Author Archives: Serenity Huang

Why Is Plenum Cable Important to High Density Data Centers?


In most terrible building or house fire disasters, combustible plastics (PVC) used in the wiring are always among the very things that contribute to the rapid spread of fire and toxic smoke. And the air conditioning systems even help the toxic smoke given off by the burning cables to spread throughout the building quickly. As for the high density data center with high speed computing equipment and large amount of plastic jacketed cables, it is crucial to take measures to reduce the damage in an unwanted fire. How to make the crucial data center a safer place? The high quality plenum cable (eg. MTP fiber) can be one of the best choices.

break-down of precision air conditioning system might lead to fire incident

The break-down of precision air conditioning system might lead to fire incident.

Plenum Cable for Data Center Environment

Plenum space is an area used for return of air circulation or air conditioning systems. In a data center, the spaces covered by the precision air conditioning unit are often necessary to deploy plenum products. They include not only plenum containment that separates cold aisle and hot aisle, but also plenum cables that meet the highest fire code requirements. Both electric cables and fiber optic cables are required to be plenum rated (CMP) when they are installed in inner walls and inner ceilings of data center buildings.

Cold aisle containment (left) and hot aisle containment (right) in a precision air conditioning unit.

Cold aisle containment (left) and hot aisle containment (right) in a precision air conditioning unit.

MTP Plenum Fiber: Get Safety and High Bandwidth at the Same Time

The integration of plenum jacket and MTP fiber is a perfect solution for high density data center applications. The safety feature of the plenum cable and the high fiber port count of MTP connector endow the data center with two essential components. By using MTP plenum cable, the possible danger that might be caused by cables located at cold aisle and hot aisle can be minimized when a fire incident occurs. In addition, the high bandwidth demands within a limited space in data center can be satisfied.

safe and high speed mtp plenum cable

Use MTP plenum cable to get safety and high bandwidth at the same time.

When buying MTP fibers, be sure to check if it is genuine plenum rated and the MTP connector should meet the physical connection standard for acceptable insertion loss. When burnt, plenum cable will give off little smoke, and the color of the smoke is light instead of dark. FS MTP plenum cables are made of Corning fiber and U.S. Conec MTP connector. They are all tested and guaranteed by 3D interferometry and the insertion loss is no more than 0.35 dB. No matter it is the plenum jacket, the inner fiber, the connector, or the end face geometry, they are all genuine parts and in high quality. They can be checked by any user without a problem.

FS high quality MTP plenum cable made of U.S. Conec connector and Corning fiber.

FS high quality MTP plenum cable made of U.S. Conec connector and Corning fiber.

Cost Comparison of 10G Leaf Switch and 40G Spine Switch

The growth of network users and the virtualized and automated trend in network have called for a great change in the macro network infrastructure. The “old” three-tier network architecture, namely core, aggregation and access, is quickly proved to be inefficient in modern complicated network environment. So what’s next to replace the outdated frame? After my research, the best I currently find is the flatter leaf-spine network architecture, which surpasses the traditional one with improved switch capacity and much lower latency.

three-tier network is outdated

In order to help buying a suitable switch when scaling the efficient leaf-spine architecture, I will compare some different brands’ 10G leaf switch (48*10G SFP+ and 6*40G QSFP+) and 40G spine switch (32*40G QSFP+) in this post. It should be noted that each switch price shown below is based on my own research, and I cannot promise that you can get one at the same price, since there are cases of discount, charged shipping, or tax. But you can take it as a reference. And welcome to add new information or to correct mistakes.

10G Leaf Switch (48*10G SFP+ and 6*40G QSFP+), switch capacity being 1.44Tbps.

