Netronome_Web_Logo_UPE9ULO.original.png

10 Myths about SDN, NFV and Data Center Switches: Debunked: Part Four

By Netronome | Dec 14, 2016

In this blog, I continue debunking myths about the role of data center switches in SDN and NFV deployments.

Myth #4: In the largest data centers, SDN and NFV and their related efficiency benefits are being realized through significant feature innovations in high-density networking switches.

The rise of east-west traffic in the data center and the adoption of leaf and spine network infrastructure topologies have revolutionized how data centers are built and efficiencies are realized. The demand for high port densities, high bandwidth and lower latency in leaf and spine switches is greater than ever. ODM white box switches based on merchant switch silicon are being increasingly deployed. More recently, innovations in the area of open switch abstraction APIs have helped the disaggregation of the control plane software from the underlying hardware, allowing for supply chains similar to servers.

Layer 2 and Layer 3 table scale and efficient packet buffer management to optimize cost and performance are hallmarks of a successful data center switch silicon design. Innovations in the switch silicon pipeline or datapath have been related to support new tunneling methods such as VxLAN and more sophisticated statistics gathering for analytics. These are significant feature innovations that have helped scale and efficiencies. However, it is worth noting that the sophistication of features in popular white box switches is far less than what vendors like Cisco supply to the enterprise market.

At the largest cloud data centers, the smarts of networking have moved to the server and most, if not all, innovation is occurring with server-based networking under the umbrellas of SDN and NFV. The ability to innovate and roll out new networking features is easier and most natural on servers. Coupling that with hardware acceleration using SmartNICs increases server efficiency by improving throughput and latency while freeing up CPU cores for revenue-generating VMs and applications. The recent announcements by James Hamilton of AWS at the Amazon re:Invent conference (https://www.youtube.com/watch?v=AyOAjFNPAbA) are an excellent testament. This trend has in fact further simplified the feature requirements in leaf and spine switches, fostering adoption of simpler ODM switch designs and control plane software, significantly reducing the cost of the switching infrastructure.

Myth #5: Networking data paths implemented in servers – as used in SDN and NFV - can be accelerated using DPDK, enabling high performance and server efficiency.

Stay tuned for part 5 in this series.

Read the Blog, "10 Myths about SDN, NFV and Data Center Switches: Debunked: Part 3" by Sujal Das.
Read the Blog, "10 Myths about SDN, NFV and Data Center Switches: Debunked: Part 2" by Sujal Das.
Read the Blog, "10 Myths about SDN, NFV and Data Center Switches: Debunked" by Sujal Das.