Delivering next generation infrastructure with the Intel Ethernet 800 Series

Data is multiplying—doubling every 2-3 years, yet only a small percentage of the data is analyzed and used. The ability to process and analyze data to improve products, services, and customer experiences as well as to glean insights is vital to business competitiveness. Infrastructure needs to evolve to address growing business demands by moving faster, storing more, and processing more data.

To move data faster, we’ve delivered new exciting capabilities with our Intel® Ethernet 800 Series network controllers and adapters. The 800 Series products have been deployed at key customers and this summer we delivered new adapters with speeds up to 100GbE that are now widely available. New capabilities, like Application Device Queues, Dynamic Device Personalization and Non Volatile Memory over Fabric, are driving meaningful performance improvements for our customers.

At large scale, application response time predictability becomes more critical to meet Service Level Agreements. We introduced Application Device Queues (ADQ) which provides advanced traffic steering dedicating queues to high priority apps, including web, caching, databases with great results. For example, with Memcached we’ve seen up to 60% improvement in predictability and up to 60% latency reduction using ADQ. Customers are excited over ability to use ADQ to better meet SLAs.

For Comms Service Providers, new and custom networking protocols are emerging to deliver data. If the NIC cannot recognize the new protocols, they will be processed in software which can reduce performance. To address this, we’ve enhanced our Dynamic Device Personalization, or DDP, which provides for dynamic configuration of the 800 Series programmable pipeline in order to recognize a wide variety of protocols. This delivers value to customers by increasing throughput, reducing latency and improves total cost of ownership by reducing CPU overhead. A great example of this is collaboration between Intel and SK Telecom to demonstrate a 5G Standalone (SA) User Plane Function (UPF) capability based on 2nd gen Intel® Xeon® Scalable Processors and our 800 Series Network Adapters. Using intelligent classification, steering, and processing of traffic we demonstrated up to 78% reduction in latency and 88% reduction in jitter1.

A year has passed since I shared my thoughts on high performance storage. I’m happy to share the next drill-down into how we improve the performance of storage networks with NVMe over Fabrics implementations. With the 800 Series, we support multiple Ethernet storage transport protocols to provide customers with choice and performance—iWARP and RoCEv2 RDMA and NVMe over TCP – which we accelerate using ADQ.

At the Storage Developer Conference last week, we gave more details on our support for iWARP and RoCEv2 RDMA, NVMe/TCP with ADQ. We delivered technical presentations including how to tune and optimize NVMe protocols, how to leverage SPDK, and we did a joint discussion with Lightbits. For more information, go to the SNIA website.

Earlier I mentioned organizations want to improve their throughput, lower their latency, and increase their overall predictability. We recently collaborated with Lightbits to work together to improve Non Volatile Memory Express (NVMe) over TCP performance and the results were very impressive with up to 30% predictability increase and up to 50% latency reduction2.

IOPs Single Thread vs. 32 Thread

Latency Single Thread vs. 32 Thread

NVMe/TCP is an industry storage transport standard that enables disaggregated SSD storage to operate at efficiency levels previously possible only through direct-attached solutions. While prior technologies have made network fabric-based NVMe storage possible, they typically involve a limited ecosystem, specialized hardware, and extra complexity in deployment. NVMe/TCP, combined with the Intel Ethernet 800 Series Network Adapters with Application Device Queues (ADQ), helps to remedy these concerns. ADQ enhances NVMe/TCP by lowering latency while retaining ease of implementation, efficiency, and scalability benefits. When combined with Lightbits Labs’ LightOS™ and Intel 3D NAND SSDs, this approach provides a comprehensive and convenient NVMe/TCP-based storage solution.

Traditional DAS vs Scalable NVMe/TCP

At Intel, we believe in the importance of disaggregated storage. We have entered into a strategic partnership with LightBits including investment from Intel Capital.

Choosing the most appropriate NVMe-oF for your organization is only one piece of a larger puzzle. To fully realize the advancements being made in data center connectivity requires a holistic view of operations. Intel Ethernet 800 Series network controllers and adapters combined with 2nd Generation Intel Xeon Scalable processors, Intel Optane™ DC persistent memory, and Intel SSDs, provide an excellent foundation for our customer’s evolving data-centric workloads. For more information, visit intel.com/ethernet.


Notices & Disclaimers;

1 Source: Low Latency 5G UPF Using Priority Based 5G Packet Classification Whitepaper
2 Source: Performance claims made by Lightbits Labs based on their internal testing in July, 2020 (for further information, please see www.lightbitslabs.com/ty-wp-scalable-low-latency-nvme-tcp-storage/). For more complete information about Intel performance and benchmark results, please visit www.intel.com/benchmarks.

Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors.
Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more complete information visit www.intel.com/benchmarks.
Performance results are based on testing as of dates shown in configurations and may not reflect all publicly available updates. See backup for configuration details. No product or component can be absolutely secure.
Your costs and results may vary.
Intel technologies may require enabled hardware, software or service activation.
© Intel Corporation. Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the property of others.

Published on Categories Data Center NetworkingTags , , ,
Patricia Kummrow

About Patricia Kummrow

Patty Kummrow is a VP in the Data Center Group and the GM of the Ethernet Networking Division at Intel Corporation. She leads the design, development, manufacturing, and marketing of Intel® Ethernet Network Adapters, Controllers, Switch Silicon, along with next generation solutions used to accelerate networking, storage, and network security in data centers. Kummrow has two decades of experience in CPU design and technical leadership. She has led multiple teams developing Intel processors for data center, networking, storage, and autonomous driving applications. She holds a B.S. in Electrical Engineering from the University of Texas, El Paso and a M.S. in Technology Management from Lehigh University.