ConnectX -4 Lx EN - aravision.com

1 ETHERNET. ADAPTER CARDS. PRODUCT BRIEF. ConnectX -4 Lx EN . 1/10/25/40/50 Gigabit Ethernet Adapter Cards supporting RDMA, . Overlay Networks Encapsulation/Decapsulation and more ConnectX -4 Lx EN Network Controller with 1/10/25/40/50Gb/s Ethernet connectivity addresses virtualized infrastructure challenges, delivering best-in-class and highest performance to various demanding markets and HIGHLIGHTS. applications. Providing true hardware-based I/O isolation with unmatched scalability and efficiency, achieving the most cost-effective and flexible BENEFITS. Highest performing boards for solution for Web , Cloud, data analytics, database, and storage platforms. applications requiring high bandwidth, low latency and high message rate With the exponential increase in usage of data and I/O Virtualization Industry leading throughput and latency the creation of new applications, the demand for the ConnectX -4 Lx EN SR-IOV technology provides dedi- for Web , Cloud and Big Data highest throughput, lowest latency, virtualization and cated adapter resources and guaranteed isolation applications sophisticated data acceleration engines continues and protection for virtual machines (VMs) within Smart interconnect for x86, Power, ARM, to rise.

2 ConnectX -4 Lx EN enables data centers to the server. I/O virtualization with ConnectX -4 Lx and GPU-based compute and storage leverage the world's leading interconnect adapter EN gives data center administrators better server platforms for increasing their operational efficiency, improving utilization while reducing cost, power, and cable Cutting-edge performance in virtualized servers' utilization, maximizing applications complexity, allowing more Virtual Machines and overlay networks productivity, while reducing total cost of ownership more tenants on the same hardware. Efficient I/O consolidation, lowering data (TCO). center costs and complexity Overlay Networks Virtualization acceleration ConnectX -4 Lx EN provides an unmatched In order to better scale their networks, data center Power efficiency combination of 1, 10, 25, 40, and 50 GbE bandwidth, operators often create overlay networks that carry KEY FEATURES.

3 Sub-microsecond latency and a 75 million packets traffic from individual virtual machines over logical 1/10/25/40/50Gb/s speeds per second message rate. It includes native tunnels in encapsulated formats such as NVGRE Single and dual-port options available hardware support for RDMA over Converged and VXLAN. While this solves network scalability Erasure Coding offload Ethernet, Ethernet stateless offload engines, issues, it hides the TCP packet from the hardware Virtualization Overlay Networks,and GPUD irect Technology. offloading engines, placing higher loads on the host Accelerated Switching and Packet High Speed Ethernet Adapter CPU. ConnectX -4 Lx EN effectively addresses this Processing (ASAP2). by providing advanced NVGRE, VXLAN and GENEVE Low latency RDMA over Converged ConnectX -4 Lx offers the best cost effective hardware offloading engines that encapsulate Ethernet (RoCE).

4 Ethernet adapter solution for 1, 10, 25, 40 and and de-capsulate the overlay protocol headers, CPU offloading of transport operations 50Gb/s Ethernet speeds, enabling seamless enabling the traditional offloads to be performed Application offloading networking, clustering, or storage. The adapter on the encapsulated traffic for these and other Mellanox PeerDirect communication reduces application runtime, and offers the tunneling protocols (GENEVE, MPLS, QinQ, and so acceleration flexibility and scalability to make infrastructure run on). With ConnectX -4 Lx EN, data center operators Hardware offloads for NVGRE, VXLAN. as efficiently and productively as possible. can achieve native performance in the new network and GENEVE encapsulated traffic architecture. End-to-end QoS and congestion control Hardware-based I/O virtualization RoHS-R6.

5 2016 Mellanox Technologies. All rights reserved.. For illustration only. Actual products may vary. ConnectX -4 Lx EN 10/25/40/50 Gigabit Ethernet Adapter Card page 2. ASAP2 Mellanox PeerDirect redundancy and performance improvement. Mellanox ConnectX -4 Lx EN offers Acceler- PeerDirect communication provides high ConnectX -4 Lx EN's Reed-Solomon capability ated Switching And Packet Processing (ASAP2) efficiency RDMA access by eliminating introduces redundant block calculations, technology to perform offload activities in the unnecessary internal data copies between which, together with RDMA, achieves high hypervisor, including data path, packet parsing, components on the PCIe bus (for example, performance and reliable storage access. VxLAN and NVGRE encapsulation/decapsula- from GPU to CPU), and therefore significantly Software Support tion, and more.

