Adisak Mekkittikul

Achieving 100% throughput in an input-queued switch

It is well known that head-of-line blocking limits the throughput of an input-queued switch with first-in-first-out (FIFO) queues. Under certain conditions, the throughput can be shown to be limited to approximately 58.6%. It is also known that if non-FIFO queueing policies are used, the throughput can be increased. However, it has not been previously shown that if a suitable queueing policy and scheduling algorithm are used, then it is possible to achieve 100% throughput for all independent arrival processes. In this paper we prove this to be the case using a simple linear programming argument and quadratic Lyapunov function. In particular, we assume that each input maintains a separate FIFO queue for each output and that the switch is scheduled using a maximum weight bipartite matching algorithm. We introduce two maximum weight matching algorithms: longest queue first (LQF) and oldest cell first (OCF). Both algorithms achieve 100% throughput for all independent arrival processes. LQF favors queues with larger occupancy, ensuring that larger queues will eventually be served. However, we find that LQF can lead to the permanent starvation of short queues. OCF overcomes this limitation by favoring cells with large waiting times.

A practical scheduling algorithm to achieve 100% throughput in input-queued switches

Input queueing is becoming increasingly used for high-bandwidth switches and routers. In previous work, it was proved that it is possible to achieve 100% throughput for input-queued switches using a combination of virtual output queueing and a scheduling algorithm called LQF However, this is only a theoretical result: LQF is too complex to implement in hardware. We introduce a new algorithm called longest port first (LPF), which is designed to overcome the complexity problems of LQF, and can be implemented in hardware at high speed. By giving preferential service based on queue lengths, we prove that LPF can achieve 100% throughput.

CORD: contention resolution by delay lines

The implementation of optical packet-switched networks requires that the problems of resource contention, signalling and local and global synchronization be resolved. A possible optical solution to resource contention is based on the use of switching matrices suitably connected with optical delay lines. Signalling could be dealt with using subcarrier multiplexing of packet headers. Synchronization could take advantage of clock tone multiplexing techniques, digital processing for ultra-fast clock recovery, and new distributed techniques for global packet-slot alignment. To explore the practical feasibility and effectiveness of these key techniques, a consortium was formed among the University of Massachusetts, Stanford University, and GTE Laboratories. The consortium, funded by ARPA, has three main goals: investigating networking issues involved in optical contention resolution (University of Massachusetts), constructing an experimental contention-resolution optical (CRO) device (GTE Laboratories), and building a packet-switched optical network prototype employing a CRO and novel signaling/synchronization techniques (Stanford University). This paper describes the details of the project and provides an overview of the main results obtained so far.

Tiny Tera: a packet switch core

In this paper, we present the Tiny Tera: a small packet switch with an aggregate bandwidth of 320Gb/s. The Tiny Tera is a CMOS-based input-queued, fixed-size packet switch suitable for a wide range of applications such as a highperformance ATM switch, the core of an Internet router or as a fast multiprocessor interconnect. Using off-the-shelf technology, we plan to demonstrate that a very highbandwidth switch can be built without the need for esoteric optical switching technology. By employing novel scheduling algorithms for both unicast and multicast traffic, the switch will have a maximum throughput close to 100%. Using novel highspeed chip-to-chip serial link technology, we plan to reduce the physical size and complexity of the switch, as well as the system pin-count.

Novel distributed slot synchronization technique for optical WDM packet networks

Recently, a number of projects aimed at the implementation of optical packet-switched network prototypes have been announced. A common and critical element of these networks is the need to achieve time-slot synchronization over the whole network. Stanford University is currently implementing CORD, a two-node optical wavelength-division multiplexing (WDM) packet-switched network testbed with a star topology, featuring all-optical receiver contention resolution by means of optical switches and delay lines. This paper describes a flexible, distributed digital slot synchronization technique for CORD which is robust and scalable.

Small high-bandwidth ATM switch

The tiny tera is an all-CMOS 320 Gbps, input-queued ATM switch suitable for non-ATM applications such as the core of an Internet router. The tiny tera efficiently supports both unicast and multicast traffic. Instead of using optical switching technology, we achieve a high switching-bandwidth by using less expensive and proven CMOS technology. Because of limitations in memory and interconnection bandwidths, we believe that to achieve such a high-bandwidth switch requires an innovative architecture. By using virtual output queuing (VOQ) and novel scheduling algorithms, the tiny tera will achieve a maximum throughput close to 100% without the need for internal speedup.

Multiterabit-per-second ATM interconnection through optical WDM

Optical switching of Tbit/sec ATM streams through dynamic WDM is proposed. Optical manipulations are concentrated in a small area, overcoming technological obstacles. Central control is avoided via optical carrier detection. Multimedia and parallel computing applications are supported.

8-Tb/s ATM interconnection through optical WDM networks

We propose a novel scheme for interconnection of multiple high-speed (2.5/10 Gb/s) ATM streams through an optical WDM network, with a total network capacity of up to 8 Tb/s. The proposed architecture is based on placing the optical WDM portion of the network in a physically small area, i.e., one central office, or in a single rack. This helps to avoid technological obstacles such as power budget, dispersion, and synchronization limitations, and optical output buffering. The interconnection is an ATM packet switched network, and provides optical contention resolution. We show that the implementation of such a network is possible using currently available optoelectronic technology. Simulation results are presented, indicating network throughput of up to 100%.

CORD - Toward Optically Transparent, WDM Packet Switched Networks

In an optically transparent network, payload data are transported from source to destination completely in the optical domain. A primary advantage of optical transparency is avoiding the throughput limitation imposed by intermediate electronic switching, routing, and optical to electrical and electrical to optical conversions. An additional benefit of optically transparent networks is their indifference to modulation format and bit rate, allowing them to efficiently evolve to handle higher speed and heterogeneous traffic. Wavelength division multiplexing (WDM) is a promising method to access the huge bandwidth of single mode optical fiber. The bursty nature of local area networks (LANs) and the projected dominance of the asynchronous transfer mode (ATM) protocol make it desirable for optical LANs to be capable of packet switching. WDM is well suited for efficient packet switching in optical networks, but limited network resources may give rise to packet contentions. Stanford University, GTE Laboratories, and the University of Massachusetts, have formed a consortium to investigate methods to resolve packet contentions in the optical domain. Our consortium, CORD (Contention Resolution using Delay lines), has developed a WDM network test-bed to demonstrate our Contention Resolution Optics (CRO). The CRO, consisting of optical switches and delay lines, is a space and time switch, capable of rearranging the order of arrival of information packets. As is shown in Fig. 1, the CRO can be used in conjunction with a wavelength division demultiplexer to resolve contentions between packets of different wavelengths in a WDM network. Optical amplifiers are included in the CRO to offset coupling and component losses. Theoretical calculation shows that the 2 stage CRO shown in Fig. 1 can decrease the packet loss probability due to contentions by one to two orders of magnitude, depending upon the traffic load [l]. A significant portion of the CORD effort is devoted to the development of a compact, lossless CRO consisting of semiconductor optical switches and optical amplifiers, integrated on an InP substrate. To experimentally demonstrate the effectiveness of the CRO, a 2-node star network testbed has been built (Fig. 2). Nodes transmit 53 byte ATM packets on unique wavelengths at a payload channel rate of 2.488 Gb/s/node. A CRO is used to resolve receiver contentions at one node.