Optics Telecom

In recent years, 40 Gigabit Ethernet (GbE) has gained more popularity and the market of 40GbE is encouraging. But with the rapid growth of the new standard 100GbE, a new voice is announcing, namely 25GbE. As the increasing bandwidth requirements of private and public cloud data centers and communication service providers, 25GbE will to have a significant impact on server interconnect interfaces. Now you have two upgrade paths to 100G, 10G-25G-100G and 10G-40G-100G. Which one to choose? This post will make a comparison of 25GbE and 40GbE cabling, hoping it can help you make an appropriate decision.

25GbE is a standard developed by developed by IEEE 802.3 task forces P802.3by, used for Ethernet servers and switches connectivity in a datacenter environment. The single-lane design of 25 GbE gives it a low cost per bit, which enables cloud providers and large-scale data center operators to deploy fewer switches to meet the needs while still scaling their network infrastructure.

25GbE physical interface specification supports two main form factors, SFP28 (1x25 Gbps) and QSFP28 (4x25 Gbps). 25GBASE-SR SFP28 is an 850nm VCSEL 25GbE transceiver available in the market. It is designed to transmit and receive optical data over 50/125µm multi-mode optical fiber (MMF) and support up to 70m on OM3 MMF and 100m on OM4 MMF (LC duplex). In fact, using an SFP28 direct attach copper (DAC) cable for switches direct connection is a preferred option now. In addition, a more cost-effective solution is to use a QSFP28 to 4xSFP28 breakout cable to connect a 100GbE QSFP28 switch port with four SFP28 ports. DAC cable lengths are limited to three meters for 25GbE. Thus, active optic cable (AOC) solutions are also used for longer lengths of applications.

40GbE is a standard developed by the IEEE 802.3ba task force. The official development of 40GbE standards first began in January 2008, and were officially approved in June 2010. At the heart of the 40GbE network layer is a pair of transceivers connected by a fiber optic cable, OM4 or OM3 fiber cable. Fiber optic transceivers are plugged into either network servers or a variety of components, including interface cards and switches.

There are several standard form factors of 40GbE transceivers in the whole evolution. The CFP (C form-factor pluggable) transceiver uses 12 Tx and 12 Rx 10Gbps lanes to support one 100GbE port, or up to three 40GbE ports. With its large size, it can meet the needs of single-mode optics and can easily serve multi-mode optics or copper. But it is gradually falling behind since the increasing demand for high density. Another form factor is the CXP. It also provides twelve 10Gbps lanes in each direction, but is much smaller than the CFP and serves the needs of multi-mode optics and copper. At present, the most commonly used 40GbE form factor is the QSFP+ (quad small form-factor pluggable plus). It has the similar size with CXP but can provide four Tx and four Rx lanes to support 40GbE applications for single-mode, multi-mode fiber and copper.

Fiber optic cabling and copper cabling are both available for 40 GbE. The supportable channel length depends on the cable and the transceiver type. For data center 40GbE fiber optic cabling, OM3 and OM4 multi-mode cables are generally recommended because they can support a wider range of deployment configurations compared to copper solutions, and lower costs compared to single-mode solutions. MPO/MTP connectors are used at the multimode transceivers to support the multifiber parallel optics channels. For copper solutions, you can use QSFP+ direct attach copper cables, such as Cisco QSFP+ breakout cable. There are a lot of options, both active and passive, like Cisco QSFP-4SFP10G-CU5M compatible 40G QSFP+ to 4x10G SFP+ passive direct attach copper breakout cable (as shown below).

Compared to 40 GbE, 25GbE seems to be more suitable and cost-effective for cloud and web-scale data center applications. Using 25GbE with QSFP28 transceivers, users can deliver a single-lane connection, similar to the existing 10GbE technology but with 2.5X faster performance. In addition, 25GbE can provide superior switch port density by requiring just one lane (vs. 4 lanes with 40 GbE). Thus, it costs less and requires lower power consumption. Benefits of 25GbE compared to 40GbE are shown as below:

• Greater port density vs 40 GbE (one lane vs. four lanes)
• Maximum switch I/O performance and fabric capability
• Lower cost versus 40 GbE
• Reduced capital expenditures (CAPEX) and operational expenditures (OPEX)
• Fewer ToR switches and fewer cables
• Requires less power, cooling, and footprint
• Leverage of the existing IEEE 100GbE standard

25GbE seems to be a preferred option in the next step. It can provide up to 2.5 times faster performance than the existing 10GbE connections while maximizing the Ethernet controller bandwidth/pin and switch fabric capability. It can also provide greater port density with lower cost compared to 40GbE solutions. The trend will always be wider band and higher speed and port density. 25GbE or 40 GbE, let's wait and see how things play out.