On contention resolution in the data vortex optical interconnection network

Abstract

͑Doc. ID 80158͒ Alternative contention resolution techniques are studied in the data vortex interconnection network, namely, the insertion of fiber-delay-line (FDL) buffers into the switching nodes. The performance of each technique is evaluated according to relevant performance metrics: acceptance rate, mean latency, and latency variance. A detailed discussion concludes that while traditional data vortex networks perform better in terms of throughput, FDL-based switching nodes have a favorable impact in reducing the system latency.

The Data Vortex Optical Packet Switched Interconnection Network

“…Various scenarios have been simulated to illustrate the effect of the topology and network load on [9], [11]. The number of cascaded nodes directly affects the overall network latency.…”

“…A time-domain method for contention resolution for the data vortex architecture has been proposed involving the insertion of FDL-based recirculating buffers into the switching nodes [11]. Methodologies for contention resolution remain a central focus for the data vortex, which uses a combination of timeand space-domain contention resolution to implement virtual buffering.…”

“…Furthermore, nondeterministic routing through the network gives rise to the possibility of unbounded latency distributions; these not only are detrimental to latency sensitive applications, such as memory accesses, but also contribute to physical-layer-induced limitations on network scalability. Investigations have explored the utility of variations on the traditional data vortex network in an attempt to alleviate this shortcoming [11].…”

“…The data vortex topology [3] was first described in [4] and first investigated in [5]- [7], and its architecture was further analyzed in [8]- [11]. A 12-port data vortex prototype was implemented and its routing performance investigated [12].…”

New nodal design for high throughput data vortex optical interconnection network

“…One inherent limitation is the traffic backpressure due to the deflection based routing, and this leads to limited throughput and latency performance under heavier traffic conditions. Several approaches have been proposed to enhance the network performance, and most of them use additional hardware in the hope that future technological development will drastically reduce the cost of switching elements within the network [16][17][18][19]. Among these schemes, one relatively easy implementation is adding a buffer path within each of the routing nodes.…”