Bharat Tarachand Doshi

Dual (SONET) ring interworking: High penalty cases and how to avoid them

The Dual (that is, two-node) Ring Interworking (DRI) feature is required to guarantee sub-second demand restoration with multiple interconnected SONET rings. There is a capacity (and, indirectly, cost) penalty for implementing DRI, as rings with larger capacities will be required to canT the same set of demands with DRI than with single (that is, one-node) ring interworking (SRI). This paper considers in detail the DRI capacity penalty with a pair of interconnected rings. In this case, the DRI capacity penalty can range from 0% to 100%, and be higher than 1000% in a contrived example. The above observation also holds if we consider a chain of interconnected rings spanning a country and carrying primarily east-to-west or north-to-south traffic. This paper analyzes the components of this DRI penalty and outlines how this penalty can be reduced. The paper also shows that almost all of this penalty can be mitigated with Triple Ring Interworking (TRI). In TRI, the essential idea is to balance the inter-ring load across 2 links involving 3 nodes rather than a single link. Finally, in a large network of interconnected tings where each ring connects to multiple rings an extension of the TRI strategy to balance the total inter-ring load across multiple separate links involved in interworking between pairs of rings will be shown to be useful in reducing the DRI penalty.

Link layer retransmission schemes for circuit-mode data over the CDMA physical channel

In the last few years, wide-area data services over North American digital (TDMA and CDMA) cellular networks have been standardized. The standards were developed under three primary constraints: (i) compatibility with existing land-line standards and systems, (ii) compatibility with existing cellular physical layer standards that are optimized for voice, and (iii) market demands for quick solutions. In particular, the IS-95 CDMA air interface standard permits multiplexing of primary traffic (e.g., voice or circuit data) and secondary traffic (e.g., packet data) or in-band signaling within the same physical layer burst. In this paper, we describe two radio link protocols for circuit-mode data over IS-95. The first protocol, Protocol S, relies on a single level of recovery and uses a flexible segmentation and recovery (FSAR) sublayer to efficiently pack data frames into multiplexed physical layer bursts. We next describe Protocol T, that consists of two levels of recovery. Protocol T has been standardized for CDMA circuit-mode data as IS-99 (Telecommunications Industry Association, 1994). We provide performance comparisons of the two protocols in terms of throughput, delay and recovery from fades. We find that the complexity of the two level recovery mechanism can buy higher throughput through the reduced retransmission data unit size. However, the choice of TCP (and its associated congestion control mechanism) as the upper layer of recovery on the link layer, leads to long fade recovery times for Protocol T. The two approaches also have significant differences with respect to procedures and performance at handoff and connection establishment.

Adaptive digital access protocol: new features and performance improvements

This paper reports on a broadband multiple access protocol for bi-directional hybrid fiber-coax (HFC) networks. Referred to here as the enhanced adaptive digital access protocol (ADAPt+TM), it builds upon earlier work to define a medium access control (MAC) protocol amenable to a multiple service environment supporting subscriber access in HFC networks with tree and branch topologies. ADAPt+ efficiently supports different access modes such as synchronous transfer mode (STM), asynchronous transfer mode (ATM), and variable length (VL) native data (e.g., IP, IPX). This enhanced protocol adapts to changing demands for a mix of circuit- and packet-mode applications, and efficiently allocates upstream and downstream bandwidth to isochronous and bursty traffic sources. This paper describes: ADAPt+ for upstream communication and multiplexing/demultiplexing for downstream communication; its applicability to STM, ATM and other native data applications; and performance attributes such as bandwidth efficiency and latency.