White box OS startup Arrcus today launched 100 gigabit and 400G Ethernet switches based on its network operating system. These switches target hyperscalers, service providers, and enterprises that are running data-intensive applications.

Last July shortly following a $15 million Series A round, the San Jose, California-based startup launched an independent hardware-agnostic operating system. Its ArcOS operating system comes with standards-based APIS and an intelligent data plane adaption layer (DPAL) that allows it to work closely with the underlying silicon.

ArcOS also uses standards-based OpenConfig/YANG programmable APIs, a hyper-performance, resilient control plane, and a data-model driven telemetry. It is supported on Broadcom’s StrataDNX Jerico+, StrataXGS Trident 3 and Tomahawk 3 switch silicon platforms.

The new 100G and 400G switches are based on ArcOS and developed with four of Arrcus’ ODM partners: QCT, Delta Networks, Celestica, and Accton Technology Corporation.

According to a blog post written by Arrcus founder and CEO Devesh Garg, the new white box switches provide four times the bandwidth “at a fraction of the power consumption of existing switches.” He noted that this increased capacity would help with the requirements — namely flexibility, choice, and programmability — for 5G networks and hyperscale cloud deployments.

In addition, Arccus made a number of on-chip scale and traffic management updates to ArcOS. These updates include traffic-aware scheduling for elephant flow detection and re-prioritization; dynamic load balancing that optimizes equal-cost multi-path routing (ECMP) in cloud and high-performance computing (HPC); a shared buffer architecture that reduces tail latencies; group management; and larger route scale.

With the launch of its 400G switches, Arrcus joins the three big switch players — Cisco, Arista Networks, and Juniper Networks — which all three offer switches integrated with their own software.

Last July, Juniper Networks said it was infusing its 400G innovations with its PTX IP series, QFX data center series, and its WAN series MX. Cisco, in October, launched its own line of 400G switches. Two of these are for its webscale customers and two of the new switches are for its Application Centric Infrastructure (ACI) architecture. Also in October, Arista Networks debuted its 400G switching line based on Broadcom’s 12.8 Tb/s Tomahawk 3 silicon.

“Now, with the pent-up demand among hyperscale cloud providers for higher capacities, transition to 400G is imminent but I see the technology becoming more mainstream only towards late 2019 (primarily due to the availability and cost-effectiveness of the optics),” wrote Garg. He noted that until the end of 2019, he expects customers to use 100G platforms to build-out their data center networks.