The operator’s paradox for the past several years has been that while there is an explosion in data traffic volumes to the tune of 45-65% yearly, the corresponding revenue growth is in the single digits at best. To bridge this gap, providers must assess how to better architect their networks to reduce the transport-cost per bit, conserve space and power, and improve network performance to lower the opex. (Statistics show that service providers spend almost 5 opex dollars for each capex dollar![1]). They must also optimize their networks to efficiently carry high growth services like Internet access, packet traffic from 3G/4G wireless networks, and video.

Figure 1. Neilsen’s Law of Internet Bandwidth, predicted in 1998, has shown accurate and consistent for over a decade (From Jakob Neilsen’s Alertbox http://www.useit.com/alertbox/980405.html ).

Achieving this efficiency entails a tighter integration between the packet and the optical/photonic layers, since the photonic layer is the cheapest perbit, per function, thus motivating the packet-optical integration.

Major Solution Drivers

Technological advances (e.g. cloud computing, remote diagnostics, multimedia collaboration), bandwidth in tensive applications (e.g. video services with HD, Carrier Ethernet enterprise services, remote data backup and disaster recovery), and fast connection speeds, which according to Nielsen’s Law double every 21 months (Figure 1), lead to a proliferation of data packets and drive the demand for a better networking solution.

In addition, some key enterprise trends contribute to this traffic. For instance, almost 95% of enterprise traffic is now Ethernet-based. Indeed, business Ethernet port demand was up 43% in 2008 alone. Further, almost 80% of traffic now leaves the enterprise (the reverse of what it was just a little over a decade ago) implying a much greater load in the metro and core[2].

Thus, a key impetus for carriers is to increase the effectiveness and efficiency of transporting these packets over an optical transport network in the WAN environment.

Today, the IP/Ethernet packets are wrapped into SONET/SDH or G.709 TDM circuits, and transported over wavelengths on an optical infrastructure. One disadvantage of this is that when all switching occurs in a Layer 3 router/switch rather than judiciously leveraging Layer 2 Ethernet or Layer “2.5” MPLS switching, the cost of the network begins to increase. Con sequently, control layer mechanisms, such as MultiProtocol Label Switching-Transport Profile, MPLS-TP (e.g. RFCs 5654, 5317, 5718, 5860;[3]), or Provider-Backbone Bridging-Traffic Engineering, PBB-TE (IEEE 802.1Qay standard), are becoming important. Plus, the transport of IP/Ethernet over optical infrastructure is moving to sending native IP/Ethernet over wavelengths via WDM, which requires newer packet-optical solutions.

The particular solution adopted will be dictated by a number of factors. For example, the balance between the extent of connection-oriented (TDM) traffic and pure datagram traffic; the exist ing capital investment in SONET/SDH ADMs, ROADMs, switches and routers; the degree of equipment consolidation needed/desired to reduce opex; desire to use the wavelengths better; the OAMP&T (operations, administration, maintenance, performance and troubleshooting) provided by the deployed technologies; and whether IP/MPLS expertise and transport expertise resides in a common team or in different parts of the providers’ organization.

Characteristics of a Packet-Optical Solution

So what are the key ingredients being looked upon by operators in a packet-optical solution? It turns out that the following 4 elements are becoming table stakes:

  • Reconfigurable Optical Add/Drop Multiplexer (ROADM) infrastructure with support for routing wavelengths at multi-degree junctions, as well as the simpler two degree nodes [4].
  • The ability to efficiently carry existing SONET/SDH services without compromising support for high-growth packet and OTN traffic.
  • Connection-oriented Layer 2 Ethernet switching and aggregation.
  • Carrier-grade OAM — merging what exists in the optical domain with what exists in the packet domain to give an operator a comprehensive view of the network.

Thus a general industry consensus is emerging on the requirements of a Packet-Optical Transport System (POTS).

Three Key Areas of Advancement

The development of packet-optical solutions has involved advancements in 3 key areas: subsystems such as ROADMs and PICs; systems and ASICs (such as Packet-Optical Transport Systems); and control and management plane software.

1) ROADMs & PICs

ROADMs have played a key role in moving the transport network toward greater agility/flexibility by reducing the manual intervention needed to set up new lightpaths. A ROADM is composed of a number of sub-systems such as Wavelength Selective Switches (WSSs), optical amplifiers, optical channel monitors, transponders, and control and management software. A ROADM eliminates costly optical-to-electrical conversions at intermediate nodes by allowing wavelengths to pass intermediate nodes in the optical domain.

