Bringing Cloud Scale Capacity and Performance to a Metro Network Near You
Our highly connected world is on the cusp of the next big wave of innovation, with technology advancements in current applications and a plethora of demanding new applications and services about to hit our networks. New applications such as the Internet of Things (IoT), connected cars, and virtual and augmented reality are affecting both fixed and mobile networks, while existing video services are moving to higher resolutions and enterprise services continue to become centralized in the cloud (Figure 1). Collectively these applications require huge amounts of bandwidth, low latency, and flexible networks that can cope with increasingly dynamic traffic flows.
To support these demanding new applications, operators are building new fiber-deep network architectures that push fiber closer to end customers to support higher capacity and performance. This is seen in the cable world with the move to Remote PHY (physical layer), elsewhere in the migration to Long-Term Evolution Advanced (LTE-A) and ultimately 5G in mobile, and deeper fiber penetration in residential and business networks with fiber to the curb or the premises (FTTx).
These fiber-deep networks greatly increase the potential capacity per user with significantly more backhaul capacity for the access network, be it hybrid fiber-coaxial (HFC) in cable, wireless, or fixed access such as digital subscriber line (DSL) or xPON. However, they come with a critical challenge that many operators are now facing – deploying higher capacity latency-sensitive access and aggregation networks in network facilities with severely limited space and power.
The Space and Power Challenge!
Many of these networks need to bring considerable capacity upgrades to parts of the network that will use a combination of facilities for the rollout, including central offices, telecoms huts, and even street cabinets. These locations were never designed to house high-capacity transport equipment and are often limited in available space and power. Site expansion or power upgrades would be highly uneconomical or not possible at all. Even where space and power are available, they are at a premium in these facilities, and therefore both are treated as critical factors in the network design process.
This issue is further compounded by some of the new architectures that are being rolled out now or are in the planning stages. A good example is the move towards mobile edge computing (MEC) and fog computing, where the high capacity and low latency of the new network are used to simplify the end users' devices. These network architectures use an array of virtualized resources throughout the network to provide shared compute and storage capabilities that previously would have been dedicated to each end device. This brings many advantages, such as higher performance for the user/application and simpler, cheaper end-user devices with lower power consumption.
However, the new compute and storage resources need to reside somewhere in the network. While some are centralized in mega-data centers, a lot of these resources are still needed close to the end customer for latency reasons. This means they also need to utilize the same limited space and power that the high-capacity metro cloud network needs to use, compounding the problem (Figure 2).
Can DCI Platforms Play a Role?
In recent years, the growth of the cloud and cloud-based services has led to astronomical growth in the number and size of data centers and the volume of traffic handled by data center interconnect (DCI) networks. The optical networking industry has risen to the challenge of supporting these networks with an array of optimized high-capacity DCI products that offer incredible density and capacity per fiber and significant advances in reducing power consumption per bit. However, these products do not address the segment of operators' networks referenced in this article, namely cloud-scale metro networks that connect end users and access networks to the mega-data center.
These high-capacity DCI-specific products are highly optimized for Ethernet-based transport over point-to-point connections with primary focus on cost per bit, maximizing density with the most 100-Gbps (100G) services per rack unit (RU) of space, and maximum capacity per fiber pair. Many other factors come into play for vendor selection, including power consumption and ease of use, but the focus is on economics, density, and total fiber capacity.
Furthermore, because of this optimization for high-capacity 100 Gigabit Ethernet (GbE) transport, they are inherently not designed to support the aggregation of sub-10GbE, 10GbE, and 100GbE connections in a multi-service environment that is required in the metro network. Metro networks also need to consider legacy and non-Ethernet traffic, such as Fibre Channel (FC) for storage area networking (SAN) services or Common Public Radio Interface (CPRI) for mobile fronthaul, which must also be carried over the same infrastructure, if required.
Metro Core as an Option?
So, if DCI-specific products can't solve this challenge, could existing high-capacity metro core products? On the surface, from a feature checklist perspective, it may appear that they could fit the bill. However, while high-capacity metro core products support the key requirements of cloud-scale metro networks, these products also suffer from significant power and space challenges, as they are designed for central office environments where space and power are considerations but not severely limited resources.
To put these constraints into perspective, operators today are pushing metro aggregation networks out into locations where only three to five rack units of space are available, and possibly only 200 to 300 W of electrical power. Many metro core-focused products run up to or above the kilowatt power level for chassis of this size, and up to many kilowatts over for bigger chassis.
So, What Does the Industry Need to Address This Challenge?
The optical networking industry needs to blend the characteristics and performance of the DCI world, metro core, and metro access/aggregation into a new breed of cloud-scale metro products that can meet the next wave of operator transport challenges. These products need to provide a step change in networking capacity with 200G per wavelength, low latency, and support for the increasingly dynamic traffic flows that these new applications drive. Low power consumption and high density are also critical.
Furthermore, these products need to prepare operators for the next steps in high-capacity transport with an open flexible grid (flex-grid) line system that is prepared for the advanced modulation formats that will be needed as networks move to 400G, 600G, and 800G per wavelength and beyond. Finally, future metro networks will play a critical role in underpinning many architectures based on network functions virtualization (NFV) and software-defined networking (SDN), requiring complete control of the platform.
Many metro core products address aspects of this but do not have the focus on low power consumption and the required range of smaller and mid-sized chassis options to enable deployment closer to the edge in metro access or metro aggregation locations.
A New Breed of Optimized Cloud-Scale Metro Platforms Is Required
Today, a new breed of cloud scale metro platforms is becoming available that provides the functionality required with low power, low latency, and high density. These platforms use 16QAM-based 200G-per-wavelength technology to drive power consumption down to as low as 20 W per 100G service within a plug-in card. This low power consumption enables high-capacity technology to be deployed in locations that could never support the power requirements of alternative platforms. These platforms combine Layer 1 transponder and muxponder options to transport a broad range of services and, critically, also contain Layer 2 packet-optical technology to directly support many Layer 2-based access architectures. The inclusion of Layer 2 means that these platforms have flexibility at Layer 0, 1, and 2 to support bursty traffic flows created by the next wave of applications and services that networks will need to support (Figure 3).
These platforms support 200G per wavelength today, but are also ready for 400G per wavelength and beyond with flex-grid optical layer and other advances such as optimized hybrid Raman/erbium-doped fiber amplifier(EDFA) optical amplifiers. Together these capabilities enable the Layer 0 components of the network to support both the higher baud rates and advanced modulation formats that will be needed as networks move beyond 200G per wavelength. SDN control enables these platforms to support the virtualized-based future that is emerging for optical networking.
Operators face a huge challenge in pushing transport networks to higher capacities with low latency at the same time as pushing the optical network closer to the edge to support fiber-deep access architectures. Cloud-scale metro platforms will play a key role in connecting users to datacenters and transporting operators to the cloud-based future.
Jon Baldry is director of metro marketing at Infinera