“Higher bandwidth and capacity requirements are driving more and more fiber optic deployments. Fifteen years ago, most optical fiber backbone networks in data centers had no more than 96 fiber cores and covered various redundant routes. Today, the number of fiber optic cores is generally 144, 288, and 864. Interconnect cables and cables used in ultra-large-scale and large-scale cloud data centers are migrating to 3,456 fiber bundles. Some optical fiber manufacturers now also provide 6,912-core optical cables, and 7,776-core optical fibers are also available.
“
Wu Jian, Technical Director, North Asia, CommScope
Higher bandwidth and capacity requirements are driving more and more fiber optic deployments. Fifteen years ago, most optical fiber backbone networks in data centers had no more than 96 fiber cores and covered various redundant routes. Today, the number of fiber optic cores is generally 144, 288, and 864. Interconnect cables and cables used in ultra-large-scale and large-scale cloud data centers are migrating to 3,456 fiber bundles. Some optical fiber manufacturers now also provide 6,912-core optical cables, and 7,776-core optical fibers are also available.
New optical cable structure design realizes density increase
Cables with a large number of fiber cores occupy valuable space in the pipeline. Due to the limited bending radius, larger cable diameters will bring performance challenges. In order to solve these problems, cable manufacturers are moving towards a crimpable ribbon structure and 200-micron optical fiber. The traditional ribbon fiber has a 12-core fiber bundle in the entire cable, while the crimpable ribbon fiber is that the parallel fibers are intermittently bonded together, so that they can be crimped and do not need to be laid flat. On average, this type of design can accommodate 3,456 fiber bundles in a two-inch pipe, and if a flat fiber structure is used in the same space, only 1,728 fiber bundles can be accommodated.
The 200-micron fiber retains the standard 125-micron cladding and is fully compatible with current and emerging optical devices. The difference is that the typical 250-micron coating is reduced to 200 microns. When used with a crimpable ribbon fiber, because the fiber diameter becomes smaller, the cable equipment manufacturer can maintain the same size of the cable, and the number of fibers can be doubled compared to the traditional 250-micron flat ribbon cable.
Hyperscale data centers have deployed technologies such as crimpable ribbon fiber and 200-micron fiber to meet the growing demand for connectivity between data centers. In the data center, the connection distance between the LEAF switch and the server is much shorter and the density is higher. The main consideration is the investment and operating cost of the optical module. Therefore, many data centers have been using low-cost vertical cavity surface emitting laser (VCSEL) transceivers based on multimode fiber. Others adopt a mashup approach, that is, single-mode is used in the upper SPINE mesh network layer, and the server is connected to the first-layer leaf (LEAF) switch through multi-mode. As more and more devices adopt 400GE, and 50G and 100G fiber connections with servers become the standard, network administrators will need to use these methods to weigh cost and performance.
80 km of DCI space: coherent light technology and direct detection technology
As the trend toward regional data center clusters continues, the demand for large-capacity and low-cost data center interconnect (DCI) links has become increasingly prominent. The new IEEE standard provides a variety of low-cost ways to provide plug-and-play point-to-point deployment. The transceiver used for direct detection is based on traditional PAM4 (four-level pulse amplitude modulation) and will be able to provide a link up to 40 km, while being directly compatible with the latest 400G data center switches. In addition, there are other developments aimed at similar functions of traditional DWDM transmission links.
As the link distance increases from 40km to 80km or more, the coherent optical system can provide stronger support for long-distance transmission and is expected to occupy most of the high-speed communication market. Coherent optics overcome limitations such as chromatic dispersion and polarization dispersion, making it an ideal technology choice for longer links. Traditionally, coherent optical devices are highly customized (and expensive), so a customized “modem” is required, which is the opposite of plug-and-play optical modules. With the advancement of technology, the size of coherent optical solutions is expected to be reduced, and deployment costs are expected to decrease. Ultimately, the relative cost difference may be reduced to the extent that shorter links can benefit from the development of the technology.
Overall control and continuous migration to high-speed
The data center’s progress towards higher speed needs to be carried out in an orderly manner. As applications and services evolve, the speed of storage and servers must also increase. Using a modular approach to handle repetitive and regular upgrades helps reduce the time and cost required to plan and implement changes. We recommend a holistic approach, in which switches, optical devices and fiber optic cabling should be used as a coordinated transmission path. Ultimately, how all these components work together will determine the network’s ability to provide reliable and effective support for new and future applications. The challenge today is 400G, and the future will be 800G and 1.6T. Although network technology is constantly changing, the basic requirements for high-quality fiber optic infrastructure will continue.
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.