Optical taps allow network and storage engineers to gather valuable data analytics. As the need to monitor data becomes more prevalent, performance considerations for incorporating taps in the data center cabling infrastructure should be taken into account. Both insertion loss as well as system parameters such as bit error rate should be considered when evaluating tap performance. While there is an inherent power penalty associated with using a tap, the structured cabling can be designed to support bandwidth needs through the use of system models generated by standards committees. These models reveal the trade-off between power penalties and supportable distance at a given data rate. Using these models, along with consideration for design and performance of system components, system designers are able to successfully deploy optical taps for network monitoring.
Optical taps are devices used to passively extract network data for analysis. Typically, the data is used to monitor for security threats, performance issues, and optimization of the network. An optical tap includes an optical splitter, which “splits” off a percentage of the input power and sends it to a monitoring device such as a probe or analyzer. Taps can be used in both the Ethernet network and storage area network (SAN); most commonly, split ratios of 50/50 are used in Ethernet systems and 70/30 split ratios are used in Fibre Channel SAN systems.
The use of taps introduces additional loss into a channel due to the nature of the technology, splitting either 30% or 50% of the power from the transmission. As the practice of inserting optical splitters (taps) into the cabling infrastructure for network monitoring has become a common practice, the question is how the power penalty impacts cabling channel performance and distance. The loss can range from 1.8 dB to 4.5 dB depending upon the split ratio and performance or type of splitter used. Figure 1 depicts a two patch panel cabling link connecting two devices. On one end of the channel a standalone optical tap has been introduced in the link to allow for monitoring of the network traffic for security and performance. If a 50/50 split ratio multimode splitter were placed in this channel, there would be an additional loss of 3.8-4.5 dB due to the optical split (depends on vendor specification), as well as the loss due to the LC matings at the input and output of the tap.
For a 10 GbE link, IEEE specifies a maximum connector loss of 1.5 dB (maximum channel loss of 2.6 dB) to support a maximum distance of 300 meters on OM3 fiber. With a two module link (assuming each module is a 0.5 dB specification) the overall link insertion loss would be 1.0 dB. However, when incorporating a tap as shown in Figure 1, the connector insertion loss increases by a total of 4.8 dB (4.5 dB for splitter plus 0.3 dB for the additional two LC connector pairs); adding that to the connector loss associated with the MTP-LC modules gives a total insertion loss of 5.8 dB. This does not mean that the channel will not support 10 GbE; it simply means the distance over which 10 GbE is supported is reduced from the original 300 meter maximum distance. Utilizing the IEEE system model to exchange distance for available dB loss, the above channel with an LC standalone tap module included, supports a distance of 59 meters over OM3 fiber and 73 meters over OM4 fiber. Hence the channel distance capability is reduced by nearly 80% as compared to the original non-tapped link.
To maximize distance capability, the connectivity design of the tap itself can factor into the loss introduced into a tapped channel. In a traditional tapped network today, a stand-alone LC-based tap is typically utilized in the design, as previously depicted in Figure 1. Alternatively, implementing an MTP integrated tap module in the cabling infrastructure reduces the number of connectivity components in the channel, thereby lowering the associated channel insertion loss. As shown in Figure 2, an MTP integrated tap module has MTP port connectivity for the input ports, allowing the use of it in place of the traditional MTP to LC module and stand-alone tap.
Using a high-performance, integrated MTP based tap reduces the loss budget for the channel as shown in the comparison in Table 1, where the number of LC matings in the channel is reduced by two. Low loss performance connectivity is assumed in both scenarios; in the “traditional” non-integrated tap design, a splitter loss of 4.5 dB is used based on typical solutions available in the market today. With the use of high performance splitter technology in the tap module, the loss impact of the tap is additionally reduced. In the case of a 50/50 MM splitter, the loss across the splitter is reduced to 3.8 dB. With a 1 dB reduction in total insertion loss, the distance capability increases from 73 meters to 240 meters, over OM4 fiber at 10GbE. This results in nearly three times the distance capability over the traditional non-integrated tap design.
Utilizing an integrated solution with high-performance components reduces the impact of the additional loss experienced from the introduction of a tap into a system, yielding a longer distance or reach.
Up to this point we have only considered the insertion loss impact of introducing a tap module into an optical link. However, there are other penalties which can affect the signal integrity such as jitter, differential modal delay, etc. These transmission performance penalties cannot be captured by a simple insertion loss, or power thru, measurement; instead, measurement of bit error rate (BER) is required.
Not all multi-mode splitters utilize the same technology, and this could have varying impact on the BER. TAPs available in the market typically use fused biconical taper (FBT) technology for both single-mode and multi-mode applications. However, alternate technologies such as thin-film splitters are available for high performance multi-mode applications. Characterization testing of these two technologies helps to define when transmission penalties are introduced in the system due to differential mode delay, depending upon the technology used in the splitter.
IEEE specifies a minimum receive power of -9.9 dBm in order to maintain operation at acceptable BER levels of 10-12. Characterization testing of systems with each of the above-mentioned splitter technologies reveals performance differences between the splitter types. To fully understand the implications of FBT vs thin film technology, two systems should be evaluated, varying the placement of the splitter relative to fiber length. Both of the system setups to be tested are depicted in Figure 3:
- TAP module with 300 meters of OM3 fiber on the transmit (Tx) before the split
- TAP module with 300 meters of OM3 fiber on the receive (Rx) after the split.