6 Reduces application run time. ConnectX -4. All Mellanox adapter cards are supported Lx EN advanced acceleration technology ASAP2 allows offloading by handling the data by Windows, Linux distributions, VMware, enables higher cluster efficiency and plane in the NIC hardware using SR-IOV, while FreeBSD, and Citrix XENS erver. ConnectX -4. scalability to tens of thousands of nodes. maintaining the control plane used in today's Lx EN supports various management software-based solutions unmodified. As a Storage Acceleration interfaces and has a rich set of tools for result, there is significantly higher performance configuration and management across Storage applications will see improved without the associated CPU load. ASAP2 has operating systems. performance with the higher bandwidth two formats: ASAP2 Flex and ASAP2 Direct . ConnectX -4 Lx EN delivers.

7 Moreover, One example of a virtual switch that ASAP2 can standard block and file access protocols can offload is OpenVSwitch (OVS). leverage RoCE for high-performance storage access. A consolidated compute and storage RDMA over Converged Ethernet (RoCE) network achieves significant cost-performance ConnectX -4 Lx EN supports RoCE specifications advantages over multi-fabric networks. delivering low-latency and high- performance Distributed RAID. over Ethernet networks. Leveraging data center bridging (DCB) capabilities as well as ConnectX -4 Lx EN delivers advanced Erasure ConnectX -4 Lx EN advanced congestion control Coding offloading capability, enabling hardware mechanisms, RoCE provides efficient distributed Redundant Array of Inexpensive low-latency RDMA services over Layer 2 and Disks (RAID), a data storage technology that Layer 3 networks.

8 Combines multiple disk drive components into a logical unit for the purposes of data COMPATIBILITY. PCI EXPRESS INTERFACE CONNECTIVITY OPERATING SYSTEMS/DISTRIBUTIONS*. PCIe Gen compliant, and Interoperable with 1/10/25/40/50Gb Ethernet RHEL/CentOS. compatible switches Windows , , or link rate x8 Passive copper cable with ESD protection FreeBSD. Auto-negotiates to x8, x4, x2, or x1 Powered connectors for optical and active VMware Support for MSI/MSI-X mechanisms cable support OpenFabrics Enterprise Distribution (OFED). OpenFabrics Windows Distribution (WinOF-2). 2016 Mellanox Technologies. All rights reserved. ConnectX -4 Lx EN 10/25/40/50 Gigabit Ethernet Adapter Card page 3. FEATURES SUMMARY*. ETHERNET Advanced memory mapping support, allowing TCP/UDP/IP stateless offload 50 GbE / 40 GbE / 25 GbE / 10 GbE / 1 GbE user mode registration and remapping of LSO, LRO, checksum offload IEEE , 100 Gigabit Ethernet memory (UMR) RSS (can be done on encapsulated packet), 25G Ethernet Consortium 25, 50 Gigabit On demand paging (ODP) registration free TSS, HDS, VLAN insertion / stripping, Receive Ethernet RDMA memory access flow steering IEEE 40 Gigabit Ethernet STORAGE OFFLOADS Intelligent interrupt coalescence IEEE 10 Gigabit Ethernet RAID offload - erasure coding (Reed-Solomon) REMOTE BOOT.

9 IEEE Energy Efficient Ethernet offload Remote boot over Ethernet IEEE based auto-negotiation and KR OVERLAY NETWORKS Remote boot over iSCSI. startup Stateless offloads for overlay networks and PXE and UEFI. Proprietary Ethernet protocols tunneling protocols PROTOCOL SUPPORT. (20/40 GBASE-R2, 50 GBASE-R4) Hardware offload of encapsulation and OpenMPI, IBM PE, OSU MPI (MVAPICH/2), EEE , Link Aggregation decapsulation of NVGRE and VXLAN overlay Intel MPI. networks IEEE , VLAN tags and priority Platform MPI, UPC, Open SHMEM. HARDWARE-BASED I/O VIRTUALIZATION. IEEE (QCN) Congestion TCP/UDP, MPLS, VxLAN, NVGRE, GENEVE. Notification Single Root IOV. iSER, NFS RDMA, SMB Direct EEE (ETS) Multi-function per port uDAPL. IEEE (PFC) Address translation and protection MANAGEMENT AND CONTROL. IEEE Multiple queues per virtual machine INTERFACES.

10 IEEE 1588v2 Enhanced QoS for vNICs NC-SI, MCTP over SMBus and MCTP over Jumbo frame support ( ) VMware NetQueue support PCIe ENHANCED FEATURES VIRTUALIZATION Baseboard Management Controller interface Hardware-based reliable transport SR-IOV: Up to 256 Virtual Functions SDN management interface for managing the Collective operations offloads SR-IOV: Up to 16 Physical Functions per port eSwitch Vector collective operations offloads Virtualizing Physical Functions on a physical I2C interface for device control and port configuration PeerDirect RDMA (aka GPUD irect . communication acceleration SR-IOV on every Physical Function General Purpose I/O pins 64/66 encoding 1K ingress and egress QoS levels SPI interface to Flash Extended Reliable Connected transport (XRC) Guaranteed QoS for VMs JTAG IEEE and IEEE Dynamically Connected transport (DCT) CPU OFFLOADS.)