Thread: PPP versus ATM from the DSLAM to the subscriber

PPP versus ATM from the DSLAM to the subscriber

The Asynchronous transfer mode "cell tax" makes ATM an inefficient trans¬port method for data-only networks. We have already seen that ATM is a tedious and inefficient way of deploying IP access due to the huge number of PVCs that are generated by such an approach. Let us now look at the cost ATM brings in terms of efficiency. Data encapsulated in ATM typically uses AALS: An ATM cell bearing AALS has the following overhead.

IP is transported over ATM in many ways, all of which use AALS. One can use "Classic IP over ATM", PPP over ATM (PPPoA), or Multi-protocol Encapsulation over Asynchronous transfer mode (MPoA). Regardless of the means used to transport IP over ATM, the amount of overhead used is the same since LLC or LLC/SNAP headers are used to encapsulate the start of the data and an AALS-specific trailer, plus padding is added at the end of the IP data. See figure I for protocol details of the different encapsulation types using a sample 64-byte IP control message.

Table 1 shows the overhead impact created by the use of ATM encapsulation. Different packet sizes are examined including the average Internet package size. In all cases, the smaller the packet size, the more bandwidth that is wasted on encapsulation and therefore the lower the throughput of the connection.



On the other hand, PPP has a much lower overhead than ATM (see table 2). If the idea is to provide an Internet access service over the local loop, using ATM instead of PPP can mean a difference in overhead of more than 18%. In other words, if the last-mile solution employs ATM, there will be a loss of throughput due to bandwidth inefficiency of at least 18%.

Overhead differences between ATM and PPP have a significant impact when converting subscribers from Tl/El links requiring IP encapsulated services DSL loops. When replacing a traditional Tl/El circuit that is used for Internet access with a symmetric DSL solution such as G.SHDSL, an Asynchronous Transfer Mode based service will actually reduce the bandwidth available to the customer by 18%. With PPP the loss of bandwidth is 0%, assuming that the subscriber is already using PPP or Frame Relay over the Tl/ El. If Asynchronous Transfer Mode is used, the subscriber will probably notice a degraded level of service.


Conclusion

In conclusion, ATM has failed to deliver on many of its promises. The use of ATM in the local loop for DSLAM based IP services has numerous drawbacks:

• Compared to PPP, ATM uses 18% more overhead, reducing overall throughput to the subscriber.

• When converting a subscriber from a frame-based protocol over Tl/El lines, the subscriber will notice a degradation of service.

• The need for ATM PVC aggregation platforms increases the cost and increases operational inefficiencies.

You will be better off with a PPP-based DSLAM than an ATM-based DSLAM if the intention is to increase operational efficiency, increase throughput to the customer when replacing Tl-based services, and deploy an overall more scalable solution.