UltraFlow Access Testbed: Experimental Exploration of Dual-Mode Access Networks

Abstract

Electrical packet switching is well known as a flexible solution for small data transfers, whereas optical flow switching (OFS) might be an effective solution for large Internet file transfers. The UltraFlow project, a joint effort of three universities, Stanford, Massachusetts Institute of Technology, and University of Texas-Dallas, aims at providing an efficient dual-mode solution (i.e., IP and OFS) to the current network. In this paper, we propose and experimentally demonstrate UltraFlow Access, a novel optical access network that enables dual-mode service to the end users: IP and OFS. The new architecture cooperates with legacy passive optical networks (PONs) to provide both IP and novel OFS services. The latter is facilitated by a novel optical flow network unit (OFNU) that we have proposed, designed, and experimentally demonstrated. Different colored and colorless OFNU designs are presented, and their impact on the network performance is explored. Our testbed experiments demonstrate concurrent bidirectional 1.25 Gbps IP and 10 Gbps per-wavelength Flow error-free communication delivered over the same infrastructure. The support of intra-PON OFS communication, that is, between two OFNUs in the same PON, is also explored and experimentally demonstrated.

UltraFlow: Elastic, flexible and dual-mode long-reach optical access network

“…These scheduling nodes communicate scheduling information in order to assign dedicated wavelengths along the network path between the source and destination. This distributed scheduling structure can be seen as a single scheduling process that serves the incoming Flow requests in a First-Come-First-Served (FCFS) manner, such that they are buffered in a scheduling queue [3].…”

“…We follow the analysis in [3], and assume that the Flow request arrival process for each OFNU follows a Poisson distribution P(λ), with arrival rate λ = 1 request/sec., and the Flow transmission time follows a Uniform distribution U(a, b), with its minimum and maximum value to be 0.1 sec. and 0.8 sec.…”

“…Here, we employ the Next Available Supported Channel (NASC) [6] as the wavelength scheduling policy. The process is in general an M/G/N queuing problem [3], which doesn't have close form analytical solutions. When transceiver number N at the OFLT is large, it converges to an M/G/1 queuing problem (analytical close form Twaiting solutions available) [3].…”

“…Conventional IP network still offers the flexibility in handling bursty traffic, which is especially important in access network. Therefore, we proposed and experimentally demonstrated [2][3] the Stanford UltraFlow access network, a novel access technology that enables dual-mode service, IP and Flow, to the end-users, as shown in Fig. 1.…”

“…With a bank of tunable transceivers, the OFLT is able to perform dynamic wavelength allocation (DWA), and provides network-monitoring functions. The gateways are constructed with edged filters to separate or combine the Flow and IP channels based on their distinct wavelengths [3]. Furthermore, in order to provide flexible optical Flow service with extended service reach (distance up to 100 km), large subscriptions (end-user number up to 1024), lower capital expenditure (CapEx) and higher energy efficiency, we proposed and demonstrated a reconfigurable long-reach (R-LR) UltraFlow access network with centralized software-defined network (SDN) controller [4][5], as illustrated in Fig.…”