Ethane

Context

• The original goal of SDN was to conduct isolation for security purposes.
• At the time, configuration for routers/switches include a lot of allowing, but lack the ability to define denying.
• Datacenters really appreciate the features brought with SDN due to following reasons:
  • Isolation for customers.
  • Performance optimization.
• SDN however, is not appreciate by everyone because:
  • It adds extra complexities.
  • There's a single point of failure (the control plane server).
• During the time when Martin Casado developed Ethane, the data plane already had pretty clean and efficient abstract/interface, but the control plane does not.

Goal

Implementation

• A simple FPGA used to implement the first customized switch because:

  • FPGA is programmable and efficient → they can customize their own hardware interfaces and mechanism such as forwarding the first packet of the flow to the control plane server
  • FPGA is cheap ← this research is not supported and accepted by the industries at the time.

• The first packet of each network flow is copied and sent to the control plane server. Meanwhile, all other traffic for the flow will be buffered in the switches. This turns out to be working just fine. While many others concern that a single server may not be able to handle such amount of traffic. Martin and his colleagues proved that a single cheap/simple control plane server can handle this task pretty well. This is because compared to flow traffic, a packet is very small in proportion.

Takeaway

• In the network research world, the academia and the industry are often disconneted (i.e. Academia doesn't know what's really needed in the industries).
• Sometimes, a solution might seem very reasonable, but there could be many reasons in reality to hinder the change/deployment of these improvements.
• Martin is rich now.

My Question