Johannesburg, 07 Apr 2017
Data centres have seen it all. Demand for bandwidth continues to grow at a phenomenal rate. Network traffic is changing and hybrid IT is the new normal. There are low bandwidth Internet of things (IOT) data streams and large individual streams that are capable of accepting latency - and all this data is travelling to and from the cables, pipes and servers at data centres, says Ed Gastle, Senior Product Manager, Enterprise, Viavi Solutions.
Traditionally, fibre has been considered superior over copper in cabling infrastructure, especially given the scale of changes just described. However, there are clear trade-offs between the two media, with fibre being more costly and complex to install, and copper being more limited in capacity and range. With recent developments in copper cabling standards, and as data centre topologies evolve, there should be room for both technologies. Regardless of the technology under discussion, proper certification of installations is vital. Following are considerations and best practices in these areas.
Copper: Considerations and standards update
Specifications for the next generation of copper cabling, called Category 8, were approved by TIA TR-42 Telecommunications Cabling Systems Engineering Committee in June. ISO/IEC has also circulated a final version for review and comment. These complementary standards address balanced twisted-pair cabling supporting 25G and 40G BASE-T interfaces, over a length up to 30 meters and frequency up to 2 GHz. With these attributes, copper becomes significantly more relevant in data centre applications, especially for rack-level interconnect or in topologies such as leaf-spine. It will no longer be necessary to make a blanket trade-off between copper and fibre, but rather use each where it is a better fit with respect to capacity, range, cost and ease of deployment.
With Cat 8 being so new, not all tools may be immediately standard-compliant, especially when it comes to testing up to 2 GHz, two to four times the frequency range of the previous standards. It is essential for data centre managers and installers to be careful in assessing inspection and test instruments to ensure compliance.
Fibre: Four factors in accurate certification
Fibre and associated test and certification standards have been around in their current form since 2011, so there's significantly more to discuss. The four key factors in installation and certification are: end-face inspection; encircled flux for multimode sources; use of test reference cords; setting and performing test references.
1. End-face inspection
When two fibres are mated together, there are three key requirements to ensure light passes from fibre to fibre without excessive loss or back reflections. Today's connector design and production techniques have eliminated most of the challenges to achieving core alignment and physical contact. What remains challenging is maintaining a pristine end face. A single particle mated into the core of a fibre can cause significant insertion loss, back reflection, and even equipment damage.
As a result, the condition of fibre end faces is likely the single-most controllable factor for the consistency of loss results as well as the ability for a system to perform as designed. This impacts all industries that use fibre optics, not just enterprise and data centre networks. In an effort to guarantee a common level of performance from optical connections, the IEC developed Standard 61300-3-35 that specifies pass/fail requirements for end-face quality before connection. While telecommunication service (wireline, wireless, cable), aerospace, and other industries have widely adopted this as standard practice, enterprise and data centres have yet to follow suit even though all current standards require it. The introduction and widespread rollout of higher-speed systems with tighter loss tolerances is changing this.
2. Encircled flux for multimode sources
Different multimode sources have different modal power distributions, also known as launch conditions. What this means is that different sources fill the large multimode core with different amounts of light. Some sources overfill the multimode core, while others tend to underfill the core. In a very simplistic sense, overfilling the fibre results in measured losses being too high and underfilling the fibre will result in measured losses being too low. The result is a variation of measured losses from tester to tester, assuming all else is equal. IEC 61280-4-14 defines a standard method to characterise the launch conditions of multimode test sources. Known as encircled flux (EF), this is measured as a ratio between the transmitted power at a given radius of the fibre core. An important aspect of encircled flux is that it is measured at the output of the launch cord rather than at the output of the source. TIA has created a Telecommunications Systems Bulletin (TSB- 4979) that describes two methods for implementing light sources to fulfil compliance requirements for the EF launch condition.
3. Test reference cords
The connection between the test cords and the system under test is a leading cause of uncertainty and variability of loss measurements. Using test reference cords (TRCs) greatly reduces this variability and increases the chances of consistent and repeatable loss measurements. TRCs are not just any fibre cord - they are built to different specifications with much tighter tolerances. In particular, they use high-performance connectors that have optimal geometrical and optical characteristics. The result is that when two reference-grade connectors are mated together there should be nearly no loss.
4. Setting and performing test references
People familiar with copper systems are used to the terms link and channel. These terms apply to fibre systems as well. A link is between two optical patch panels and may include connections and splices (such as an intermediate patch panel). Adding equipment cords at both ends creates a channel. During the construction phase of an enterprise or data centre network, the link is typically what is tested. Only rarely are channels tested. This is important to understand because the test reference method specified by the various standards is different depending on if a link or a channel is being tested.