SEARCH FINANCIAL SERVICES INFRASTRUCTURE SECURITY SCIENCE INTERVIEWS

 

     

Packet, Netronome Team for Cloud-Native Workloads at the Edge

December 18, 2018

Packet and Netronome detailed a new hardware platform that efficiently powers cloud-native workloads at the edge. The joint design, which will be deployed throughout all of Packet’s cloud and edge sites, combines the latest in disaggregated computing architecture in a cost-effective bare metal form factor.

Cloud-native and edge workloads demand stringent cost, power and latency parameters without compromising security and the pace of innovation. Packet’s new compact microservers meet these requirements while providing the tenant isolation needed in a bare metal environment. Due to its ultra-compact and low-power design (<100W per microserver), up to 120 microservers can fit in a single 42U rack, enabling the build-out of performance-driven edge infrastructures while maintaining real estate and power requirements.

“With this platform we are bringing the lessons of hyperscale providers directly to the edge with a disaggregated, accelerator-based architecture and a custom designed hardware model,” said Zachary Smith, CEO of Packet. “We see cloud-native network workload driving huge demand at the infrastructure edge and the combination of SmartNIC technology and eBPF expertise from Netronome brings not only significant cost savings but also unique performance capabilities to our content, IoT and Telco customers.”

Each of Packet’s new Edge cloud microservers with integrated Netronome SmartNIC technology delivers up to 25Gb/s throughput while enabling highly scalable security policies and control close to the cloud-native applications. As a result, Packet is able to deploy less complex, lower power and more cost-effective top-of-rack switch solutions. At the same time, developers are empowered with access to deep network acceleration capabilities through the use of eBPF programming for innovating in the areas of switching, security, telemetry and load balancing.

“We have witnessed a dramatic increase in the usage of eBPF due to the inherent advantages it provides to developers and operators of cloud-native infrastructure – owing to the programmable sandbox environment it runs in,” said Joe Stringer, Linux eBPF subsystem developer and contributor at the Cilium project. “The ability to transparently offload eBPF in edge cloud microservers with SmartNICs brings that flexible sandbox into hardware, allowing developers to more rapidly build and tune their applications to run at line-rate.”

With the end of Moore’s Law and the rise of cloud-native principles, hardware is evolving to take advantage of new disaggregated architectures. As the data collected and processed at the edge grows, the need for low-latency networking and application offload at breakthrough economics becomes critical.

“New edge-based technologies and applications such as autonomous vehicles, AI, and augmented reality are heavily latency dependent and require the strongest of security without sacrificing performance,” said Niel Viljoen, CEO and founder of Netronome. “Packet’s new Netronome SmartNIC-accelerated, low-latency Edge cloud infrastructure with eBPF offload will enable developers to effectively utilize the massive amounts of collectable data on the edge to build and test highly secure, real-time applications.”

Packet and Netronome worked closely with ODM partners to design and manufacture the dense, high performance networking model needed to support powerful microservers that are highly efficient on a cost per watt basis. The investments made by the companies are cross architecture and will extend to Intel, AMD and Ampere.

The microserver is built leveraging the open infrastructure designs from the Open19 Foundation, enabling hyperscale efficiency on common rack and power infrastructure, while enabling an edge-friendly operating model that lowers the cost to deploy compute in remote locations.

Terms of Use | Copyright © 2002 - 2017 CONSTITUENTWORKS SM  CORPORATION. All rights reserved. | Privacy Statement