Netris

Netris operates as a software company delivering network automation, abstraction, and multi-tenancy solutions tailored for the operators of modern Artificial Intelligence (AI) and cloud infrastructures. Founded in 2017 by CEO Alex Saroyan, the company has its headquarters in Santa Clara, California. Saroyan's background as a network engineer for internet service providers and large-scale data centers, including an early role at the search engine Lycos, provided him with a distinct perspective on the operational differences between traditional data centers and hyperscale cloud providers. This experience was the genesis for Netris, which was created to provide the advanced, proprietary-style network automation used by hyperscalers to the broader market of public and private cloud providers.

The company targets a clientele of AI infrastructure operators, GPU cloud providers, and both public and private cloud service providers who need to manage complex, dynamic, and multi-tenant network environments. Netris addresses the significant challenge these clients face in configuring networks automatically and on-the-fly to accommodate constantly changing services without disrupting other users. Revenue is generated by offering its specialized network automation software as a product. The business model focuses on equipping network engineers and operators with the tools to build and run their own cloud services, rather than being a cloud provider itself.

The core product is a network automation platform centered around the Netris Controller, which serves as the single source of truth for the network. It allows engineers to model, simulate, and deploy network infrastructure using an infrastructure-as-code approach with tools like Terraform. A key feature is its ability to deliver AWS-style networking constructs such as Virtual Private Clouds (VPCs), elastic IPs, and load balancers on-premises. This is achieved through components like Netris's SoftGate, a horizontally scalable software gateway that provides functions like NAT and load balancing. The platform is designed to be multi-vendor and manages both Ethernet and InfiniBand fabrics, integrating with NVIDIA's UFM for InfiniBand and acting as the direct fabric manager for Ethernet switches from vendors like NVIDIA, Dell, and Arista. This provides hardware-level isolation for multi-tenancy, a critical requirement for maximizing the utilization of expensive GPU resources in AI clusters securely and efficiently.

Keywords: network automation, multi-tenancy, AI infrastructure, GPU cloud networking, private cloud networking, hybrid cloud solutions, network abstraction, VPC on-prem, cloud service provider networking, NetOps, network operations, data center networking, infrastructure as code, network fabric management, Alex Saroyan, EVPN, VXLAN, BGP, Ethernet fabric, InfiniBand management