Newly Developed Demartek Technical Guide to Facilitate RoCE Deployment in 2017
RDMA over Converged Ethernet (RoCE) has proven to be effective at accelerating the performance of existing Ethernet infrastructure without being too costly. To help technical professionals implement RoCE capabilities in their systems, Demartek and the InfiniBand Trade Association’s RoCE Initiative have teamed up to create the RoCE Deployment Guide. The document explains the advantages of RoCE and shows IT managers how to deploy RoCE-enabled solutions quickly and with ease.
The Demartek RoCE Deployment Guide provides the following:
- A description of RDMA and RoCE technology
- How to deploy RoCE technology in Microsoft Windows and Linux environments
- Ethernet switch and adapter configuration guidance
- Diagrams showing how all the components were connected
- Performance of 25GbE and 100GbE RoCE technology compared to traditional TCP/IP
- Description of the products used from multiple vendors, including HPE, Mellanox, QLogic
The Deployment Guide is available for download from the RoCE Initiative’s resource library.
Network is Key to Achieving Higher Performing Systems
The industry continues to adopt more powerful CPUs, faster storage, and software-defined architectures designed to accelerate system performance when managing increasingly larger data sets. In a recent Network Computing article, Mellanox’s Kevin Deierling explores how these three industry trends have also shifted the performance bottleneck to the network level. He explains how advanced network protocols, such as RoCE, are key to achieving additional cost savings, scalability, and speed.
The full article provides a detailed overview outlining RoCE’s advantages for next generation networks.
Nonvolatile Memory Express (NVMe) Protocols Continue to Gain Popularity in IT Industry
NVMe is gaining traction in the network community and Demartek’s Dennis Martin believes more IT professionals should start paying attention. NVMe, NVMe over Fabrics, and other types of nonvolatile memories may eventually replace or supplement NAND flash due to its higher performance and lower latency capabilities. However, in order for that to happen, IT professionals must have a better understanding of the new protocols as well as the technologies that support them – such as RDMA interconnects.
To learn more about NVMe and its impact on the IT industry, read TechTarget’s full article.
TOP500 List Demonstrates InfiniBand and RDMA Impact on HPC Networking Trends
Many in the High Performance Computing (HPC) industry use the TOP500 List as a way to spot adoption trends in interconnect technologies. The current list shows that the standardized high performance interconnects, namely InfiniBand and Ethernet, currently account for 79 percent of the world’s fastest systems. However, Ethernet’s share is concentrated toward the end of the list representing less demanding systems while InfiniBand’s distribution is even.
A recent insideHPC article provides additional insight around HPC trends in the TOP500 List.
InfiniBand Overcomes Performance Bottlenecks in the CPU
A pressing topic for many in the networking industry is the comparison between onloading and offloading network function to the CPU for improving overall system performance. While onloading interconnect technologies may be easier to build, they fail to address the need for increased CPU utilization. A recent insideHPC article takes a closer look at the issue and explains how offloading interconnects, like InfiniBand, can overcome CPU performance bottlenecks, increasing the CPU’s availability for compute functions and improving overall system efficiency.
For a detailed overview of onloading versus offloading, read insideHPC's full article.
Mellanox Announces: New iFLYTEK Machine Learning Center to Leverage RoCE Solutions
Mellanox recently announced that a leading intelligent speech and language technologies company, iFLYTEK Co., will leverage end-to-end 25GbE and 100GbE RoCE solutions for their next generation machine learning center. The resulting partnership will enable iFLYTEK to achieve a remarkable 97 percent speech recognition rate. This new RoCE network infrastructure provides iFLYTEK with an architecture that is flexible, scalable, and easy to manage.
View Mellanox’s official announcement to learn more.
OpenFabrics Alliance (OFA) Workshop 2017 Call for Sessions & Registration Now Open
The 13th Annual OpenFabrics Alliance (OFA) Workshop will be taking place on March 27-31, 2017, at the Hyatt Regency in Austin, Texas. Each year, the workshop generates lively exchanges among Alliance members, developers, and users who all share a vested interest in high performance networks. OFA hosts its workshop annually as a joint effort among open source networking community members. The objective is to create an event driven by the needs and interests of its attendees.
The Call for Sessions is now open and industry representatives are encouraged to help shape this year’s discussions by leading a session focused on critical networking issues.
Suggested session topics, formats, and detailed submission instructions can be found in the Call for Sessions flyer.
Proposals must be received by February 3, 2017 at 5:00 p.m. PST. The Technical Program Committee will respond to proposals and announce session topics by February 17, 2017.
Given InfiniBand and RoCE’s impact on the high performance networking industry, we encourage IBTA members to contribute their insights and expertise at the OFA Workshop 2017. Submit your session proposal today.
The OFA has also opened early bird registration, which is available through February 13, 2017. For details on registration, lodging, and the event schedule, visit the OFA Workshop registration page or download the Registration and Sponsorship flyer.IBTA encourages you to share the call for sessions and registration information with peers and colleagues who might be interested as well.
For questions or comments, please contact email@example.com