Brand Model CPU ASIC NOS Online Base Price (US$)
Edge-Core AS5812-54X X86-64 Broadcom Trident2 Cumulus Linux $4,438.07 ~ $4,889.75
Agema AG7648 X86-64 Broadcom Trident2 Agema OS & OcNOSTM $5,995.00
Cisco Nexus 9372PX X86-64 Broadcom Trident2 Cisco Internetwork Operating System (IOS) $9,505.00 ~ $21,318.16
Dell S4048-ON X86-64 Broadcom Trident2 Cumulus Linux $2,250.00 ~ $2,850.00
Arista 7050SX-72Q X86-64 Broadcom Trident2 Arista Extensible Operating System (EOS) $21,408.95
Juniper QFX5100-48S X86-64 Broadcom Trident2 Juniper Operating System (JunOS) $24,299.00 ~ $25,942.00
Brocade VDX 6740 X86-64 Broadcom Trident2 Brocade Fabric OS $16,815.82 ~ $19,369.99
HPE Altoline 6920 X86-64 Broadcom Trident2 HPE Comware $11,209.66 ~ $12,792.00
Huawei CE6851 X86-64 Broadcom Trident2 Huawei Operating System (OS) $6,379.95 ~ $11,238.08

40G spine switch (32*40G QSFP+), switch capacity being 2.56 Tbps.

Brand Model CPU ASIC NOS Online Base Price (US$)
Edge-Core AS6712-32X X86-64 Broadcom Trident2 Cumulus Linux $7,571.95 ~ $14,124.16
Agema AG8032 X86-64 Broadcom Trident2 Agema OS & OcNOSTM $8,495.00
Cisco Nexus 9332PQ X86-64 Broadcom Trident2 Cisco IOS $17,617.21 ~ $18,673.15
Dell S6000-ON X86-64 Broadcom Trident2 Cumulus Linux $ 29,476.80
Arista 7050QX-32S X86-64 Broadcom Trident2 Arista EOS $24,208.84 ~ $27,988.95
Juniper QFX5100-24Q X86-64 Broadcom Trident2 JunOS $29,313.83 ~ $32,949.99
Brocade VDX 6940 X86-64 Broadcom Trident2 Brocade Fabric OS $21,546.46 ~ $24,036.81
HPE Altoline 6940 X86-64 Broadcom Trident2 HPE Comware $15,354.38 ~ $15,739.68
Huawei CE7850 X86-64 Broadcom Trident2 Huawei OS $13,737.60 ~ $23,000.00

These open networking leaf and spine switches are almost all adopting Broadcom Trident 2 chip and the Intel 64 processor, the major difference between them lies in the software. Some big brand switches deploy their own network operating system while some support licensed Cumulus Linux OS.

FS provides not only high performance 10G leaf switch (S9000-48S6Q, 48*10G SFP+ and 6*40G QSFP+) and 40G spine switch (S9000-32Q, 32*40G QSFP+) that have excellent Broadcom Trident 2 switching chip and licensed Cumulus Linux OS, and customers also enjoy the easiest addressing of both hardware and software problems. When you buy open networking switch at FS, you can also get certified optics and cables, and professional software support effortlessly and at low cost.

FS S9000 series 10G leaf and 40G spine switch

The base prices listed above do not include accessories like optics and fiber optic cables, or the software and hardware support cost. And only a few vendors I know can provide certified optics and cables, and expert software and hardware support at the same time. One-stop experience could only be found at vendors like Cumulus Networks (Cumulus Express service) and FS (Fiberstore). As you know that building a high performance networking is never only about switch, it would be better if simple and all-in-one approach could be provided in switch installation.

100G QSFP28 PAM4 or Coherent CFP?

The ever-increasing need for higher data rate in mobile data traffic, data centers and cloud services has pushed the access streams from 2.5Gb/s to 100Gb/s, and is demanding for 100Gb/s beyond without a stop. In today’s core network that has deployed 100G rates, there are QSFP28 optical transceivers including SR4, PSM4, CWDM4, LR4, ER4, etc., serving for a maximum 25km transmission distance. And there are 100G AOC, DAC and breakout cables generally for applications of tens of meters. 100G CFP/CFP2 modules including SR10, LR4 and ER4 support transmission distances of 150m to 40km. Until recent years, the telecom service providers are adopting new 100G DWDM technologies in their high capacity and long distance backbone applications. Coherent 100G DWDM transceivers are the first to be deployed for 100G long-haul applications, and then new technologies like PAM4 (Pulse Amplitude Modulation) are developed to meet lengths requirements for 100G metro network. This post is to discuss the issues on coherent and PAM4 100G DWDM transceivers.

