SAMUEL LIU ROUTER INNOVATIONS
As network complexity increases, router technology can no longer rely on a one-size fits all approach. Samuel Liu of Nokia explains. Challenge Solving THE ROUTER
to 2km applications and SR4 for 100m applications using parallel MMF. Other QSFP28 modules include PSM4 for 500m using parallel SMF, CLR4 for 2km, SWDM4 for 100m using one pair of MMF. It is worth noting that most of the QSFP28 need clause 91 FEC on router host line cards. One challenge for the industry is how to support the QSFP28 standard ER4 module. CFP and CFP2 ER4 exist in the market. Carriers’ facilities will not change overnight when they upgrade routers to support higher density 100G. After a few trials, the industry is converging to develop an “ER4 Lite” design – since QSFP28 ER4 is not feasible with current technology. To address the ER4 distance, a FEC on router host board is a must. Another challenge is how to support an extended-temperature-range 100G optical module, especially for MBH and some A&A applications. Such a ETR module may have much higher than QSFP28 MSA defined 3.5W power consumption, which leaves the door open to continue supporting CFP4 on router. The industry is naturally moving to beyond 100G for next-generation router optical interfaces, targeting 400Gbits/s and even 1Tbits/s. The early adopters in the industry are more interested in reducing LAG and fiber management complexity than per 100G cost comparison with 100G interface on routers. There are two different approaches for 400G in the foreseeable future: CFP8 for short term and QSFP56-DD or other smaller form factor for midterm. The latter one possibly will offer not only higher density 400G interface on router but also a more competitive price compared to 100GBits/s normalized to Gbits/s. For even higher density with 400G or beyond 400G, on board optics design using silicon photonics becomes a potential option. On the other hand, higher 100G densities on router is emerging as a new requirement since more customers believe that 100G will become the new 10G. In other words, 100G will stay in the market for a long time. One approach is to drive OSFP, μQSFP or 100G SFP, which can push router 100G front panel density higher than QSFP28. It remains difficult to say how much cost reduction this approach will bring to the market initially since the industry is fragmented for the overall high density 100G market. Another approach is to use QSFP-DD / OSFP/ CFP8 to drive Nx100G break out through a MTP cable. This approach could double or even quadruple QSFP28 based 100G density for router. The challenge is that some of the new 100G modules are needed to support the other end due to the electrical and optical lane data rate difference. Either 2x50G or 1x100G optical modules are required to support 100G breakout solution.
SAMUEL LIU
W hen people talk about using optics on a router, they often just want to use similar optics as are developed for optical transport equipment. However, optical transport equipment development is typically focused on line side or WDM transmission technology development. In addition, more and more optical connections are now required between routers or switches only, with different requirements to those between routers or switches and transport nodes. Routers actually encompass a wide range of products, such as core routers, edge routers, access and aggregation routers, mobile gateway routers, and so on. And their optics requirements will vary depending on the different applications. The persistent trend of network bandwidth growth is set to continue for many years to come considering drivers such as 5G and IoT. Thus routers, as the fundamental building blocks for modern packet -ased networks, are in high demand, especially service routers at the edge of networks. For carriers, the challenge is how to control CapEx and OpEx while continuing to meet fast growing network bandwidth demand. For network equipment vendors, this means developing higher-capacity routers at lower costs and with better power consumption. ASIC DEVELOPMENT The 100G port cost for routers is mainly dictated by two key building blocks: packet-processing ASIC chipsets and optical modules. ASIC design has been consistently following Moore’s Law for
50 years. Silicon technology evolving from 40nm to 28nm to 16nm and now to even smaller feature sizes, enables routers to support higher 100G port density and higher than 100G per port capacity per ASIC with reasonable electrical power consumption. Due to front panel space limitation, power consumption and thermal management challenges, optical interface capacity on routers is outpaced by ASIC chipset capacity. Therefore to match ASIC capacity on router, the network equipment industry is evolving in two orthogonal directions: higher 100G port density and 400G per port capability, similar to what previously happened in the 10G to 40G/100G migration. It took the industry more than five years from initially offering a 100G optics interface on a router to widespread deployment. And it will take several years for the industry to move to 400G and beyond with major deployment levels. A technical challenge here is to control total power consumption and thermally manage very high-speed port density routers. Improving the watts/Gbits performance on router will help to reduce carriers’ operation cost. Optical module cost becomes more significant to the overall router solution cost. Router 100G interface started from 2010 with the initial CFP design. The industry is steadily increasing router density to match ASIC capacity by using CFP, CFP2, and then CFP4 formats. The new industry norm in 2016 is now QSFP28. There are multiple flavours of QSFP28 modules, which address different distances and applications with different price ranges. Typical QSFP28 modules include QSFP28 LR4 for up to 10km applications, CWDM4 for up
24
| ISSUE 7 | Q3 2016
www.opticalconnectionsnews.com
Made with FlippingBook - Online catalogs