格罗夫徽标

RDMA

Low-Latency RoCE v2 at 100/200Gbps

The GROVF RDMA IP core and host drivers provide RDMA over Converged Ethernet (RoCE v2) system implementation and integration with standard Verbs API. The RDMA IP is delivered with a reference design that includes the IP subsystem itself, the 100/200G MAC IP subsystem, the DMA subsystem, host drivers, and example application on software. The system drivers are integrated with OFED standard Verbs API and are compatible with well-known RNIC cards and software. The IP core also provides a low-latency FPGA implementation of RoCE v2 at 200Gbs or 100Gbps throughput.

RNIC和软性RoCE v2兼容

200Gb/s throughput, 2µs latency

Configurable number of
RDMA queue pairs

特点

  • 硬件操作的RC、XRC、RD、UC、UD服务
  • 传入和传出的SEND、RDMA READ、RDMA WRITE
  • 在FPGA和ECN中实现的内存保护域
  • 第三方MAC和DMA IPs
  • 主机上的标准Verbs API
  • 使用Verbs API的动态配置
  • 硬件重传和重新排序
  • 可定制的IP

通过基于FPGA的SmartNIC实现RNIC的使用案例

方框图、数据表和产品细节

产品操作

The solution is a soft IP implementing RDMA over Converged Ethernet protocol. It consists of FPGA IP integrated with MAC and DMA, plus the host CPU drivers and is compatible with a variety of BittWare’s FPGA cards. The 200Gbps IP is compatible with BittWare’s IA-440i Agilex 7 I-Series FPGA card, and the 100Gbps IP is compatible with BittWare’s IA-840f and IA-420f Agilex 7 F-series cards and XUP-VV8 and XUP-P3R UltraScale+ FPGA cards. The solution complies with Channel Adapter and RoCE v2 requirements as stated in the IB specification. The diagram above shows a simplistic architectural overview of the system. The data plane and reliable communication is hardware offloaded, and the implementation does not include CPU cores in the FPGA.

详细功能列表

  • 完全兼容已知的RNIC产品和软性RoCE实现(RoCE v2)。
  • 100 or 200 Gb/s throughput
    • 可配置的RDMA队列对
    • 1023个或更多
  • 200Gbps IP: under 2.7 µs software to software latency (roundtrip) and under 1 µs hardware to hardware latency (roundtrip)
  • 100Gbps IP: under 2.0 µs software to software latency (roundtrip) and under 300 ns hardware to hardware latency (roundtrip)
  • 硬件重传管理
  • 在FPGA中实现的存储器保护域
  • 使用ECN、PFC的拥堵控制
  • Can work with 3rd party MAC 
  • 使用Verbs API的动态配置
  • 在主机用户/内核空间的标准Verbs API
  • 硬件实现的可靠连接(RC)、扩展可靠连接(XRC)、可靠数据报(RD)、不可靠连接(UC)和不可靠数据报(UD)。
  • 传入和传出的SEND、RDMA READ、RDMA WRITE 

参考设计

该参考范例由三部分组成:

  • 带有参考设计的加密FPGA IPs,实现了RDMA协议
  • 为基于FPGA的RDMA适配器提供标准Verbs API支持的软件驱动程序
  • 建立在Verbs API之上的应用实例,展示乒乓测试结果:延迟和带宽

实施结果样本

器材 LUTs 片上存储器
UltraScale+ VU9P 170K 6Mb
Agilex 7 AGF014 170K 6Mb

兼容的FPGA卡

对价格或更多信息感兴趣?

我们的技术销售团队随时准备提供可用性和配置信息,或回答您的技术问题。