Figure 2. Conceptual Operation of a Wavelength Selective Switch.

First generation ROADM’s allowed a lightpath’s direction to be changed, while its wavelength remained fixed. They were typically 2-dimensional nodes that enabled ring architectures. Sub sequent ROADMs had higher degrees of between 4-8, allowing for mesh architectures.

Second generation ROADM’s used tunable lasers and wavelength selective switches (WSSs), allowing both the direction and the wavelength of a lightpath to be changed. WSS modules are the building blocks for ROADMs that can handle any wavelength on any port (and so are known as ‘colorless’) and can connect signals flowing in any direction on any port to any other port (hence ‘directionless’).

The next-generation of ROADMs will be gridless and contentionless. A contentionless ROADM allows multiple copies of a given wavelength (coming from different directions) to be dropped at a node, while a gridless ROADM has the capability to accommodate wavelengths that do not fit on the ITU 50 GHz or 100 GHz grid, but will utilize a flexgrid with a less rigid channel spacing (where some or all of the channels could use more than the standard 50GHz bandwidth). This allows for variable channel widths and enables operators to efficiently use spectrum to maximize fiber capacity. They will also incorporate fast switching speeds to decrease latency, and superior optical channel monitoring at the ROADM ports to better regulate signal power.

Photonic Integrated Circuits (PICs) have shown to be very effective in reducing the cost of the DWDM systems deployed by operators [5]. For example, Infinera’s PIC-based transport system is the #1 most widely deployed DWDM system in North America and includes a PIC-based Line Module with more than 100 optical components (lasers, modulators, wavelength lockers, etc.) integrated on a single monolithic Indium Phosphide chip approximately 5mm square. Next generation PICs are now under development to incorporate more complex modulation schemes such as QPSK and QAM, which are required to achieve 100Gbps per wavelength and higher and achieve aggregate capacities of 500Gbps or 1Tbps per PIC, and more than 10Tbps per fiber over long-haul networks.

2) Packet-Optical Transport Systems

Packet-optical transport systems/ platforms (P-OTS or P-OTP) are a new class of networking platforms that combine the functions and features of SONET/SDH/OTN ADMs or cross-connects, Ethernet switching and aggregation systems, and WDM/ROADM transport systems into either a single network element or a small set of network elements.

The goal of a Packet-Optical Transport System is to combine the best features of all of the legacy technologies, such as SONET/SDH, IP, ATM, and Ethernet. As a result, the requirements can be thought of as drawing upon the features of each technology in the following way:

  1. From SONET/SDH: Resilience — 50 ms recovery, path provisioning, and OAM.
  2. From ATM: Sophisticated Traffic Management and QoS as in ATM, including traffic engineering and guaranteed QoS.
  3. From IP/Ethernet: Very high efficiency from statistical multiplexing of packets/frames, and packet-flow control that are key for multimedia traffic.
  4. Flexible grooming or the ability to efficiently map a rich service mix onto the underlying transport layer by switching at the wavelength (lambda) level, sub-wavelength (ODU) level, port (TDM or SONET/SDH) level, and sub-port/packet (Ethernet, MPLS) level.

P-OTS architectures may be divided into three broad types:

  1. IP-over-Glass or Layer 3 routers with integrated transponders connected to a DWDM system. These rely on the router to perform switching function and eliminate O-E-O interfaces. Network architecture is simplified by eliminating SONET/SDH, thus reducing Capex and Opex.
  2. Carrier Ethernet Switch Routers with Connection-Oriented Ethernet (COE) controlled using PBB-TE or MPLS-TP plus a DWDM layer. The goal here is to leverage the low cost points of Ethernet, while getting the advantage of its traffic management and traffic engineering capabilities.
  3. Packet-Optical Devices combining SONET/SDH and IP/Ethernet switching/aggregation with DWDM transport. They emphasize a modular architecture, where sub-wavelength multiplexing and packet switching are done and traffic is groomed onto DWDM transport. These systems permit router bypass of non-IP traffic (e.g. L2 traffic, TDM traffic, and transit traffic), and minimize wavelength requirements by integrating SONET/SDH, MPLS, and OTN switching onto a single system.

