Chapter 3.9.3 - The RingO Project

3.9.3 The RingO Project

The RingO (Ring Optical network) project [71, 79] is carried out by a consortium
of Italian Universities coordinated by the Optical Communication and Network
groups of Politecnico di Torino and is focused on experimentally studying the
feasibility of a WDM optical packet network based on a ring topology. The
researchers participating in this project aim at the design and prototyping of
network architectures for metro applications, taking an approach based upon
optical packets, but limiting optical complexity to a minimum, and trying to use
only commercially available components [71]. In order to exploit the advantages
of both optical and electronic technologies, the bulk of raw data is handled in the
optical domain, and more complex network control functions (such as distributed
resource allocation and contention resolution) are mostly implemented in the electronic
domain [2].

Three alternative architectures have been proposed for the RingO network. The
first architecture is the simplest to implement and the most restricted in terms of

FIGURE 3.17 The simplest architecture proposed for the RingO metro network (unidirectional ring with the number of nodes equal to the number of channels).

performance. In this architecture (Fig. 3.17) network nodes are interconnected via a
unidirectional fiber ring. As a result, a packet transmitted from node 1 to its adjacent
node (node 2) must traverse nodes 5, 4, and 3 before it reaches its destination.

The main features of the first architecture suggested in RingO are the
following [71]:

Time is divided in slots and packet transmissions are synchronized across all
wavelengths.
The length of packets is fixed and equal to one timeslot.
The number of nodes N in the network is equal to the number of wavelengths
W; a given node i is thus identified by a wavelength λ_i. Node i is the only node
able to receive a packet using this wavelength and is also responsible for physically
removing it from the ring, using a fixed-wavelength optical drop filter.
Each node is equipped with a tunable transmitter in order to enable direct
communication with all other nodes. Before transmitting a packet to node k,
node i must tune its transmitter on wavelength λ_k. Tuning times are assumed to
be short with respect to the slot duration.
Each node is able to check the state of all wavelengths (busy or free) on a
slot-by-slot basis before transmitting data. The applied method is called
λ-monitoring and is used for collision avoidance. Input packets are queued
electronically and access to channels for transmission is regulated by a suitable
medium access protocol (MAC).

The fixed relation between a destination node and a wavelength in the architecture
described above allows a significant simplification in terms of optical hardware.
Routing decisions are not based on packet headers, because the destination address
is “coded” onto the used wavelength. Additionally, packets do not need to be
actively routed along the network, but are passively dropped at the destination by
the node optical drop filter. Optical headers are not required for monitoring purposes
either as l-monitoring can be implemented by simply measuring the power level
in each slot and wavelength [71].

If single-hop communication between all nodes is required, then either the node
transmitters or receivers must be capable of tuning over all used wavelengths.
Tunable transmitters are preferred over receivers because they are considered to
be easier to implement.

The proposed RingO architecture efficiently combines optical and electronic
technologies. Transmission of data is performed entirely in the optical domain
with a wavelength granularity, and packet queuing, MAC protocol, and statistical
time multiplexing are handled in the electronic domain at the speed of a single
data channel. The resulting solution, although not optimal, offers a set of interesting
features for metropolitan area networks connecting a limited number of veryhigh-
capacity nodes over a ring. The paragraphs that follow describe the structure
of a node in the RingO network.

The basic subsystems are common to all node architectures incorporated in
RingO. The main functions supported by each node (from input to output) are the
following:

Amplification of the optical signals in order to compensate for the losses due to
node passive elements and the downstream fiber link.
Demultiplexing of the WDM comb after the amplifier. In the first
RingO testbed this function was implemented using arrayed waveguide
grating filters.
Channel monitoring for each timeslot. This is performed by tapping a fraction
of the power on each fiber at the output of the demultiplexer and by sending it
to an array of photodiodes.
Detection of the incoming data stream on the preassigned wavelength.
Generation of local packet traffic. Instead of a tunable transmitter, an array of
lasers was used in the RingO testbed. An important advantage of this approach
is the efficient multicasting capability that it offers. If a node wishes to transmit
a packet to multiple destinations in the same timeslot, it can instruct the
corresponding lasers to emit at the appropriate wavelengths. In this way, multi-
casting in a single timeslot can be implemented without increasing electrical
bandwidth requirements at the transmitter, because the “replication” of
packets is performed in the optical domain.

