ATM (with the answer!!!)

Sat Jul 2 16:17:14 UTC 2005

O ye of little packets (or cells), if you are going to complain about 
the question at least answer it.

ATM Adaptation Layer (AAL) 1, 2, 3 / 4 and 5

AAL 1 - CBR constant bit rate traffic for voice trunks and video.

AAL 2 - VBR variable bite rate with a timing relationship between sender 
and receiver aka voice over ATM VBR Codec.

AAL 3 / 4 -  VBR traffic with no timing relationship between sender and 
receiver allows both connection oriented and connectionless traffic types.

AAL 5 - is a thinner interface for those bearer services that do not 
require the services of AAL 3 / 4 it also has Available Bit Rate to take 
up any excess capacity in the circuit.

 From an engineering and implementation perspective, which vendor's 
equipment do you want to use and how do they implement AAL services.

In general on the ATM edge devices and switches I have worked on the 
propagation delay was 15 to 20 micro seconds per switch or edge device. 
Assuming your data will be traversing one to two switches at each end 
and several in the WAN then you can expect reasonably fixed delays of 
150 to 200 micro seconds with jitter parameters of less than 5 -10 micro 
seconds not including speed of light issues.

With routers you will need to turn buffering off and you will still have 
propagation in the double to triple milli-seconds range with jitter in 
the multi milli-seconds range.

With ATM you setup your VPI/VCI combinations and with current ATM 
equipment look at QOS for both the overall VPI and different VCCs. For 
widely disparagent QoS parameter sets, setup a separate VPI for each 
parameter set.

For an ATM circuit you can setup either a uni-directional channel or a 
bidirectional pair of channels with either symmetric or asymmetric 
bandwidth. You can setup either a point to point or point to multipoint 
configuration. With current ATM switches you can setup one VPI/VCI 
channel for a single channel distribution until you get to a common 
switch closest to the customer where that switch can replicate the ATM 
cell stream to as many output ports as required to minimize duplicate 
streams of data in the network.

Having been one of the MPLS "experts" as the vendors were building the 
equipment trying to do ATM style of QoS at layers three and four does 
not work. IP needs to enhance the PPP (point to point protocol) that 
runs on serial links such as POS to Muti-channel PPP. Multi-link PPP was 
developed to bond smaller channels together to have a larger one. MCPPP 
needs to implement SAR for different size packets to be able to run VoIP 
and large 2k, 4k, 8k and 16k packets on the same serial link. Several 
major router vendors have implemented their routers as large cell 
switches with packet SARing at input and exit. They have also 
implemented proprietary inter router links that keep the packets in cell 
form until they leave the companies routing domain. These cell switching 
routers use larger cell sizes than ATM but the switching is extremely 
fast with most of the qos and routing table functions occurring on the 
interface cards in either standard or custom network processors.

Being an "Enganeer" I want to apply the right tool to the right job. 
Tools come and go.

First there was point to point and then circuit switched. Then packet 
route, point to point rings, broadband circuit switched, broadband 
packet route and switch, fiber based point to point with WDM and DWDM.

ATM gave us zero hop routing of IP packets from ingress to egress 
Ethernet.(who cares about the overhead since typical IP packets have 
more overhead in them than ATM cells) Fiber DWDM with ROADM and the like 
will give us zero hop optical routing of IP packet streams from ingress 
IP aggregation point to egress point. This will probably use GMPLS with 
the L being a lambda not a label but who knows.

Time to get off my White hobby horse and go back to work.

If you have any questions e-mail me directly.

John (ISDN) Lee

I Still Do not kNow
Its Suites Dennis's Needs

Philip Lavine wrote:

>I plan to design a hub and spoke WAN using ATM. The
>data traversing the WAN is US equities market data.
>Market data can be in two flavors multicast and TCP
>client/server. Another facet of market data is it is
>bursty in nature and is very sensitive to packet loss
>and latency (like voice). What type of ATM AAL format
>would be best for this topology? Is there any other
>concerns I should be aware of.
>
>Thx,
>
>Philip
>
>__________________________________________________
>Do You Yahoo!?
>Tired of spam?  Yahoo! Mail has the best spam protection around 
>http://mail.yahoo.com 
>
>
>  
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.nanog.org/pipermail/nanog/attachments/20050702/b28b2747/attachment.html>