Prior to testing the above scenarios, a reference BER waterfall curve must be generated. To create a reference BER waterfall curve, the BER is measured over a short length of fiber with a variable optical attenuator (VoA) and tested over a range of receive power levels. When the BER vs. receive power is plotted, the BER waterfall curve is generated. After creating a reference waterfall curve at a very short length, a longer length of fiber (300 meters) is tested, and BER is again measured over a range of receive power levels until the BER reaches 10-12. Power penalties such as differential mode delay will cause this waterfall curve to shift to the right, with a BER of 10-12 occurring at a slightly higher receive power threshold.
Once both reference waterfall curves are generated, the above test setups are measured, with 300 meters of fiber placed on each side of the splitter, and on each splitter output leg (70% and 30%). The results of these measurements compared to the reference curves indicate any effects of the splitter performance on BER. The desired output is for the BER curves of the splitter outputs to coincide with the 300m reference curve, indicating no additional penalties (other than insertion loss) incurred when introducing a splitter in the link.
When generating the short fiber length reference curve for test equipment validation, the acceptable BER (10-12) threshold occurs at approximately -14 dBm, showing a 4 dB margin compared to the -9.9 dBm specification stated in IEEE 802.3 for 10 Gigabit Ethernet. As expected, when the reference fiber length is increased to 300 meters, the BER curve shifts to the right due to length dependent effects such as differential mode delay. The acceptable BER (10-12) threshold occurs at approximately -12.8 dBm, well within the minimum requirement of -9.9 dBm stated in IEEE 802.3 for 10 Gigabit Ethernet.
System setups 1 and 2 were characterized with measurements taken on both the 30% and 70% output legs, resulting in four total scenarios measured against the 300 meter reference waterfall curve:
- TAP module with 300 meters of OM3 fiber before the splitter, measured on the 70% output leg
- TAP module with 300 meters of OM3 fiber before the splitter, measured on the 30% output leg
- TAP module with 300 meters of OM3 fiber after the splitter, measured on the 70% output leg
- TAP module with 300 meters of OM3 fiber after the splitter, measured on the 30% output leg
The waterfall BER curves of these four scenarios, shown in Figure 4, overlay directly on the 300 meter reference system curve with no tap (or splitter). This indicates that the splitter does not introduce any BER penalties in the systems. As shown in the below graph, both output legs from the splitter (70% and 30% outputs) have the same performance with respect to BER, indicating there are not modal effects induced through the splitter. The performance is also shown to be the same whether the 300 meter length of the system is inserted before or after the split, eliminating possible concerns over BER effects due to placement of the splitter in the system.
For comparison, an FBT technology used in many of the MM taps available in the market was tested in each scenario as well. As shown in Figure 5, the BER waterfall curves vary for each of the systems, dependent on where the TAP is placed relative to the 300 meter system length (before or after the splitter) as well as variation between the output legs (70% vs. 30%) of the splitter. This disparity in BER seen between the output legs of the splitter indicates that there are variations in the power distribution across the splitter, resulting in additional penalties for each splitter output. As shown in Figure 5, on the 30% leg output the acceptable BER rate of 10-12 occurs at a receive power level of -11 dBm, which provides only 1 dB margin from the IEEE specification.
The characterization testing completed verifies that a TAP module using the high performance thin film technology does not introduce BER penalties. Although these additional penalties are not measurable in the field by a traditional light source and power meter attenuation measurement, they affect the transmission signal. Thus, it is important to consider the type of splitter technology being deployed in the network link in order to ensure there will be no negative impact on link performance.
There are many factors to take into account when designing data center cabling infrastructure. Optical taps are often used in order to obtain a copy of network traffic for the purpose of monitoring. Due to the insertion loss introduced by the taps, the overall link distance is reduced, sometimes by nearly 80% of the original non-tapped length. This length impact can be minimized by selecting high-performance tap modules which integrate the taps into the structured cabling. It is also important to consider the construction of the splitter to insure that a thin film technology is being utilized in multi-mode links to guarantee no additional BER penalties are introduced to the link. With these considerations in mind, system designers are able to successfully deploy optical taps when monitoring network links, with no negative link effects other than shortened distances.
Jennifer Cline is the enterprise networks data center global market development manager for Corning Optical Communications in Hickory, N.C. Jennifer started 15 years ago in technology as a graduate from North Carolina State in Mechanical Engineering. She has since held positions in systems engineering, enterprise marketing, and enterprise sales, including being a member of the Global Accounts Team. For more information, please email Jennifer.Cline@corning.com or visit www.corning.com/opcomm.
With over 10 years of experience at Corning Optical Communications, Brian Rhoney has held positions in product engineering, systems engineering, and product line management. He is currently the data center business development program manager for Enterprise Networks. In 2005, Brian received recognition as the Dr. Peter Bark Inventor of the Year for Corning Optical Communications with Pentagon Cable. He also received his professional engineer’s license in December 2005. Brian graduated from North Carolina State University with a Bachelor of Science in Mechanical Engineering and a Master of Science in Mechanical Engineering. He also received a Master of Business Administration from Lenoir-Rhyne University. For more information, please email Brian.Rhoney@corning.com or visit www.corning.com/opcomm.
Bigdata and data center