`

Feedback

Reader's Poll

Which of the following technologies/concepts are likely to witness significant traction this year?
 
Any data to show

Teledata

Tele Data

Mobile Subscribers Yearwise comparision

Network Migration: Overcoming connection reflectance and connectivity failures

August 23, 2017
E-mail Print PDF

By Nicholas Gagnon, Business Development Manager, EXFO

The inevitable obsolescence of copper infrastructure and the migration to optical fibre 10G and higher gigabit Ethernet environments will bring exciting opportunities, challenges and complexities for data centre network managers and contractors. One such network complexity that is often neglected is connection reflectance, which can lead to serious downtime and network failure that can also compromise the success of the migration. By using suitable testing equipment and changing the current testing practices to more future-oriented methods for accommodating the surge of new changes in connectivity such as parallel optics, pulse amplitude modulation 4 (PAM4), encircled flux multimode optical fibre (MMF) requirements, and OM5 wideband MMF with short wavelength division multiplexing, a serious network problem can be easily and quickly solved.

Connection reflectance at the patch panel

As transmission speeds increase, especially for 40G, 100G and 400G on the horizon, fibre insertion loss and connection reflectance requirements must become more stringent. For example, migrating from 10 GbE to 25 GbE provides 2.5x throughput per lane, while choosing a 25G line rate for a 100G (4x25) link may require a new transceiver. It is, therefore, debatable that the reflectance specifications published by some major industry standard bodies of -20 dB (ISO/IEC 11801 2010; TIA-568.3D 2013) for multimode fibre and -35 dB UPC (TIA-568.3-D 2013, ballot 6) for single mode fibre are inadequate. A -35 dB multimode and -40 dB to -45 dB reflectance for single mode may be a more realistic customer requirement for an existing patch panel equipped with lucent connector (LC)/ultra physical contact (UPC) (for use with QSFP28 4x25G, 100G). Reflectance thresholds should also consider the fact that the permanent link is changing for the high density and scalable data centre network, with its increased adoption in optimising parallel optics using only multi-fibre push-on (MPO) connectors that through patch cords are directly connected to the transmitter. This approach is ideal for the 40G-100G and future 400G migration and uses ribbon fibre for fast and easy installation, saving both time and labour.

MPO connectors for both multimode and single mode, however, may require tighter loss tolerances. Unlike subscriber connector and LC single-fibre connectors, MPO connectors can be more susceptible to developing poor physical contact after multiple connects and disconnects that can lead to degraded connector insertion loss and reflectance. For multimode, Elite grade or UPC and for single mode, angled physical contact (APC) polishes help to minimise reflectance issues.

Modifying testing procedures for reflectance and changes in the data centre

Data centre fibre infrastructure advancements have evolved rapidly in the past few years, and the fibre infrastructure and transceiver roadmap is becoming clear. Therefore, network operators and contractors are now in a position to plan their own roadmap for bandwidth growth. Through studies and research and development, EXFO engineers have become cognisant of macro trends that are dictating network infrastructure deployments and bandwidth migration for years to come, and have designed and launched testing equipment to reflect those trends and technology advancements. One such trend is that many hyper-scale and other large data centres are adopting single mode fibre only, while negotiating good price points for the cost of the more expensive single mode transceivers. Leveraging new modulation formats such as parallel single mode (PSM4) and PAM4, and parallel optics into their infrastructure, large data centres are deploying 100G (4x25G) today while simultaneously planning their future 200G and 400G deployments.

Since not all data centres are created equal, a second trend observed is that for smaller footprint enterprise data centres, multimode is still a good option, even if maximum fibre distances to meet the requirements of 100G/200G/400G over MMF are  constrained at maximum 60-70 metres to allow a safe migration to 200G/400G. The recent ratification by the Telecommunications Industry Association’s (TIA) TR-42 Engineering Committee of OM5 fibre allows WDM (4 wavelengths) in the 850 nm region and serves as yet another example that multimode fibre is a credible future-proofing medium. A successful plan to meet new technology requirements, however, like it or not, requires new testing practices and procedures.

