Carding 4 Carders
Professional
The new technology will make networks faster and more efficient.
Google announced the launch of a data transfer technology called Falcon (hardware-accelerated transport layer) and its inclusion in the Open Compute project, which is focused on the joint development of open hardware specifications for equipping data centers. Falcon is touted as the next generation of Ethernet, capable of increasing throughput and data transfer efficiency in existing standard networks based on Ethernet and TCP / IP that are critical to bandwidth and latency, such as networks for high-performance computing and artificial intelligence systems.
The Falcon protocol is designed to be scalable for use in data centers and provide predictably high performance, low latency, flexibility, and extensibility. Falcon is the first to be supported in the Intel IPU E2000 (Infrastructure Processing Unit) series of network accelerators, which combine an Ethernet adapter with a programmable processor that can handle operations normally performed on the network stack or system side, such as traffic management, congestion control, and high-level protocol parsing.
To achieve minimal latency in high-speed Ethernet networks with packet loss capability, Falcon uses three principles: accurate measurement of the delay between sending a request and receiving a response (RTT), hardware traffic management for individual streams, and fast and accurate packet retransmission. These features are complemented by the ability to use multiple channels simultaneously (Multipath) and support for connection encryption.
Falcon includes the following technologies:
Falcon supports top-level protocols such as RDMA and NVM Express, and provides the ability to add support for other top-level protocols through extensions. This protocol is divided into three layers: the top-level protocol mapping layer, the transaction layer, and the packet delivery layer, which provides data flow management, resource planning,and data transfer reliability.
Google announced the launch of a data transfer technology called Falcon (hardware-accelerated transport layer) and its inclusion in the Open Compute project, which is focused on the joint development of open hardware specifications for equipping data centers. Falcon is touted as the next generation of Ethernet, capable of increasing throughput and data transfer efficiency in existing standard networks based on Ethernet and TCP / IP that are critical to bandwidth and latency, such as networks for high-performance computing and artificial intelligence systems.
The Falcon protocol is designed to be scalable for use in data centers and provide predictably high performance, low latency, flexibility, and extensibility. Falcon is the first to be supported in the Intel IPU E2000 (Infrastructure Processing Unit) series of network accelerators, which combine an Ethernet adapter with a programmable processor that can handle operations normally performed on the network stack or system side, such as traffic management, congestion control, and high-level protocol parsing.
To achieve minimal latency in high-speed Ethernet networks with packet loss capability, Falcon uses three principles: accurate measurement of the delay between sending a request and receiving a response (RTT), hardware traffic management for individual streams, and fast and accurate packet retransmission. These features are complemented by the ability to use multiple channels simultaneously (Multipath) and support for connection encryption.
Falcon includes the following technologies:
- Carousel: A traffic management mechanism for regulating the throughput and intensity of streams at the individual host level.
- Snap: A microkernel-based network subsystem that can be extended with modules to add additional functionality, such as network virtualization and traffic management.
- Swift: A congestion management mechanism for data centers that provides minimal latency while maintaining 100Gbps throughput per server at maximum load.
- RACK-TLP: A packet loss detection algorithm for TCP.
- PLB: A load balancing mechanism using congestion signals, which resulted in reduced load imbalance, packet loss, and latency in Google networks.
- CSIG (Congestion Signaling): A telemetry information exchange protocol used for congestion signaling and traffic management.
- PSP (PSP Security Protocol): A traffic encryption protocol.
Falcon supports top-level protocols such as RDMA and NVM Express, and provides the ability to add support for other top-level protocols through extensions. This protocol is divided into three layers: the top-level protocol mapping layer, the transaction layer, and the packet delivery layer, which provides data flow management, resource planning,and data transfer reliability.
