

> I don't think this is a technical issue at all.  At Home did just fine 
> without 
> crimps, despite Napster and the Windoze worms of the day.  What's changed 
> between then and now?  What kind of technical issues drove them to dhcp 
> instead of simpler static ip addresses?  How about port blocking?  Looks to 
> me like Cox has had to bow to other, stronger interests.  

I'm not gonna argue with you what may have changed in the past few 
years.  I'd say a great many things have changed, but the upstream 
path's inherent limitation is the primary engineering problem for cable 
connectivity and always has been.  The math involved is well understood.

Customers on a single downstream connection in the US share a 6Mhz band 
and I'm going to surmise that Cox has gone to 256QAM modulation giving a 
raw throughput rate of 27Mbps on a single downstream.  On the upstream 
side, the widest available channel width is 3.2Mhz and the most 
efficient available modulation scheme is 16QAM in DOCSIS 1.0/1.1 
yielding 10.24Mbps raw to share amongst customers on a single upstream 
path.(1)  DOCSIS signalling and TCP/IP overhead further erode the 
available bandwidth. Other factors as well will limit available 
bandwidth, typically on the upstream side of the equation.

Getting to those maximum widths and modulations schemes for cable 
vendors is recent as much work had to be done to get the plant to 
support that without  major issues. (2) A lot of companies out there are 
still at 64QAM on the downstream and 1.6Mhz width and QPSK on the 
upstream.  Cox had to do a significant amount of work and buildup to 
support a 512kbps upstream; I talked to a couple of engineers at Cox 
about this, actually.

The biggest improvement DOCSIS 2.0 will offer if it ever gets deployed 
is changes in modulation schemes on the upstream, changes in contention 
rules on the upstream, changes in resiliency for operating in a "noisy" 
RF environments, and great QoS enhancements for supporting 
bandwidth-intensive "real time" apps like VoIP.  It's all geared towards 
increasing the upstream available bandwidth.

(1)http://www.cisco.com/warp/public/cc/so/cuso/sp/hfcn_wp.pdf page 67, 
appendix table. This doc is not an easy read.
http://www.cisco.com/en/US/tech/tk86/tk168/technologies_tech_note09186a0080094545.shtml
"Understanding Data Throughput in a DOCSIS world" is a better read 
mostly for providers but there is some info on customer changes in PC's 
and OS that can improve throughput.
http://www.cisco.com/en/US/tech/tk86/tk804/technologies_tech_note09186a00800a9702.shtml
is a more "read friendly" document and though a little long in the tooth 
  in some respects it gives a good overview of how they're making decisions.
www.broadbandreports.com - the goldmine for users for information on 
this stuff.

(2)Did Cox in BR do a good enough job cleaning up the plant, 
reengineering, etc is a very subjective question, of course.


-- 
Scott Harney <[EMAIL PROTECTED]>
"Asking the wrong questions is the leading cause of wrong answers"
gpg key fingerprint=7125 0BD3 8EC4 08D7 321D CEE9 F024 7DA6 0BC7 94E5

Reply via email to