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

RE: stds-80220-requirements: Latency and packet error rates




Branislav,

There's one class of traffic that doesn't appear on your matrix: if you
intend to use the networks resultant from this standard in some process
applications then you need _determinism_ or _bounded_delay_delivery.

On Fri, 2004-02-27 at 16:54, Branislav Meandzija wrote:

> 
> Class                Attributes of Traffic
> -----------------------------------------------------------
> Conversational  |    Two-way, low delay, low data loss
>                 |     rate, sensitive to delay variations.
> -----------------------------------------------------------
> Streaming       |    Same as conversational, one-way,
>                 |    less sensitive to delay. May require
>                 |    high bandwidth.
> -----------------------------------------------------------
> Interactive     |    Two-way, bursty, variable
>                 |     bandwidth requirements moderate
>                 |    delay, moderate data loss rate
>                 |    correctable in part.
> -----------------------------------------------------------
> Background      |   Highly tolerant to delay and data
>                 |   loss rate has variable bandwidth.
> -----------------------------------------------------------
Deterministic	|	Application requires bounded maximum delay
		|	characteristics.


This requirement is omitted from the IETF diff-serv docs because it
simply isn't attainable across a multiple-segment routed network without
doing terminal violence to the connectionless nature of the network
layer.  But it's practical to meet such requirements by providing some
form of ___synchronous service in the bottleneck network segments
(inevitably the radio networks) and overprovisioning the wired segments
so they never become the bottleneck.  

We've been here before in wired networks; token networks had the
theoretical ability to deliver determinism and FDDI had practical
implementation.  We've been here before in some radio networks too --
Link 16 (aka JTIDS, aka TADIL-J) in the military delivers synchronous
service using a TDMA structure.  

Most of the requirements justification that I can produce is military in
nature -- weapons control is an obvious example.  But I would guess that
SCADA applications in commercial industry would be another example.  


BTW, viewing QoS requirements as purely latency ones is viewing QoS
through a keyhole.  QoS is adjusting the balance across several
competing needs:

1.  Bandwidth efficiency (aka throughput, but macro view)
2.  Micro performance issues, incl:
	a.  latency
	b.  interactivity
	c.  jitter
(You've got a good handle on this in your post)
3.  Determinism (aka bounded delay delivery)

Help?

> > 
-- 
Rex Buddenberg
Naval Postgraduate School
Code IS/Bu
Monterey, Ca 93943

831/656-3576