• <Maggiori informazioni su Intel.com

Understanding iWARP: Delivering Low Latency to Ethernet

Questo PDF è disponibile solo per il download

Understanding iWARP: Delivering Low Latency to Ethernet

For years, Ethernet has been the de facto standard local area network (LAN) technology for connecting users to each other and to network resources. Low cost and increasing Ethernet data rates have simplified growth for existing data networking applications and removed the wire-speed barriers to deployment in storage and clustering environments. However, for certain types of applications such as large-scale financial services, cloud computing, and high-performance computing (HPC), Ethernet’s inherent latency and limited message-processing rates have presented unacceptable performance barriers. As the data center evolves toward a more virtualized model, with a higher degree of abstraction applied to the network and servers, Internet Wide Area RDMA Protocol (iWARP) has emerged as an important enabler to help businesses get the full network throughput benefits of the latest 10 Gigabit Ethernet (10GbE) technologies.

What is iWARP?

WARP delivers converged, low-latency fabric services to data centers through Remote Direct Memory Access (RDMA) over Ethernet. The key iWARP components that deliver low-latency are as follows:

• Kernel Bypass. Removes the need for context switching from kernel-space to user-space
• Direct Data Placement Eiminates intermediate buffer copies by reading and writing directly to application memory
• Transport Acceleration Performs transport processing on the network controller instead of the processor

The iWARP specification, maintained by the Internet Engineering Task Force (IETF), supports transmissions over TCP and is implemented on top of IP networks using an existing Ethernet infrastructure.

Read the full Understanding iWARP Paper.