Enterprise networks are so last century. These days, anyone who is anybody is busy deploying fabrics – and not just on the server interconnect but across data center, campus and even wide area infrastructure as well.
Demand for fabric architectures has been ramping up since the beginning of the decade, driven primarily by the complex workloads hitting the enterprise and the need for broad flexibility when shuttling data between servers, storage and other points. It is only lately, however, that both fabric connectivity and data management capabilities have risen to the point that they can handle modern production workloads and the Big Data/IoT tidal wave to come.
A key requirement for advanced fabric topologies is broad, even universal, connectivity. This is leading many top IT vendors to unite under common formats that strive to be as open as possible. The OpenCAPI Consortium, for instance, is built around IBM’s Cache-Coherent Accelerator Processor Interconnect (CAPI) system that is seeking to supplant PCIe as the primary interconnect for modular infrastructure. Offering speeds as high as 150 GBps, the format has drawn support from Dell-EMC, HPE and Nvidia, as well as relative newcomers like Xilinx.
Meanwhile, many of these same companies support a similar initiative called Gen-Z that aims more toward rack-scale infrastructure through its use of peer-to-peer communications and semantic networking. The system is more attuned to handling large volumes of data than bursts of small-packet traffic, although it still has the capability to scale into the 100 GBps range and above while maintaining sub-100 ns latency. Gen-Z would likely be optimized for in-memory applications, real-time analytics and other functions that support Big Data workloads, while at the same time fostering advanced memory pooling and data-centric computing.
Proprietary fabric solutions are also hitting the channel as start-ups see an opportunity to edge their way into network markets that have been dominated by Cisco, Brocade and other top vendors for decades. A company called Datera just released the latest version of its Elastic Data Fabric that provides universal application and orchestration support to enable automated networking across public and private clouds. The fabric supports the VAAI (vStorage API for Array Integration) API and is certified under the VMware Reach for Storage program, and it can also be integrated into vCenter and vRealize for cloud provisioning and scaling. The company says this makes it ideal for cloud distributions of SQL, Exchange and Sharepoint, as well as homegrown MySQL applications.
Fabrics are also moving in on the Data Center Interconnect (DCI), the wide area infrastructure that links disparate data facilities under a distributed computing model. Pluribus Networks recently released a new Open DCI solution that utilizes standard VXLAN connections to create integrated software-defined architectures across remote locations. The system uses Pluribus’ Virtualization-Centric Fabric architecture to push layer-2 switching capabilities to the edge, with automated change management to ensure universal connectivity across regional infrastructure. Pluribus is also looking to integrate IT automation frameworks like Puppet and Ansible to provide scale across virtual and container-based environments.
Fabric architectures are not only more flexible, they provide added resiliency against failure since traffic can be easily routed away from malfunctioning hardware. For enterprises looking to expand their footprints for increasingly heavy data loads, and provide more reliable service to boot, fabric networking will be one of the go-to technologies in the coming decade.
Arthur Cole covers networking and the data center for Enterprise Networking Planet and IT Business Edge. He has served as editor of numerous publications covering everything from audio/video production and distribution, multimedia and the Internet to video gaming.