A common testing practice in today’s data centres is to first use advanced optical loss test sets (OLTSs) to assess the important end-to-end loss budget of the fibre optic interconnect infrastructure and length of the fibre links to ensure that signals are successfully transmitted by the transceiver as per the specific applications. An encircled flux compliant OLTS such as the MAX 940/945 iCERT QUAD can provide bidirectional testing in two wavelengths in less than three seconds. How and when an OLTS is used for testing must take into account factors like substantial shrinkage of the loss budget (1.5 dB over MMF for 40G/100G), the precise assessment of where the insertion loss is distributed along the fibre link, and what portion goes to cassettes and connections versus the fibre itself. The reduction in loss budget and headroom becomes an even more complex issue when testing 10G deployments with 25G and/or future 50G lane deployments for 100G/200G/ 400G. Moreover, higher speed networks require additional testing including the reflectance of connections at the patch panel. Bypassing testing these systems, as some companies do, can lead to disastrous network failures that could have been easily avoided.

One of the primary sources of fibre link loss and multimode reflectance is fibre mismatch that creates connector offsets when using a vertical-cavity surface-emitting laser (VCSEL), which underfills the core with light and underestimates losses, or an LED that overfills and yields overestimated losses when testing. Studies indicate that there can be a variance of up to 60 per cent in power measurement due to the VCSEL’s susceptibility to underfill the fibre core, leading to inaccurate and random measurements. To achieve repeatability, consistency, reproducibility and validity of testing results, Encircled Flux metrics and methods should be employed (Figures 1, 2 and 3).

Poor physical connectivity can result in high bit error rates, potential packet loss and latency, which is why an encircled flux compliant light source such as the MAX 940/945 iCERT QUAD is advised.

Whenever OLTS results suggest a fault, there is a greater than 95 per cent chance that the problem lies with the connectors. The OLTS, however, cannot evaluate or measure insertion loss or reflectance and it cannot detect the source of the problem or the location of the faulty connector(s). That is the job of an encircled flux compliant optical time domain reflectometre (OTDR), such as the FTB-720C QUAD iOLM, that measures localised insertion loss and reflectance for each connector on the fibre link and precisely detects the position of the fault(s) for all input and output connectors to minimise network downtime. It also provides information about the total optical return loss (ORL), link loss and the length of the fibre link.

Since reflectance is such a serious issue, the logical recommendation is to test first with OTDR to ensure that all connectors are working efficiently, ensuring first-time-right migration to high speed links. It should be followed with OLTS testing to meet industry standards-related compliance requirements and ensure optimum end-to-end physical infrastructure transmission performance.

One of the most prevalent causes that affect return loss is contamination of the connector by transference of oil from the technician’s hand or fingertips during installation and testing. Oil contamination results in major changes to return loss (10dB to 12dB at 10G) in this single mode fibre example. The oil does not create an air gap at 10G so it does not impact the insertion loss, but at 25G, the return loss could be great enough to cause serious network failure due to significant degradation of the bit-error-rate test performance.

Since there is zero tolerance for any dirt, dust, or debris on connectors, fibre inspection and cleaning using an FIP-435B optical scope is a must each time a connector is manipulated and prior to any reconnection of connectors and OLTS or OTDR testing to keep the network up and running 24x7.

The future is now

With the rapid growth of bandwidth requirements and migration beyond 10G, the future is here and now. Old 10G connectivity and testing procedures and habits are no longer sufficient or economically justifiable. Continuing to run the data centre network with older small form-factor pluggables (SFPs), for example, is no longer viable as it would require too much fibre and active 10G ports. Moreover, the fact that a link was running smoothly at 10G is not a guarantee that 100G/200G/400G will run fine on the same existing architecture; it is a risk to assume that it would. It is not a guarantee either that new network build installations using the best quality components will not run into flaws, faults, losses and reflectance.

The best assurance for a successful migration and a low maintenance zero-downtime network is to test effectively and frequently as recommended, ultimately saving the organisation  significant time, labour costs and worry, since the network will be running at peak performance.

 
 
  • Most Viewed
  • Most Rated
  • Most Shared
  • Related Articles
 Your cart is empty

Monday morning

Monday morning