The best alternative will depend on the existing and projected traffic mix (TDM to packet balance in the operator’s network), existing capital investment in network assets (SONET/SDH ADMs, ROADMs, switches/routers), need for efficient utilization of optical resources (wavelengths), and the carrier’s operations model (i.e., whether the IP/MPLS and transport teams are separate or common).

3) Photonic Control-Plane Software

The data plane, comprising flexible ROADMs and packet-optical transport systems, must be complemented by a highly integrated management and control plane that spans the packet, TDM, and optical domains. This control plane software is critical for future agile optical networks.

Figure 3. Packet-Optical Transport Systems (P-OTS): Architectures in use today

The control plane, which uses routing and signaling to set up the connections between nodes, coupled with an efficient management plane, is essential to orchestrate the operations of the data plane. Developments in the control plane are occurring within the IETF, which has developed the GMPLS control plane that is now being refined to include wavelength switched optical network (WSON) requirements. This will allow the control plane to have simplified knowledge of the optical parameters (such as chromatic dispersion and polarization mode dispersion) and simple rules that can be used to decide whether an optical path is adequate or requires signal regeneration.

The ITU-T has developed control plane requirements and architecture, under the umbrella of ASON (Automatically Switched Optical Networks). The GMPLS/ASON control plane comprises a common part and a technology-specific part to include technologies such as SONET/SDH, OTN, wavelengths, and MPLS-TP. By combining electrical and optical switching and an integrated control plane, the operators will be able to continually optimize their networks, and devolve them to the lowest-cost and most power-efficient solutions.

The User-to-Network Interface (UNI) and the External Network-to-Network Interface (E-NNI) implementations, based on the ITU-T, OIF and MEF standards could prove very useful for carriers. The UNI standards should enable operators to have packet switching devices that can signal the agile optical network, and request wavelength services for certain duration over a specific path and with a defined level of protection. E-NNI implementations will enable Wavelength Networks to share topology and availability information in a way to facilitate service deployment across multi-vendor (and possibly even multi-carrier networks) in an end-to-end manner.

Open Issues

Even as advancements in packet-optical integration continue to be made, challenges remain before a fully agile optical network is a reality.

An important consideration is providing the control plane with knowledge of the optical impairments, and enabling routing transparently between vendors. Similarly, handling increasing customer application rates, say 1, 10 or 40 Gb/s on 100 Gb/s infrastructure, will require the use of OTN (G.709) multiplexing and electrical switching, plus control plane support.

Finally, modularity of the system makes the challenge of integration for an operator much easier. This modularity comes in multiple forms, as universal switch fabrics and the ability to mix-and-match linecards (from all TDM to all packets and everything in between), or as modularity of the associated software with the ability to selectively turn on or off specific features.

This article was written by Vishal Sharma, Principal Technologist & Consultant, Metanoia, Inc. (Mountain View, CA), and Mark Allen, Ph.D., Director of Systems Engineering, Infinera Corporation (Sunnyvale, CA). For more information, contact Mr. Sharma at This email address is being protected from spambots. You need JavaScript enabled to view it., Dr. Allen at This email address is being protected from spambots. You need JavaScript enabled to view it., or visit http://info.hotims.com/34451-201.  

References

  1. Michael Kennedy, “Sizing-Up The Approaches,” Presentation, Network Strategy Partners, Fierce Telecom: PacketOptical Networking Platforms Webinar, July 14, 2010.
  2. Steven Gringeri, Bert Basch, Vishnu Shukla et al, “Flexible Architectures for Optical Transport Nodes and Networks,” IEEE Comm. Mag., Vol. 48, Issue 9, July 2010, pp. 40-50.
  3. Matt Rossi, “Enterprise Bandwidth Consumption,” Presentation, Zayo Enterprise Networks, Fierce Telecom: Making the 100 Gb/s Connection Webinar, July 21, 2010.
  4. Internet Engineering Task Force IETF, “MPLS-TP Standard,” WikiPage, http://wiki.tools.ietf.org/misc/mpls-tp/ wiki/drafts, Accessed 12/29/2010.
  5. Mark Allen, Chris Lou, Serge Melle, Vijay Vusirikala, “Digital Optical Networks Using Photonic Integrated Circuits Address the Challenge of Reconfigurable Optical Networks,” IEEE Comm. Mag. Vol. 44, Issue 12, Dec. 2007, pp. 2-11.
Photonics Tech Briefs Magazine

This article first appeared in the February, 2011 issue of Photonics Tech Briefs Magazine.

Read more articles from this issue here.

Read more articles from the archives here.