Marvell® FastLinQ® 41000 Series
●The FastLinQ 41000 Series leverages Marvell’s 15+ years of expertise in Ethernet by providing the highest levels of performance, efficiency, and scalability for the enterprise data center.
●For more effective use of the 10/25GbE bandwidth, the 41000 Series Converged Network Adapter offers switch-independent NIC partitioning (NPAR), which enables segmentation of a single 10/25GbE port into multiple network partitions and dynamic allocation of bandwidth to each port. The segmentation allows IT organizations to optimize resources while lowering infrastructure and operational costs.
●The evolution of data centers—triggered by high-density server virtualization, soft-ware-defined networking (SDN), and multitenant cloud computing platforms—demands a high-performance 10/25GbE solution that boosts CPU efficiency and reduces capital expenditures (CAPEX) and operational expenditures (OPEX) of the migration to 10/25GbE. The FastLinQ 41000 Series Adapters are the best choice for workload-inten-sive computing environments, providing reliable, high-performance 10/25GbE connectivity solutions.
■Features:
●Universal RDMA technologies—RDMA over Converged Ethernet (RoCE), RoCEv2, and iWARP
●PCI Express® (PCIe®) Gen 3 x8 (8GT/s) support
●Full line-rate performance across dual and quad ports
●Broad operating system (OS) and hypervisor support
●iSCSI-Offload and FCoE-Offload support
●Network boot support
◆iSCSI remote boot
◆FCoE remote boot from LUN
◆Preboot Execution Environment (PXE) 2.0
◆Unified Extensible Firmware Inter-face (UEFI) support
●Simplifies deployment and trouble-shooting using Windows® Admin Center, QLogic® Control Suite (QCS) CLI, QConvergeConsole® (QCC) PowerKit, UEFI human interface infrastructure (HII), in-OS utilities, QCC vCenter GUI and ESXCLI Plug-ins, and OpenStack® integration
●Switch-independent NPAR with up to 16 physical functions (partitions) per adapter
●Marvell Data Plane Development Kit (DPDK) high-speed packet processing engine delivers up to 38 million packets per second at 64B frame sizes
●Energy Efficient Ethernet (EEE) support for reduced idle power consumption in RJ-45-based networks (10GBASE-T variants only)
●MSI and MSI-X support
●IPv4 and IPv6 stateless offloads
●PCI-SIG® single root input/output virtualization (SR-IOV) with up to 192 virtual functions
●Comprehensive stateless offloads (see page 6)
●Auto negotiation: 1G/10G (BASE-T) and 10G/25G (on direct attach cable (DAC) cable using 10GBASE-KR/25GBASE-CR)
●FastLinQ SmartAN™ for intelligent connectivity that determines link speed, cable type and length, and Forward Error Correction requirements.
●RX/TX multiqueue
◆VMware® NetQueue
◆Windows® Hyper-V® Virtual Machine Queue
◆Linux Multiqueue
●Tunneling offloads
◆Windows Network Virtualization using Generic Routing Encapsulation (NVGRE)
◆Linux Generic Routing Encapsulation (GRE)
◆VMware, Windows, and Linux Virtual Extensible LAN (VXLAN)
◆Linux and VMware Generic Network Virtualization Encapsulation (GENEVE)
●Receive side scaling (RSS)
●Transmit side scaling (TSS)
●Support for virtual LAN (vLAN) tagging
●Support for jumbo frames larger than 1,500 bytes (up to 9,600 bytes)
●Network teaming, failover, and load balancing
◆Switch independent NIC teaming/bonding
◆Switch dependent NIC teaming/bonding such as link aggregation control protocol (LACP) and generic trunking
●Data center bridging (DCB)
◆Data center bridging capability exchange protocol (DCBX) link layer discovery protocol (LLDP)
◆Priority-based flow control (PFC)
◆Traffic Class over VLAN's 3-bit priority code point (PCP) field or Traffic Class over the IP header's 3-bit differentiated services code point (DSCP) field
◆Enhanced Transmission Selection (ETS)
◆Explicit Congestion Notification (ECN or CN)
◆Data Center Quantized Congestion Notification (DCQCN)
●Non-offloaded Storage over Ethernet
◆iSCSI using OS-based software initiators
●Offloaded storage over Ethernet
◆Increases server performance with full hardware offload for storage traffic
◆Industry-leading FCoE-Offload performance of up to 3.6 million IOPS, suitable for high-density server virtualization and large databases
◆Industry-leading iSCSI-Offload performance of up to 2.9 million IOPS, suitable for a diverse set of applications leveraging the flexibility of iSCSI
●Marvell Flow Filtering is supported on Linux® using the ethtool -u/-Ucommands. See the n-tuple Flow Filtering and Steering FastLinQ 41000/45000 Series Adapters Deployment Guide for more information.
41000 Series 、 41000 、 45000 Series 、 45000 |
|
|
|
Supplier and Product Introduction |
|
|
|
Please see the document for details |
|
|
|
|
|
|
|
English Chinese Chinese and English Japanese |
|
11/20 |
|
Rev. 1 |
|
TD-000998 |
|
1 MB |
- +1 Like
- Add to Favorites
Recommend
All reproduced articles on this site are for the purpose of conveying more information and clearly indicate the source. If media or individuals who do not want to be reproduced can contact us, which will be deleted.