Tomorrow’s High-Performance Networks

Increasing network performance in the enterprise

Today we are moving beyond the “giga era” and into the “tera era”. In fact, data is multiplying so quickly worldwide that it’s now being measured by the exabyte—a million terabytes. Teraflops, tera-IOPs (I/O operations per second) and terabytes of information represent a new scale of workload.

Consider, for instance, the huge gains in performance necessary to handle Wal-Mart’s 7.7 million terabytes per day if all items in its U.S. retail stores adopt radio frequency identification (RFID) tags. Changes like these will dramatically increase the need for more efficient communications and data processing in tomorrow’s enterprises.

Intel is developing performance networking solutions to meet these challenges. We are driving Intel® silicon and software innovation in key areas from network processors (NPUs) to data centers to provide superior processing performance for the computationally demanding enterprises of the future.

Intel focus areas in performance networking include:

• Server network acceleration
• Infrastructure and edge networking
• Self-healing platforms
• Learn more

Server network acceleration

Intel is developing innovative processing techniques to improve overall server application throughput and reduce network input/output (I/O) overhead. This is important research, as the rate of improvement in CPU, memory and bus frequencies can no longer keep up with the rapid growth and faster arrival rates of packet traffic.

The higher packet volumes and faster rates of arrival directly increase CPU overhead and reduce available cycles for user application. Server application throughput is throttled and data center performance degrades.

Infrastructure and edge networking

A significant amount of processing in the network occurs in platforms at the edge of IP networks, such as edge routers and multiservice switches. To meet the demand for new voice, video and data services, edge platforms will be required to support increasing bandwidth requirements and expand new services while preserving the existing infrastructure. Additional capabilities needed from these platforms include reliable aggregation and forwarding, robust quality of service (QoS), enhanced reliability, enhanced security, and the addition of new protocol standards.

Modular communications platforms (MCPs) will help the industry meet these demands. MCPs provide network equipment providers with reusable development and deployment platforms for more rapid innovation, greater solution flexibility, and faster time-to-market.

Intel® modular communications technology lab provides a cohesive, architectural vision for open, modular infrastructure and edge communication platforms based on Intel® processors. Current research focuses on developing key technologies for tomorrow’s multicore, highly threaded processor programming environments. Intel offers additional industry leadership through its work within the Intel® Internet Exchange Architecture (Intel® IXA) ecosystem and on standards ratification.

Since server network acceleration relies on so many system elements, any proposed solution must account for each facet of the process, from the network interface card (NIC) to the CPU and chipset to the protocol stack and operating system. These building blocks are interconnected elements that require a system-level solution to provide peak application throughout.

Research areas include:

• Network I/O modeling and simulation
• Lightweight threading
• Data placement
• Protocol stack optimizations
• Affinity/partitioning
• Remote Direct Memory Access (RDMA)

Self-healing platforms

Data center and infrastructure costs continue to rise. A major impediment to lowering their total cost of ownership (TCO) is the labor-intensive role people continue to play in mundane, day-to-day IT tasks.

The proliferation of Intel® architecture across enterprise computing environments (from cell phones to PDAs to mobile/desktop clients to servers) provides an opportunity for a common management solution across platforms. Providers of high-level adaptive stacks could provide management solutions that interface with Intel® building blocks across the entire enterprise, from the data center to the edge of the network.

Intel’s research focuses on ways to provide foundational technologies that will allow computer and networking industry companies to build these adaptive, self-managed platforms for future enterprises. By operating autonomously within the IT ecosystem, proactive building blocks could allow out-of-band (OOB) communications in all system states and enable management instrumentation, alerting, configuration, and remote control.

Learn more

• High Octane CRC Generation with the Intel Slicing-by-8 Algorithm, White PaperFile type/Size: PDF 235KB
• A Scalable and High Performance Software iSCSI ImplementationFile type/Size: PDF 490KB
• A Systematic Approach to Building High Performance, Software-based, CRC GeneratorsFile type/Size: PDF 341KB
• Server Network I/O Acceleration, White PaperFile type/Size: PDF 87KB
• TCP Onloading for Data Center ServersFile type/Size: PDF 144KB
• Intel Research and the Data Center of the Future, White PaperFile type/Size: PDF 163KB
• Intel® I/O Acceleration Technology
• Self-Managed Platforms, White PaperFile type/Size: PDF 135KB
• Silicon photonics