Overview on 100G DWDM Transceivers

In the past few years, the adoption of 100G DWDM technologies is mainly focused on coherent DWDM optical transceivers, including CFP and CFP2. Until the year 2016, Inphi (a specialist in this area) offers pluggable 100G PAM4 QSFP28 DWDM transceivers to support 80km data center interconnect (DCI). The alternative for 100G DWDM coherent transceiver is given much attention. Besides, this new option for 100G DWDM transceiver also arouses hot discussion on which to choose. Knowing the characteristics and suited applications of them could help in selection.

QSFP28 PAM4 and Coherent CFP/CFP2

There are significant differences between QSFP28 PAM4 transceivers and coherent CFP/CFP2 transceivers, but they also have some relations in 100G applications. Contents below will go to details of these optical modules.


Before the announcement of PAM4, binary NRZ (non-return to zero) modulation format is used for 40G and 100G long-haul transmission systems. PAM4 has four distinct levels to encode two bits of data, essentially doubling the bandwidth of a connection. Currently the single-wavelength PAM4 modulation scheme is considered the most cost-effective, efficient enabler of 100G and beyond in the data center. The 100G DWDM transceiver utilizing PAM4 signaling is in QSFP28 form factor. The advantage is that the customers who want to build an embedded DWDM network can use this transceiver directly in the switch. On this side, it is simple and cost-effective solution. But there are some prerequisites: it needs amplification to get out of the blocks and dispersion compensation to go beyond 5-6km. Therefore, a separate DWDM multiplexer with an amplification system and dispersion compensation is required to connect data canters together.

single wavelength PAM4 100G

In another case, if the QSFP28 PAM4 module is added to an existing DWDM network, it must be a network already having right dispersion compensation modules (DCMs) and amplification system in place; if it is not, changes are required when QSFP28 PAM4 is later added.

Coherent CFP

CFP digital coherent optics (DCO) have a high speed digital signal processing (DSP) chip built in. They do not require separate DCMs. This is what makes CFP different from QSFP28. Instead, they have electronic dispersion compensation built in. Although the built-in DSP requires more power and adds cost in components, it releases the switch vendors from adding DSPs to their equipment. Coherent CFPs enables transmission distance of more than 1000km between sites.

CFP2 analog coherent optic (ACO) is half the width of the CFP. Existing CFP2 coherent DWDM optical transceivers are analog and require a separate DSP on the host board to take the full advantages of the coherent features. So it is suited for switch vendors who have fitted such a DSP, but it adds additional cost and power consumption on the main board.

CFP2 digital coherent optic (DCO), expected to be released in the coming two years, is more optimized than CFP2 ACO in that it has built-in DSP. This component will open up to all switch vendors using CFP2 without DSP. With different coherent CFPs optional, customers can pay only for what they need when they need it.

QSFP28 PAM4 or Coherent CFP?

This really should depend on the applications. According to ACG research (an analyst and consulting firm that focuses on the service providers’ networking and the telecom industry), the 100G PAM4 solution and coherent DWDM solution, together with IEEE802.3ba, cover different portions of the optical fiber reach in the data center interconnect. So when deploying a long distance 100G DWDM network with DWDM transceivers, the required transmission distance and available equipment should be taken into consideration when choosing a suitable pluggable module.

IEEE, PAM4, OIF coherent optical reaches


Using pluggable transceivers for embedded DWDM, where the DWDM functionality is in the transceiver and not a separate DWDM converter platform, offers the ultimate solution in terms of cost and simplicity. Both QSFP28 PAM4 and coherent CFP/CFP2 are all suited to this approach. They can be used for embedded DWDM networking or as part of an existing DWDM installation. They all enable the advantages of pluggable modules: simple installation, easy spares handling, lower cost of ownership and quick return on invest.