From the description of the architecture, it is evident that the RingO architecture
requires an electrical data path bandwidth equal to a single channel data rate on both
the transmitter and the receiver side. In other words, equipment at each node must be
capable of operating at the channel data rate. This is true for all proposed RingO
designs and is considered part of the project rationale. Indeed, a metro architecture
capable of scaling at large aggregate capacities must avoid processing the entire
network bandwidth at each node.

In RingO, timeslots for transmission are allocated to network nodes by the MAC
protocol. From the MAC protocol design perspective, RingO is a multichannel
network, in which packet collisions must be avoided, and some level of fairness
in resource sharing must be guaranteed combined with acceptable levels of
network throughput. Collisions occur when nodes attempt to insert packets at
wavelengths that are already occupied in the given timeslot. Collisions are
avoided by giving priority to upstream nodes, that is, to in-transit traffic, via the
λ-monitoring capability.

Fairness is obtained by implementing an efficient a posteriori packet selection
strategy exploiting a virtual output queuing (VOQ) structure. The basic VOQ
idea, as applied in the RingO architecture, consists in storing packets waiting
for ring access in separated FIFO queues, each corresponding to a different destination,
and to appropriately select the queue that gains access to the channels for each
timeslot. Thus, head-of-line blocking is avoided and a packet waiting to be transmitted
on a busy wavelength does not prevent other packets from being transmitted
to other destinations.

Another important issue is the accumulation of physical impairments on the
optical signal resulting from the fact that signals propagate all-optically along the
ring without 3R regeneration. The node structure based on AWGs that was presented
earlier faces significant physical-layer performance limitations, which impair the
scalability of the proposed network. This motivated the design of an alternative
RingO configuration, which is based on an add–drop filter allowing for better
cascadability and less stringent physical constraints.

This second network structure is similar to the previous one in terms of network
functionalities, but it is significantly different from the physical layer point of view.
The input–output optical path is greatly simplified, and now consists only of a
passive optical splitter and a fixed add–drop filter tuned on the wavelength λ _drop
that must be received locally. This setup greatly reduces node attenuation, self-
filtering, and PDL effects, and allows higher node cascadability.

An important limitation of the two previously presented RingO architectures is
the fact that the number of nodes must not exceed the number of wavelengths available
on the ring; that is, N must be less than or equal to W. This constraint impairs the
scalability and flexibility of the proposed network. The third RingO architecture
overcomes this limitation by allowing multiple destinations (nodes) to receive on
the same wavelength. The same basic node architecture is maintained; that is,
each node is equipped with a fast tunable transmitter and a fixed receiver.

The distinctive feature of this third architecture is the separation between
resources devoted to transmission and resources devoted to reception. This separation
can be achieved in the wavelength domain (by using different wavelengths),
in the time domain (by using different timeslots), or in the space domain (by
using two different fibers). The third option was selected because it is easier to
implement. Therefore, the network comprises two fiber rings, one used for transmission
and the other for reception. These two rings are connected at some point
in order to allow a packet to move from one to the other. Transmitted packets
traverse the first ring and are then switched to the reception path (ring), which
they must traverse in order to reach their destination nodes.

Because there is no one-to-one relationship between nodes and channels, a given
node must not drop the packets carried on its assigned wavelength from the
ring because they may be destined for another node sharing the same channel.
Thus, headers are required to indicate packet destination addresses. Another issue
that arises concerns the optimal allocation of system wavelengths to different
receiving nodes.

< Previous Excerpt Next Excerpt >

Purchase This Book

TABLE OF CONTENTS

Chapter 3.9.3 - The RingO Project

Contact Preferences

This is embarrasing...

Customize Your GlobalSpec Experience

Select Your Free Newsletters

Industry Newsletters

Select Your Free Product Alerts

This is embarrasing...