Tag: network operating system

  • Arrcus Expands Network Services Spanning Hybrid Clouds

    Arrcus Expands Network Services Spanning Hybrid Clouds

    Arrcus has added virtualized distributed routing (VDR) and multi-cloud networking (MCN) capabilities to ArcOS, a network operating system (NOS) that can be deployed on-premises or in the cloud.

    Company CEO Devesh Garg said these additional capabilities provide the programmatic network services that IT organizations require to foster deployments of hybrid cloud computing environments.

    Arrcus MCN provides the ability to securely connect workloads and data across any cloud, while VDR eliminates the need to rely on physical routers that require their own chassis, he said.

    Arrcus has been making the case for a disaggregated approach to networking that eliminates dependencies on physical infrastructure. Instead, IT teams can deploy ArcOS on a cloud service or any white-box appliance they choose in an on-premises IT environment. That approach eliminates the possibility of becoming locked into specific network hardware platforms, he said.

    Garg added it also provides DevOps teams with a consistent foundation on which to deploy distributed applications. Rather than having to navigate different NOSes running in the cloud and on-premises IT environments, ArcOS presents a consistent set of application programming interfaces (APIs) for accessing network services.

    In fact, the need to programmatically invoke those network services is what’s driving the convergence of network operations and DevOps teams, he noted. As applications become more distributed in the age of the cloud, IT organizations are looking for more efficient ways to expose network services on-demand.

    That convergence, however, will take time. Many networking teams still tend to prefer to acquire routers and switches as physical devices that include a NOS rather than deploying an operating system on infrastructure they select and manage. However, as DevOps teams continue to push for access to network services that can more easily scale up and down as required, it’s now only a matter of time before software-defined networking services become a default requirement.

    In fact, Garg said, the rise of microservices will force that issue as DevOps teams look to deploy applications that will be updated frequently.

    In addition, he noted, the rise of edge computing platforms and 5G wireless networking services will simultaneously push organizations toward virtualized network services.

    It’s not clear at what rate network services are being transformed. Software-defined networking services are already readily available in the cloud. Progress in on-premises IT environments has been tied to the willingness of IT organizations to fund upgrades to core network infrastructure. The challenge many IT organizations now face in the wake of the economic downturn brought on by the COVID-19 pandemic is finding a way to achieve that goal cost-effectively. White-boxes provide a less costly alternative to switches from vendors such as Cisco Systems and Arista Networks. At the same time, the total cost of delivering network services across a hybrid cloud computing environment can decline when a common NOS is employed across all platforms.

    Regardless of the path forward, the one thing that is certain is legacy network infrastructure has become a major bottleneck at a time when IT organizations need to be more agile than ever.

  • NVIDIA Aims to Drive Disaggregation Following Cumulus Networks Buy

    NVIDIA Aims to Drive Disaggregation Following Cumulus Networks Buy

    Adoption of open source networking may be about to get a boost following the forthcoming acquisition of Cumulus Network by NVIDIA.

    The acquisition of the provider of a network operating system (NOS) based on Linux follows an earlier move by NVIDIA to acquire Mellanox, a provider of networking and storage systems.

    Kevin Deierling, senior vice president of marketing for the NVIDIA networking business unit, said Cumulus Networks adds a programmable NOS to the provider of graphical processor units (GPUs) that should find greater adoption as more organizations disaggregate the acquisition of network software and hardware. That trend toward disaggregation is especially strong in data centers and cloud computing environments that are building artificial intelligence (AI) applications based on NVIDIA GPUs, he said.

    As the worlds of DevOps and network operations (NetOps) continue to converge, NVIDIA is betting interest in a programmable NOS that is designed to scale will increase as well, added Deierling.

    Cumulus Networks has been trying to drive the adoption of Cumulus Linux since 2010 with mixed success. Cumulus Linux runs on more than 100 hardware platforms, including switches from Mellanox. While cloud service providers and some high-performance computing (HPC) environments have disaggregated networking software and hardware, most enterprise IT organizations continue to buy switches from Cisco Systems and Juniper Networks that come integrated with a NOS.

    Cumulus Networks CEO Josh Leslie said in time it’s apparent more enterprise IT organizations will be moving toward disaggregation of network software and hardware, especially as more application workloads move into the cloud. A NOS that is not tied to proprietary ASIC technologies embedded in a switch from Cisco or Juniper Networks makes it feasible for IT organizations to benefit from Moore’s Law as each new x86 processor generation is added, he noted.

    At the same time, he added, the tools for programming a NOS based on Linux are a lot more familiar and accessible to developers than tools designed for proprietary NOS platforms.

    It’s too early to say to what degree DevOps teams are going to need programmable access to shared networking resources. Ideally, developers will want more programmatic control over network resources as applications become more distributed than ever. That approach would enable them to also provision network resources to applications via application programming interfaces (APIs) without having to wait for network administrators to provision them manually.

    However, network bandwidth is a shared resource inside many organizations, and the concern is that one application will consume all the available network bandwidth at the expense of every other application on the network. As such, network administrators have been embracing automation and self-service portals to make network resources more readily accessible to developers without ceding total control of the network.

    Regardless of the approach, networking is becoming more agile. NetOps and DevOps teams may not necessarily converge; however, as networking becomes another class of infrastructure that can be treated as code, it’s clear the days when networking services are largely managed in isolation from the rest of the IT environment are coming to a close.