Cloud Networking Applications, Deployment Models, and Trends
April 19, 2023
By Fady Masoud
Sr. Director, Solutions Marketing
Part 2: Deployment Models
As outlined in my previous blog post on cloud networking applications, cloud networking solutions heavily rely on the quality and performance of connectivity provided by the underlying optical network. This connectivity must have the highest levels of performance, whether across distances that are short, such as within a campus or city; medium, such as within a region; or long. This puts the optical network at the heart of the evolution to the cloud. So, how do you choose the right solution set for each data center interconnect (DCI) application?
DCI Deployment Models
First, let me start with the fact that the same DCI application can be served by many different deployment scenarios. Each has its own advantages and challenges. However, some deployment scenarios are more optimized around certain application requirements, as described below:
Router-hosted coherent pluggables
The first deployment model is often referred to as IP over DWDM (IPoDWDM). It is performed using coherent pluggables hosted directly in a router at each data center. A typical deployment is over a relatively short distance, such as metro DCI within a city, where the link between data centers is less than 120 km (Figure 1). This deployment model can be achieved using 400G ZR pluggables as they enable low cost, compact footprint, and low power consumption over a simple architecture. However, there are some considerations to be taken into account, such as the host router’s support for the QSFP-DD form factor and the ~15-20 W power envelope of these coherent pluggables.
When the distance between data centers is longer than 120 km, a higher-performance coherent pluggable, such as 400G ZR+, can be used to provide the reach and required optical performance to go through multiple cascaded ROADMs between two endpoints. Management also becomes a consideration. Whether 400G ZR or 400G ZR+ pluggables are used, network operators must ensure a simple and efficient management scheme as these coherent pluggables providing optical transport functions are now hosted in a device managed by the IP domain. While basic optical functions, such as setting the DWDM frequency, can be set using a register-based MSA information model, such as CMIS or C-CMIS, advanced optical functions required to support current or future transmission rates (e.g., modulation, power, etc.) are not.
A new breed of intelligent coherent pluggables, such as ICE-X optics, combine leading optical performance for regional/long-haul DCI, including +0 dBm launch power and the ability to go over cascaded ROADMs, with support for point-to-point and point-to-multipoint configurations in the same coherent pluggable. Some of their performance metrics have already been put to the test. Infinera and Arelion announced a record-setting transmission of 400 Gb/s wavelengths using a QSFP-DD pluggable coherent transceiver over 1,800 km of SMF-28 fiber in a live production network between Dallas, Memphis, and Chattanooga. These pluggables’ performance has been pushed even further – Infinera announced another record-breaking achievement of delivering a 400G single-wavelength transmission with our ICE-X 400G QSFP-DD intelligent coherent pluggable solution across 2,400 km of Corning’s TXF® optical fiber.
In addition to optical performance, the Open XR Forum has issued specifications that introduce a dual-management paradigm, allowing for disaggregated network management of Open XR intelligent coherent pluggables. This architecture uses and preserves the current register-based management path for applications and features supported in existing and future MSAs. This specification introduces an additional dual-management IP/Ethernet paradigm that will enable the industry to accelerate deployment of innovative features and rapidly introduce advanced applications with virtualized transport functionality. Without this new capability, the industry would face a significant delay from co-development and deployment on each hosting device required in the network. Although conceived as a method to rapidly introduce point-to-multipoint coherent optical applications, it also introduces advanced transport features and remote management for point-to-point applications. Intelligent system-level features such as streaming telemetry, dynamic capacity allocation, and intergenerational interworking are also among the key capabilities offered by ICE-X optics.
Figure 1: DCI with router-hosted coherent pluggables
Compact modular platform equipped with coherent pluggables
Another deployment model is using compact modular platforms with sleds that support coherent pluggables such as 400G ZR+, ICE-X 400G ZR+, or ICE-X 400G XR in a CFP2 form factor (Figure 2). One of the benefits of this deployment model is the fact that it is compatible with legacy routers that do not support coherent pluggable form factors (e.g., QSFP-DD) or their power envelope. However, the sleds in a compact modular platform must support coherent pluggables like 400G ZR+, ICE-X 400G ZR+, or ICE-X 400G XR, which are typically in a CFP2 form factor but are also evolving to support QSFP-DD, and short-reach client optics must be used between the router and the compact modular platform.
A good example of this deployment model is Infinera’s GX Series of compact modular platforms and XTM Series packet optical platform, which support 400G ZR, 400G ZR+, ICE-X 400G ZR+, and ICE-X 400G XR intelligent coherent pluggables. IP/optical demarcation and optical layer management of the compact modular platform are among the benefits of this approach. However, one of the considerations for this deployment model is the increase in footprint, given the 1RU or 2RU added by the compact modular platform to the overall footprint. Similar to the previous deployment model, the coherent pluggables must provide the required optical performance to go over the required distance, including any ROADM nodes in the optical path.
Figure 2: DCI with compact modular platform and coherent pluggables
Compact modular with embedded optical engines
This deployment model consists of taking advantage of the optical performance in optical engines, like Infinera’s ICE6, embedded in compact modular sleds to reduce the cost of transport and power per watt per km. When high-capacity transport is required between data centers over regional or long-haul distances, this deployment model proves to be the most efficient, offering low power, compact footprint, and high spectral efficiency. Like the previous deployment model, it is suitable for legacy routers that do not support the form factor and power envelope of 400G coherent pluggables. This model is also compatible with open line systems and provides superior optical performance (capacity-reach) relative to any pluggable-based alternative. Infinera’s GX G30 Series and G42 compact modular platforms offer industry-leading optical engines with 600G and 800G per wavelength to meet any DCI capacity and reach requirements while lowering the cost of transport. Infinera’s ICE6-based solutions have repeatedly set new records in capacity-reach, including extending the economic life of many subsea cable systems:
- Hawaiki Achieves Record-breaking Performance on 15,000 km of Uncompensated Subsea Cable with Infinera ICE6
- Australia Japan Cable Trials Infinera’s Advanced Coherent Algorithms for Next-generation Optical Engines
- Unity Submarine Cable System Leverages Infinera’s ICE6 800G Solution to Double Capacity of the US and Japan Link
Similar to the previous deployment model, an additional 1RU or 2RU is added to the overall footprint.
Figure 3: DCI with compact modular platform and embedded optical engine
The advantages of each deployment solution are summarized in the following table:
DCI applications can be served by many different deployment scenarios, such as coherent pluggables hosted in routers or in compact modular platforms, or the high optical performance of embedded optical engines can be leveraged to extend capacity and reach. However, some deployment scenarios are more optimized around certain application requirements, as previously discussed in this blog post. Check Infinera’s Cloud Networking Applications website and blogs for more details and insights on how you can optimize your DCI/cloud networking deployment.