During the past few years, there is a dramatic increase in the demand for bandwidth requirement for our communications. Whether be in a communication service provider or in a public or private data centers, a development in connectivity that can cater higher speed, bandwidth is needed. That is why last July 2014, an industry consortium was formed to create a new Ethernet connectivity standard in data centers. This standard was called 25 Gigabit Ethernet or 25Bade-T, developed by IEEE 802-3 task force P802.3by. This standard was derived from the 100Gbe, however, its operation works as a four 25Gbps that are running on four fibers or coppers. Last June 2016, this technology was commercially released using new interfaces called SFP28 and QSFP28. This article will discuss about the SFP28.
The SFP28 was constructed in a four parallel 25Gpbs data lanes allowing a maximum rate of 100Gbps. This physical structure of the SPF28 is the same with the popular SFP and SFP+. This characteristic provides flexibility due to the fact that the 100Gbps can also be divided individually in to four 25Gbps connections. SFP28 uses a 28Gbps lane (25Gbps + error correction) specifically used for top-or-rack (TOR) switch to server connectivity. Moreover, SFP28 is available in both copper and fiber optic cables.
The copper cable version is manufactured in a single fixed-configuration module which means the copper cables are directly attached to an SFP+ module. This version is ideal to be used for short distances ranging from 1m to 5m. On the other hand, the optical fiber version functions in either an 850nm that utilizes a pair of multimode fiber and works to a maximum distance of 100m or in a 1320nm that is made with a pair of single mode fibers works up to 20km.
The development of 25G SFP28 has provided a wide range benefits especially in a web-scale data center environment where the trend is to toward a single port server due to cost.
Primarily, it gives way to efficiently utilize data and switch port density. The reason for this is that, existing 100G port can be used as a 4x25G with as QSFP to SFP28 break out cable instead of using for different ports. For example, a 25Gbe strand can provide 2.5 times more data than the popular 10G solution and can provide greater port density.
Moreover, it provided an extremely efficient increase in speed to server to top-of-rack(TOR) especially when using the Direct Attached Copper assembly. It also simplifies development of interoperability specification and system due to the fact that its backward compatibility and gives an easier upgrade path from an existing 10G ToR server configuration.
Furthermore, using 25G SFP28 for ToR servers are more economical. This is because it can provide higher port densities, fewer ToR switches and cables are needed. It allows a more cost-effective alternative top-of-rack server connection that uses point-to-point patch cords. It enables End of Row(EoR) or Middle of Row (MoR) by using the 30-meter structured cabling. As a result, it reduces the capital expense in the construction cost compared to other configurations like the 40GbE.
Ultimately, the 25G SFP28 assemble features a reduced power and smaller footprint requirements for data centers because it limits the power per port to under 3W.
Due to this benefits that the 25G SFP28 assembly provides, it is forecasted that it will be popular in the years to come. It is believed that the dominant next generation server connection is toward the 25Gbps speed in server and in the near future, there will be more equipment that will use the 25G SFP28 cable assembly.