Network Computing is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Increasing Port Density in Data Centers

network-2402637_1280.jpg

Data center
(Source: Pixabay)

Demands for connectivity in the data center are rising, especially in hyperscale data centers where 1728- or 3456-fiber cables are becoming more popular. Connecting such high-fiber-count cables to servers and switches is the key challenge because there's only so much rack space available. Fiber patch panels are at the center of this challenge. To address this issue, the industry is increasing port density in patch panels to accommodate the ongoing thirst for bandwidth.

Maximizing Rack Space

One challenge facing the data center architect is to maximize usage of precious rack space. Each rack typically has 42 or 48RU, and the main goal is to use as many of those units as possible for computing. But compute servers must be connected, so some of the rack volume must be used for patching. Clearly, the less volume used for passive components, the more available for active hardware.  The volume taken up in cooled racks is probably the most expensive in the data center; architects want to minimize usage of this valuable floor space for patching, and increasing port density is the solution.

At the same time, data center architects have varying applications and optical interfaces when it comes to connectivity. Some use serial connectivity, while others have moved to parallel connectivity to increase bandwidth. Ideally, the passive infrastructure and horizontal cabling should allow as much flexibility and modularity as possible to avoid having to purchase different types of patch solutions for different needs. Connector formats are also evolving: simplex SCs and duplex LCs have been popular for years;  alternate reduced form-factor serial connector formats (SN by Senko, MDC by USConec) are engaged in the latest “connector wars”;  and parallel connectivity has made a strong push within the last ten years.

But connector performance is evolving. Everyone wants the most flexibility in choosing the type of connector – whether it’s duplex connectors with two fibers per port or parallel connectors with a much higher range of connections per port. USConec has proven 72 fibers per port with MTPs, and 24 fibers per MPO/MTP port is common. Port preferences can migrate from serial to parallel and ultimately back to serial.  Without a flexible platform, future-proofing and MACs considerations are handicapped, and the lifespan of the investment is limited.

The Patch Field – Optimizing Both Sides

The patch field is where you make the connection between compute hardware and high-fiber count cabling. At Day 1 deployment, you need access to the back-side or behind-the-wall (BTW) cabling, and to the latches of the ports on the patch side.  You need good rear cable management, conceived at platform design, and deployed at installation, and you need access to the port latches. This need for access to both the front and rear of the panel impacts flexibility in some fixed panel designs.

Some vendors offer flat panels with adapters, potentially squeezing in a few more ports per panel, but they have ignored cable management concerns, and there’s no flexibility or modularity. Movable panels, including half-rack wide solutions, provide better port-latch access but tend to fall short in terms of modularity and flexibility of the platform.

Another option to flat panels is pre-terminated panels, so there’s no connection to the back of the panel. It’s not a flexible solution: vendors have minimized back-side cable management, but they haven’t improved the patching side of cable management or allowed for flexibility within the platform.

Most of these solutions provide limited modularity, so you’re committed to port type when purchasing the panel. Also, when you fill a fixed panel with ports at maximum density, you have no panel space for port identification.  With fixed panels, to get the most density, panel designs may deviate from 1RU panel-increments; you have to go to 2RU, 3RU, and even higher commitments of rack-space Day 1 to achieve highest port density.

Legacy modular patching platform designs have some manufacturers limited to 72 ports per RU. Some vendors have non-interchangeable platforms that aren’t flexible. Other solutions sacrifice port density to maintain interfaces with automated infrastructure management (AIM) solutions, and that backwards legacy has made it impossible for many vendors to increase density. Moreover, many vendors don’t want to introduce a new platform that is incompatible with previous platforms

Flexibility and Modularity are Key

The value of flexibility and modularity is that the platform can be reconfigured to suit a different need to deliver a different application. If the platform is modular at the connector level, you can accommodate serial and parallel ports in the same cassette if desired. Additionally, connector modularity enables port designation by using colored adapters, keyed adapters, shuttered and non-shuttered adapters, and more.

Modularity at the platform level also allows customers to use the same cassettes in multiple applications while maintaining port density. If you’re pulling a high fiber count cable, pre-termination of both ends may not be practical due to limited pathways such as small conduit.   Platform modularity should extend to the cassettes, so the same cassette with different internal components can accommodate field-splicing, either internal or external to the cassette.

Today, MPO is the parallel optical connector of choice in the data center, predominantly with a 12-fiber count. More recently, eight active fibers in a 12-fiber ferrule has found favor based on breakout applications.  At a minimum, the platform should support those options. An ideal solution will support 8- and 12-fiber connectivity, with additional capability to support newer parallel connectivity variants, including 16-fiber one row, as well as two rows in the same MPO ferrules, i.e., 24 and 32-fiber connectivity.  Ideally, a patching platform will have the ability to mix cassettes in the same housings, with no loss of port density. Most solutions support two or possibly three of these variants.   

There’s definitely a trend toward more use of parallel connectivity – it’s typically the first step toward bandwidth increases.  As transceiver processing speeds are increased, serial applications are typically first to market. Bandwidth increases at the same processing speeds is accomplished by adding more lanes (fibers), hence the evolution to parallel applications.  As transceiver efficiencies and speeds are increased, serial solutions tend to replace early parallel solutions. It's cyclical. Most data centers start with serial applications, increase the bandwidth with parallel, and then several years down the road go back to a serial solution at the same speed because of advancements in the efficiency of the transceiver. Patching platforms should be flexible in their ability to easily migrate from serial to parallel and back to serial as applications evolve.

An ideal connectivity solution should accommodate these trends, supporting serial and parallel connectivity with ease, and adapting to different types of connectors and cable management requirements. By designing patch panels with high modularity and flexibility, vendors can support any direction data center architectures happen to take.