Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [RE] Is anything special required??



Fred,

Thanks for the well constructed post. Couple this with DVJ's goal of
building RE networks with "no loading or topology constraints" and a
provisioning scheme is clearly indicated.

My definition of overprovisioned is no link over 75% utilized in either
direction under any realistic load scenario. A network that achieves this
will have no problem reliably delivering real-time A/V services.

With topology constraints, I believe it is possible to build home-scale
networks with commodity Ethernet equipment that meet this overprovisioning
specification.

I believe the CE market will tolerate topology constraints if they allow us
to deliver a cheaper, simpler solution.

Kevin

-----Original Message-----
From: owner-stds-802-3-re@IEEE.ORG [mailto:owner-stds-802-3-re@IEEE.ORG] On
Behalf Of Tuck, Fred
Sent: Thursday, May 05, 2005 12:13 PM
To: STDS-802-3-RE@LISTSERV.IEEE.ORG
Subject: Re: [RE] Is anything special required??

All:

Let me throw out a few assumptions that will be typical in a home
environment.  I don't think over-provisioning will always work in these
areas.  Some of these may seem a little contrived but as I pointed out in an
email a couple of months ago we have to be very liberal in our data rate
assumptions for home environments.  Trick modes, large HD file transfers,
and new minimal compressed HDV and other formats will easily consume 10s if
not 100s of Mb/sec of bandwidth each.  

Assumptions / Realities of home market.

1.	Multiple switches:  Users are going to have clusters of AV equipment
interconnected with switches that will then be connected to a whole home
switch.  Total bandwidth over a single switch can exceed 1Gb (8 100t ports
full duplex = 1.6Gbs) If all of this traffic is going to another switch it
requires a link of greater than 1Gbs.  Does anyone expect that we will be
able to do 10G Ethernet links in a home environment?

2.	Switches connected by something other than Ethernet:  There will be
other network environments in the home that use existing coax such as MoCA.
These will be used as backbone links and they will run at lower data rates.
We must interoperate with these other networks and they will have QOS and BW
reservation.  This is necessary because many people will be unwilling or
unable to run cat5 or other new wiring within their home.  In some places
like Florida, where there is a lot of cement block construction, it is
almost impossible to run new wiring.

3.	Gigabit links:  Some people have suggested that in a year or so that
all new links will be 1Gb.  This works fine if one is going to a single
switch with a full bandwidth backplane.  The above problems still exist when
you go to multiple switches or hybrid environments.

4.	Properly engineered LANs:  Over-provisioning works when you know
what your application environment and data rates are going to be.  Or you
can build an environment where the switch interconnects can handle all of
the backplane traffic between two switches.  I do not believe that you can
guarantee either of these things in a home environment.
 
5.	There will be multiple HD TVs in homes.  17 inch LCDs easily handle
HD resolution.  Even 14 inch displays in laptops and portable players have
1280 x 768 resolution.

6.	People will be making very large data transfers 10s of GB.  Large
directories of Multi Mega pixel photos, HD home videos, and transfers to
portable devices.  Multiple servers or DVRs in a home.  Etc... 

Fred Tuck
EchoStar Data Networks

-----Original Message-----
From: owner-stds-802-3-re@IEEE.ORG [mailto:owner-stds-802-3-re@IEEE.ORG] On
Behalf Of David V James
Sent: Thursday, May 05, 2005 12:13 PM
To: STDS-802-3-RE@LISTSERV.IEEE.ORG
Subject: Re: [RE] Is anything special required??

Arthur,

>> I do not agree that the term "over provisioned" is an abstract quantity.
>> If a 100Mbps network can carry the traffic with bandwidth allocation
>> then a gigabit network will be over-provisioned.

Does this mean that your definition of overprovisioned is 10%?
I'm not trying to say that 10% is good or bad, just trying to
nail down a specific counterproposal for consideration.

Without a specific number, its hard to respond intelligently.
As an extreme example, 1% may make things more deterministic,
but 1% might be an unacceptable limitation in the marketplace. 


>> > Do you disagree on those goals, or on the conclusions
>> > we have reached, based on those goals?

Can we safely assume that the disagreement was in goals,
specifically 75% vs 10% of link utilization for RE traffic?

Since one can't affect non-RE stations, such as server-to-PC
data transfers, I suppose the 10% would only apply to RE
traffic restrictions. Is this the intent of your proposal?

Respectfully,
DVJ

 

>> -----Original Message-----
>> From: owner-stds-802-3-re@IEEE.ORG
>> [mailto:owner-stds-802-3-re@IEEE.ORG]On Behalf Of Arthur Marris
>> Sent: Thursday, May 05, 2005 12:17 AM
>> To: STDS-802-3-RE@LISTSERV.IEEE.ORG
>> Subject: Re: [RE] Is anything special required??
>> 
>> 
>>  
>> David,
>> 
>> > For some of us, the goals are to allow time-sensitive
>> > traffic to occupy 75% of the bandwidth on 100Mb/s links,
>> > with no other loading or topology constraints. This is
>> > a bit easier to deal with, since the abstract quantities
>> > of "over provisioned" and "would notice" are better
>> > quantified and therefore measurable/provable.
>> 
>> I do not agree that the term "over provisioned" is an abstract quantity.
>> If a 100Mbps network can carry the traffic with bandwidth allocation
>> then a gigabit network will be over-provisioned.
>> 
>> > While the preceding goals, there do seem to be a few
>> > special subscription and pacing requirements.
>> 
>> > Do you disagree on those goals, or on the conclusions
>> > we have reached, based on those goals?
>> 
>> I think the goal at this time should be to identify the needs and
>> requirements of various CE applications in terms of throughput, latency,
>> jitter, loss etc and see what enhancements need to be made to Ethernet
>> (if any) to address these. I think Denis said something similar in a
>> previous message.
>> 
>> Arthur.
>>