Because Gnodal-to-Gnodal switch processing is handled within the ASIC hardware, latency between any node in a leaf-spine mesh, which can scale out to as many as 4,896 ports, is less than 500 nanoseconds. Other features automatically implemented in the Gnodal control plane, which works across all switches in a Gnodal network, include dynamic load balancing and distributed, interswitch packet fairness.
In this interview, I ask Gnodal's CTO, Fred Homewood, how the GS0072 evolves the company's technology from last year's 10- and 10/40-Gbps edge products and new features beyond the speeds and feeds. We also talk about what differentiates Gnodal's switch from the plethora of 40-Gbps competition, and he explains the Gnodal fabric and the capabilities it provides when built into an internal two-tier leaf-spine network.
Like all Best of Interop entrants, Gnodal isn't just brochureware; it must bring actual hardware, so we take a look at the switch and its performance when aggregating traffic between a pair of Gnodal's edge devices. The numbers don't lie, and he demonstrates results from a Spirent traffic generator showing both three- and five-hop latencies through the GS0072 never exceeding 500 nanoseconds. At $2,500 per port and able to horizontally scale to thousands of edge ports, the GS0072 will certainly attract the interest of IT organizations building wide, fast, low-latency networks.
Watch the video interview: