Building a 10G 40G 100G Datacenter Switch

The process of developing telecom network infrastructure equipment might seem rather difficult, particularly owing to a lack of sufficient experience. “Making a non-blocking/wire-speed data center switch is still a very challenging task today. In addition to just assembling hardware components, there is also a need to consider control plane SW and tons of L2/L3 data plane protocols”, says Ivan Kuten, Promwad CTO and co-founder. Concerning this fact, he wants to highlight some core aspects in designing an Ethernet data center switch hardware.

Before developing 10G 40G 100G switches it makes the most sense to start with defining all the necessary elements. First of all, you’ll simply need an outer framework or chassis, which can be customized in any possible way, and power supplies to “spice up” the chassis. Second are fans to serve as an indicator of switch state. They also cool it down. Then comes several types of PCBA (Printed Circuit Board Assembly): fan control PCBA to connect fans, the heart of the switch — CPU PCBA, and switch main board PCBA, serving to host CPLDs, and the main interface cages. You definitely should have no trouble finding all the above-mentioned elements. Just in case you're not familiar with CPDLs, we should clarify this question. CPDL (Complex Programmable Logic Devices) is an interface that allows communication between SFPs, LED, power supplies, fans, sensors with NOS (Network Operating Systems) using widely used protocols like I2C or UART. The most commonly used CPDLs are Altera MAX chips.

So here are the elements that are included in all switches. Anyways, particular solutions have their own unique patterns and approaches. You may find them out in the next pert of the article. Also, we’ll touch upon the other side of a developing process — switching software. Keep track of our blog updates. Ivan will continue to develop this topic. Stay tuned!