Over the years, data center bandwidth requirements are expanding at double-digit rates, along with the equally urgent push not to compromise the cost-to-performance ratio. To accommodate the needs of web-scale data centers and cloud-based services, efforts have been made by leading cloud and telco providers to define and drive 25Gigabit Ethernet (GbE) technology. As such, the latest Ethernet speed upgrade path would be 10G-25G-100G, or the possible 10G-25G-50G-100G, instead of 10G-40G-100G. The occurrence of 25GbE changes the data center Ethernet landscape of some enterprises and organizations, creating a viable market for high-speed, reasonably-priced connectivity. How much do you know about this 25GbE network? You may have few insights on it. Don’t worry, this article gives an detailed description about it.

25GbE is defined for 100GbE which is implemented as four 25Gbps lanes running on four fiber or copper pairs. 100G optics (e.g. QSFP28 transceivers) have four lasers, each transmitting at 25Gbps. The twisted pair standard was derived from 40GbE standards development. Using 25GbE with QSFP28 transceivers results in a single-lane connection similar to existing 10GbE technology—but it delivers 2.5 times faster performance. It’s a proposed standard for Ethernet connectivity that will benefit cloud and enterprise data center environments, enabling the transmission of Ethernet frames at 25Gbps, and promoting the standardization and improvement of the interfaces for applicable products. The following table provides a summary of key 25GbE interfaces.

One of the main challenges in data centers is the insatiable hunger for more bandwidth. In 2010, the IEEE ratified a 40GbE and 100GbE standard, and launched a new study group to work on a 400GbE standard to keep up with bandwidth demand.

However, for some cloud providers and other large-scale data center operators, the requests are not more than the simple raw capacity. 10GbE is no longer fast enough. For server to top-of-rack network connections to keep up, you would need to double the number of switches in each rack and use 10GbE NICs. This would lead to impractical budget. The 40GbE isn't cost-effective or power-efficient in top-of-rack (ToR) switching for cloud providers and others that operate at a similar scale.

In such a condition, 25GbE was proposed as a standard for Ethernet connectivity, using a single-lane 25Gbps Ethernet link protocol to deliver the best price per performance ratio. In June 2014, the 25GbE Consortium was formed to promote the technology, and subsequently an IEEE 802 workgroup was formed to develop the standard.

Compared with 40GbE solutions, 25GbE technology provides superior switch port density by requiring just one lane (vs. four with 40GbE), along with lower costs and power requirements. Since 40G short-reach QSFP+ interface is constructed from four parallel links. Extending QSFP+ onto fiber requires four parallel 10Gb streams to transport this to the receiving QSFP+ parallel optics. And long-reach QSFP+ interfaces utilize Wave Division Multiplexing (WDM) to transport the four 10Gb streams over a single pair of fiber (image below). The requirement of four lanes significantly reduces switch port density per switching chip and increases the cost of cabling and optics. While the 25GbE standard requires only a single lane, while delivering 2.5 times more throughput compared to current 10GbE solutions, saving the cost compared to 40GbE solutions.

Besides, deploying 25GbE networks enables organizations to significantly reduce the required number of switches and cables, along with the considerations for the reduction of facility costs related to space, power, and cooling compared to 10GbE and 40GbE technology. Fewer physical network components reduce ongoing management and maintenance costs.

Additionally, the 25GbE physical interface specifications support the form factors, including QSFP28 and SFP28. QSFP28 has four lanes and each lanes supports 25Gbps speed. Each lane requires a serializer/deserializer (SerDes) chipset. The proposed 25GbE standard uses the same physical silicon from a single 25Gbit/s lane, which simplifies the process with just minor changes for forward error correction and lane alignment.

25GbE standard helps reduce capital expenditures (CAPEX) and operational expenditures (OPEX) compared to 40GbE, while meeting the necessary I/O bandwidth requirements in data centers. In addition, some blade server chassis solutions today are limited to only two SerDes lanes for their LAN on Motherboard (LOM) networking ports and therefore cannot implement a four-lane 40Gbps interface.

25GbE specification enables network bandwidth to be cost-effectively scaled in support of next-generation server and storage solutions residing in cloud and web-scale data center environments. At present, Fiberstore supplies many 25G solution products, including QSFP28 to 4xSFP28 DAC cables which serve as the alternative solutions to QSFP28 to QSFP28 DAC (QSFP28 cable). With the technology being much maturer, more and more 25G products will be made available to users.

Posted by: fernxu123 at 02:44 AM | No Comments | Add Comment
Post contains 780 words, total size 7 kb.