Netronome has tuned its Agilio Server Networking Platform for delivering hardware-acceleration for OpenStack networking, allowing data centers to accelerate applications such as network virtualization, security, load balancing and telemetry using different data plane options suitable for use cases spanning traditional IT to IaaS and Telco NFV workloads. Netronome's own testing has found a 5X boost in VM performance when network functions are offloaded to its Agilio interface cards.
At this week's OpenStack Summit in Austin, Netronome, in collaboration with Ericsson and Mirantis, is showcasing acceleration of open source datapath implementations, specifically, Open vSwitch (OVS), Stateful Firewall (Connection Tracking) and OpenContrail vRouter using Agilio CX intelligent server adapters. The company said this approach will be incorporated as an enhanced OpenStack networking plug-in architecture specification. As a result, critical networking functions that would otherwise hamper performance of the OpenStack implementation are offloaded to the Agilio platform, accommodating significantly more virtual machines per server leading to up to 6X lower TCO and higher services revenue per server compared to traditional NICs.
“The OpenStack platform has become the de facto cloud and SDN orchestration tool and server-based networking has become the cornerstone of pervasive SDN and cloud-based networks,” said Sujal Das, senior vice president and general manager, strategy and marketing at Netronome. “Our flagship Agilio product-based demonstrations and proposed open specification bring much needed hardware-based efficiency to that mix.”
“Mirantis OpenStack delivers comprehensive cloud orchestration tools that are optimized for Enterprise IT and NFV deployments,” said Kamesh Pemmaraju, vice president of product marketing at Mirantis. “Utilizing Netronome’s expertise in server-based networking and its Agilio intelligent server adapters, our joint pioneering work promises to provide the industry with new OpenStack networking capabilities that will enable rapid SDN deployments, delivering significantly higher networking performance and increases in server efficiency.”
Netronome plans to release beta releases of products that integrate OpenStack with Agilio Software for OVS, Stateful Firewall and OpenContrail vRouter, and an open enhanced OpenStack networking plug-in architecture specification in Q3 2016.The specification enhancements will propose extension of the current SR-IOV OpenStack plug-in mechanism for efficient packet delivery to virtual machines and add the OVS, Stateful Firewall and vRouter data paths implemented in intelligent server adapters such as Netronome’s Agilio CX products. Through support of innovative VirtIO extensions, hardware-based performance and efficiency benefits are delivered without compromising operational agility, while virtual machine migration capabilities needed for efficient server resource utilization are preserved.
http://www.netronome.com
Netronome's Agilio Server Networking Accelerates Cloud Data Centers
The company said its hardware and software-based Agilio platform delivers up to 5X higher throughput while reducing CPU requirements by up to 80 percent compared to traditional NICs and server-based networking implemented in software.
Server-based networking is being widely deployed in cloud data centers to handle virtualization, firewalls, load balancing, telemetry, zero-trust security using micro-segmentation, virtual network functions (VNFs) and application-based analytics. The big cloud providers (AWS, Microsoft Azure, Google) are using server-based networking in their mega data centers. Netronome's Agilio solution accelerates such server-based networking functions by offloading compute-intensive flow and tunnel processing from the CPUs.
The Agilio CX intelligent server adapters (ISAs) are based on Netronome's own flow processing silicon (NFP-4000) and software architecture (Agilio Software). The Agilio ISAs use onboard memory to support up to two million security policies, and deliver 28Mpps of throughput using hardware-based acceleration.
