Over the past 10 years, Ethernet speeds have undergone a transition from 100M to 10G and 10G networks. With the arrival of the expected peak of the 10G network around 2015, it is particularly urgent to avoid network bottlenecks. Currently, servers with 10G uplink ports have been widely used in data centers, and servers with 40G uplink ports are also eager to try. In order to cope with the excessive occupation of aggregation link and backhaul link resources caused by the rapid increase of server speed, IEEE 802.3ba HSSG has started to define MAC parameters, physical layer specifications and management parameters, and has successfully realized that the next generation Ethernet can be used in 40G and 100G. rate to transmit 802.3 format frames. Just as in the past, IEEE evolved to a higher rate to improve Ethernet, smoothness is the most important consideration. However, the evolution to 40G/100G ethernet is very different than it used to be, and this article will highlight these differences, focusing on the many elements that can change and impact data center design philosophy.
Table of Contents
Overall Design of Data Center
Designers always strive to achieve high reliability, high manageability, high flexibility, and scalability of data center networks. In addition, structured cabling standards set the general framework for achieving these goals. The ANSI/TIA 942 standard has long served as a structured cabling standard, providing guidance in designing telecom network infrastructure for data centers and providing added value for future network transitions to 40G/100G.
A New Interface for Optical Transceivers: MPO Connectors
For Ethernet, the continuous emergence of new technologies and the wide application of high-speed data networks are nothing new, but this evolution to 40G/100G is fundamentally different from previous rate increases. An important difference is the increased difficulty of VCSEL signal modulation, prompting the birth of multi-core fiber parallel optical transmission technology. In 40G technology, the single-core optical fiber in 10G is aggregated into 1 signal for transmission. In the same way, in 100G technology, a single-core fiber with “10 receiving and 10 sending” is used. Another problem caused by the difference is that it is completely different from the previous 10-fold rate progression, and it evolves to these two new transmission rates. How to specify. Although some users strongly demand to bypass 40G technology and directly evolve to 100G, HSSG has formulated 40G and 100G for the consideration of different speed requirements of current growing network applications in data centers, especially for network computing and core networking applications. rate standard. HSSG believes that 40G transmission rates are ideal for servers, high-performance computing clusters, blade servers, storage area networks and network storage, while 100G technology is beneficial for switching, routing and aggregation services in data centers. Another big change is the adoption of fiber optic connector technology that supports higher speeds. With the increase of network speed, all other forms of fiber optic connectors except duplex SC and LC have been largely eliminated. 40G/100G Ethernet provides higher speed. For this reason, MPO fiber optic connector technology is introduced, and 40G and 100G optical modules with MPO interface, such as 40G QSFP SR4, 100G QSFP28 SR4, will be equipped in large-scale optical fiber ports of active equipment , and rapidly integrated into various fields of the data center. Over the past decade or so, MPO has been widely used in data center physical layer deployments as an emerging technology with industry-recognized value-added modular solutions, but has not yet achieved scale on optical transceiver ports. TIA-604-5-C Fiber Optic Connectors Interoperability standard defines MPO, an MPO type connector that is an array connector capable of implementing up to 72 fibers on the ferrule end face of a single connector.
In fact, MPO connectors can support a variety of requirements in terms of fiber core count, and the 12-core MPO is just a generic version for current applications. The type of optical transceiver port to realize the transition from the currently dominant dual-core LC to MPO technology requires the supplier to solve the replacement of the corresponding component products of the physical layer, and also depends on how the network designer builds the 40G/100G data center. planned.