Simplify the Implementation of High Density 100G CXP

Data center bandwidth demands continue to grow, requiring higher capacity and throughput. The 100G/120G Ethernet is no longer new in data center optic market, but it’s still a complex act to efficiently and effectively upgrade existing 10G/40G architectures to these higher data rates, especially in a space-constrained application. In order to explore the approaches of smooth migration to high density 100G/120G network, this post will take multimode 100G CXP module as example, and illustrate some simplified scenarios when upgrading to these higher data rates.

Overview on 100G CXP Module

High density 100G CXP is very popular in the implementations up to 100Gbps for saving-space. This deployment can then leverage the 10G-per-lane channels to distribute the 10G data anywhere in the data center. 100G CXP module is designed to connect with an MTP/MPO-24 connector, which can be divided to 10x10G or 12x10G transceiver pairs. For 120G CXP, it is also possible to separate the signals into three QSFP+ transceivers, and then to three groups of 4x10G transceivers by using an 8 fibers MTP/MPO to LC breakout module or cable.

Direct Connectivity for two CXPs

For two 100GBASE-SR10 modules, direct link can be easily made via 100G MPO cable. For connecting two 120G CXPs, a cost-effective 24 fibers MPO trunk can also work well. Here uses an 24 fibers MPO (female) to MPO (female) OM4 polarity B trunk cable.

direct link for two 100G/120G CXP modules

Figure 1: direct link for two 100G CXP modules.

Connectivity Methods for CXP and SFP+/QSFP+

In this part, the scenarios applied for 100G to 10G connection, and 120G to 40G or 10G connection will be explained.

100G to 10G

Figure 2 shows a direct link for one 100G CFP module and ten 10G SFP+ modules. By using the 24 fibers MPO to LC duplex harness cable, the whole 100G from the CFP transceiver is connected to ten SFP+ transceivers (two LC duplex legs are not used in this link). The fanout legs are available to be the same length or staggered type, so as to meet different applications.

direct link for a 100G CFP to 10x10G SFP+s

Figure 2: direct link for a 100G CFP to 10x10G SFP+s.

In figure 3, the interconnect for CFP and SFP+ transceivers is more flexible than the direct link. Here the 160 fibers MTP/MPO (male) breakout patch panel allows connectivity to any duplex path reachable by the patch panel. This method offers ultimate flexibility in allowing connectivity to any row, rack or shelf. Moreover, this breakout module can support up to eight groups of this 100G to 10x10G transmission. In such a high density link, it is suggested to use HD patch cables or LC uniboot patch cables to enable quicker and better cable management.

interconnect solution for 100G CFP to 10x10G SFP+s

Figure 3: interconnect solution for 100G CFP to 10x10G SFP+s.

120G to 10G and 40G

When directly connecting one 120G CXP to twelve 10G SFP+ transceivers, a 24 fibers MTP-24 to 12 LC harness cables can do the job well. Here we use a customized high density bend insensitive female MTP-24 to 12 LC duplex OM4 breakout cable.

 direct link for 120G to 12x10G transceivers

Figure 4: direct link for 120G to 12x10G transceivers.

An option for breaking out a 120G CXP to three 40G QSFP+s is to use the 1×3 MTP/MPO conversion harness cable. Figure 5 illustrates implementation of a 1×24 strand MTP to 3×8 strand MTP conversion harness cable. Like the 12x10G segregation mentioned above, once split, the 3×8-fiber QSFP+ channels can be distributed through patch panels and 12-fiber based trunking to any area of the data center.

hybrid link for 120G CXP to 40G QSFP+s and 10G SFP+s

Figure 5: hybrid link for 120G CXP to 40G QSFP+s and 10G SFP+s.


This article has illustrated some simplified implementation examples of 100G CXP modules. 24 fibers MTP/MPO trunk cable are suited for connecting two CXP modules. Breakout cables can achieve quick connection for CXP and QSFP+ or SFP+ optics, but when flexible patching is needed in the link, it would be better to adopt breakout patch panel. If you need 100G optics, FS.COM can offer you fully tested compatible 100GBASE-SR10, 100GBASE-SR4, 100GBASE-LR4 and 100GBASE-ER4 transceivers, etc.