Managing the Rising Cost of Optics in Networks









































































































































































































In this issue of Voices of the industry, Bob Shine, vice president of product marketing and management at Telescent, explores how the rising cost of network optics can be managed through the dynamic allocation of fiber layer resources.

automation fibers

Bob Shine, Vice President of Product Marketing and Management, Telescent

As optical networks have evolved to higher data rates to cope with astronomical increases in traffic, the percentage of network costs associated with optical components has increased significantly. Finding ways to deal economically with this rising percentage of optics costs has become the focus of many network operators and providers. The dynamic allocation of fiber layer resources is a way to improve the economy of the optical network and can be implemented today using large-scale robotic fiber cross-links.

The growing share of costs for optical components in networks is due to the need for networks to maintain higher port densities and faster speeds. This necessitates the use of higher-speed optics, as seen from the transition from 40G to 100G and now to 400G. Moore’s Law reduced the cost / bit for switches and routers with advanced ASIC development for silicon. However, while the cost / bits for interconnect optics have also decreased, they have not decreased almost as fast as the respective silicon components.

Based on the above economic trends, Cisco said that for 10G networks, optics represent about 10% of the total hardware cost of a data center network. In contrast, optics will account for over 50% of the total hardware cost of a 400G network. [Optics in the Data Center: Powering Ever-Increasing Capacity Demands – Cisco Blogs]

As you can imagine, this cost challenge occurs in a wide variety of areas – mainly in any situation where there are computing and storage components combined with high-speed transmission requirements. These include long-distance networks and now 5G networks, data center networks and even machine learning resources in data centers. Spine optical networks are often the first areas of the network to migrate to higher data rates due to the huge amount of traffic transmitted through limited and expensive long-distance fibers. The compaction of 5G networks and the requirements for the radio towers in front increase the cost of the 5G optical network. Within data centers, scaling with growth while minimizing capital costs requires effective ways to add capacity that can be managed through dynamic fiber management. Even applications such as machine learning, which is very computationally intensive, will see a growing share of the cost of optical components due to the need to transfer ever-increasing sets of parameters between learning iterations. The following sections will demonstrate how all of these cases can benefit from the dynamic resource allocation of the optical layer.

The advantage of dynamic Fiber Cross Connects for long distance networks

For core optical networks, the AT&T Group has demonstrated significant cost savings and improved stability through the use of software-defined network (SDN) control and dynamic fiber cross-links (DFCC) to provide joint IP and network optimization. optical layer. In essence, SDN and DFCC allow the disaggregation of a failed connection, which allows the running components of the failed connection to be reused in a redesigned route. For a central office with two large core routers, instead of connecting both the router port and the transponder offline, if one component fails or needs to restart software, dynamic cross-connect allows the work component to be recombined with a port in the second router. Even without failure, DFCC can be used to recombine router / transponder pairs to create more efficient routing. The addition of only automated fiber cross-linking in this case of use creates an estimated saving of 5% in the entire network stack.

The advantage of Dynamic Fiber Cross Connects while scaling the capacity of the data center

Moving to the data center, while the construction of the data center must be completed before commissioning, the data centers are usually slowly filled with servers and switches. This allows capacity to meet growing demand while avoiding empty capacity. The initial moderate-scale network will need to be expanded while the network carries live traffic. While other network topologies have been explored to allow fine-grained expansion, it is possible to extend using a traditional Clos network using a new topology-driven methodology, as shown by the work done in Google. Live extension requires maintaining sufficient network bandwidth throughout the live extension to avoid congestion. This is done through multiple automated steps, in which a limited subset of network elements are interrupted and added. As capacity is removed during these retrofits, each stage must be completed as soon as possible. To simplify reconfiguration, all server blocks and all spinal blocks are connected through a group of panel panels – creating a DCN topology that can be created and modified simply by moving fiber jumpers to the adjustment panel. The management of this reconnection during upgrades can be significantly simplified and controlled by a robotic fiber cross-linking system.

The advantage of Dynamic Fiber Cross Connects for machine learning resources

Finally, inside data centers, machine learning is becoming an increasing share of computing power. While the use case is dominated by the development of advanced GPU and TPU processing capabilities, the very large and growing training kits used are creating a high bandwidth demand for communication between GPUs. NVIDIA’s fastest ML learning platforms offer 1.2 Tbps of communication bandwidth between their GPUs. But while conventional data center workloads have unpredictable behavior with short flows dominating the workload, machine learning workloads are predictable, scarce, and consist mostly of large transfers. By taking advantage of these features, dynamic cross-linking, which regulates the communication bandwidth between GPUs based on the training model, can significantly improve performance.

An example of dynamic fiber crossing to deal with the use cases above would be Network Topology Manager (NTM). When selecting NTM, look for one that uses a robot to remotely configure and reconfigure cross-connections in minutes. Ideally, you’ll also want to choose a pay-as-you-go NTM model that can scale the number of ports while maintaining connectivity from anything to anything. You will also want a system with low lost and fixed connections and one that can be fully maintained on site without interrupting traffic. Finally, your NTM must be NEBS 3 certified to ensure the reliability that customers require.

Data traffic will certainly continue to grow, leading to the need for increased bandwidth at all levels of the data network. While cost control will require progress in a number of areas, adding dynamic fiber management to fiber is one way to more effectively address the growing share of optical costs in the network.

This article was written by Bob Shine, vice president of marketing and product management, Telescent. Telescent’s Network Topology Manager (NTM) uses a robot to remotely configure and reconfigure cross-connections in minutes. It provides dynamic fiber layer control, is NEBS certified and is available to meet your needs today. To learn more, contact us Telescent today.