Cloud Scale Multi-terabit Transport – Scale Out Instantly

October 17, 2017

By Dan Parsons
Sr. Manager, Architecture Marketing


The instant scalability of compute, storage and network capacity using a virtual infinite pool of resources is one of the significant attributes of the cloud. Scaling up, or vertically, in a data center (DC) is when a single server utilizes more of its internal resources to enhance performance for an application. Scaling out, or horizontally, is when a DC utilizes a pool of resources from all accessible servers, either inside the same DC or within other DCs, providing a virtual infinite pool of resources.

Data Center: Scale Up and Out
Figure 1: Data Center: Scale Up and Out

The concept of scaling up optical capacity in transport networks involves simply adding more capacity to a network element (NE) and subsequently a link, either by utilizing a higher line rate or adding more optical channels. However, doing so at cloud scale means doing it instantly from a flexible, software-defined, pre-qualified, virtual infinite pool of optical capacity, typically in 100 gigabits per second (Gb/s) increments. Scaling out the transport network at cloud scale means instantly accessing all the capacity, optical and sub-optical, within the entire network, not just on a single NE or link.

Transport Network: Scale Up and Out
Figure 2: Transport Network: Scale Up and Out

Why It Matters

Digital business transformation not only needs more capacity but to support real-time interactions requiring low-latency connectivity. Urbanization is driving the need for digital services to be closer to the user, as almost two-thirds of the world will live in cities by 2030. As more DCs move closer to users, connectivity that scales with services is essential. As more DCs become connected to other DCs, as well as to users within metro areas, throughout regions and across continents, communication service providers (CSPs) will need to be able to scale out transport capacity at cloud scale – like DCs can.

Is It Possible?

To compare scaling out in the DC to scaling out in the optical transport network sounds reasonable – until one compares the cost of a virtualized pool of compute and storage resources on DC servers with that of a pool of hundreds of Gb/s on optical transport platforms. The pooling of compute and storage is done at an astronomical level, on thousands of servers in every DC, while optical transport capacity pooling has not been the industry norm. As a result, the component industry has invested in processor and storage technology and continually drives cost down and performance up through silicon integration. For the same reason, the optical component industry’s investment is focused primarily within the DC, targeting the high-volume application of high-speed, short-reach, low-cost pluggable transceivers for servers, switches and routers. Investment in optical technology outside the DC is primarily focused on achieving the highest capacity per wavelength, not on the number of wavelengths from a single optical component.

What if the cost of a capacity pool was only applied when used? What if the pool was available at every transport node in the network? What if it was accessed simply via a software application, with all the optical engineering aspects already calculated? What if once used, the capacity could be returned to the pool? The result would be a cloud scale optical transport network.

However, providing a cloud scale transport network with traditional networking equipment requires pre-deploying a pool of conventional transponders at major locations in a network of colorless, directionless and contentionless (CDC) reconfigurable optical add-drop multiplexers (ROADMs) for the reconfiguration of the network’s optical channel routing. As a result, capacity can be accessed on demand and flexibly managed across the network. The up-front capital costs of a dormant pool of hundreds of Gb/s of capacity via individually pre-installed transponders, however, can be enormous and therefore not economically viable.

Infinera’s cloud scale Intelligent Transport Networks use the embedded photonic integration-based fourth-generation Infinite Capacity Engine (ICE4) and Infinera’s Instant Network solution to provide cloud scalability to optical networking.  Capacity is available as a flexible, pre-qualified, multiple terabits per second (Tb/s) pool that can be accessed instantly at any node in the network with Infinera’s software suite. What makes Infinera’s solution truly cloud scale is compact, simple rack-and-stack NEs that can serve as disaggregated functions of traditionally larger optical transport NEs. For example, a single Infinera meshponder NE delivers multiple muxponder functions with sliceability, offering 1.2 Tb/s of cloud scale capacity in a single rack unit (1RU).

Scaling for a Hybrid Cloud
Figure 3: Scaling for a Hybrid Cloud

Immediate Value

The value of cloud scale transport can be seen in a hybrid cloud application. A hybrid cloud is when a private cloud operated by a private enterprise needs additional resources, and seamlessly accesses them from a public cloud operated by an internet content provider (ICP) over a data center interconnect (DCI) service. Suppose a CSP offering 10 Gb/s DCI services receives multiple requests for hybrid cloud connectivity for a two-month period, demanding an additional 50 Gb/s of transport capacity. Even though the network may be capable of multi-terabit transport, it will need to be scaled up to support 100 Gb/s. Unless there is 100 Gb/s of capacity sitting idle waiting to be turned up, the CSP must engineer and plan the new capacity, order the new hardware, wait for delivery and then install and commission the new service. After more than two months involving three different departments, the new 100 Gb/s capacity is turned up for two months and then sits idle, not generating revenue until the next time it is needed.

Using the cloud scale Infinera Intelligent Transport Network and Infinera’s Instant Network solution for this hybrid cloud scenario, the 100 Gb/s of capacity is scaled out instantly across the network. Not only are almost two months of the traditional deployment cycle eliminated, but the CSP can respond to business opportunities quickly, realize faster time to revenue and offer new 100 Gb/s DCI services like DC replication and disaster recovery. This is a huge competitive advantage with a reduction in total cost of ownership of more than 50 percent when moving services from 10 to 100 Gb/s, compared with more traditional optical transport networks.

Cloud scale transport not only enables new capacity quickly and easily, making the operational team happy, it also opens new business possibilities. Not only are the capital and operational cost savings compelling enough on their own, CSPs can address new customers faster, supply on-demand capacity requests or optimize network capacity through time-based services via Infinera’s Instant Network solution, for even more revenue without more costs.

Don’t let tradition get in the way of the future. Infinera delivers real cloud scale transport. To learn more, contact us.

Related Links:

Legal Disclaimer

Leave a Reply

Your email address will not be published. Required fields are marked *