PRODUCT BRIEF Intel Ethernet Converged Network Adapter X520-QDA1 with Support for QSFP+ Connections Network Connectivity Intel Ethernet Converged Network Adapter X520-QDA1 Single-Port QSFP+ Converged Network (4x10GbE) Adapter Provides Ultimate Flexibility and Scalability in Virtual and Network Appliance Environments and compression. The adapter uses Intel Virtualization Technology a single PCI slot without the need to for connectivity (Intel VT-c) upgrade your switch infrastructure. Intel Ethernet Controllers include Intel Customers can deploy these adapters Virtualization Technology for connectiv- today using existing 10GbE switches ity (Intel VT-c) to deliver virtualized I/O giving IT managers an easy upgrade path performance optimizations and Quality of to 40GbE. Service (QoS) features designed directly in to the controllers silicon. Working in Intel Ethernet Converged Network conjunction with virtualization optimized Adapter X520-QDA1 (Single Port QSFP+ drivers, PCI-SIG* Single Root I/O Virtu- 4x10GbE) alization and Sharing (SR-IOV) can be The new QSFP+ single-port X520-QDA1 used to help reduce I/O bottlenecks, and connects your server to the latest 40GbE improve the overall server performance. switches with a single cable operating in 4x10GbE mode. This adapter can also Hypervisor BYPASS using SR-IOV Overview utilize existing 10GbE SFP+ switches us- Bypassing the hypervisor and allowing di- ing the QSFP+ to 4xSFP+ breakout cable. The best-selling Intel X520 dual-port rect hardware access by virtual machines, (2x10GbE) adapter is known for its high The QSFP+ adapter supports Direct At- reduces CPU overhead, reduces latency, performance, low latency, reliability, and tach Copper Cables and Intel Ethernet and increases network throughput. flexibility. The addition of a new low QSFP+ SR Optical Transceivers. Most of the current hypervisor releases profile QSFP+ 4x10GbE adapter to the have been enabled to partition a single X520 family delivers all of the capabilities Powered by the latest network controller, physical Ethernet controller in to multiple in a 4x10GbE adapter. The Intel Ethernet the Intel Ethernet 82599 Ethernet Con- virtual Ethernet controllers that can be Converged Network Adapter X520- troller, the X520-QDA1 adapter addresses used directly by VMs by taking advantage QDA1 adapter delivers a proven, reliable the demanding needs of the next-genera- the PCI-SIG* SR-IOV standard. The use of solution for deployments of high density tion data center by providing unmatched these virtual controllers, known as Virtual Ethernet for unified 10GbE network features for virtualization, flexibility for Functions (VF), enables additional QoS connections. The new single-port QSFP+ LAN and SAN networking, and proven, features in the controllers silicon to man- adapter delivers 4 channels of line-rate reliable performance. The Intel 82599 age and direct traffic such as traffic isola - 10Gb Ethernet in a single PCIe slot using controller is the industry standard for tion, port partitioning with bandwidth the latest PCI Express* (PCIe*) v3.0 bus 10GbE making it the most popular 10GbE allocation and on-chip VF-VF switching. technology for maximum bandwidth that controller on the market today. addresses the increasing capability of the Unified Networking and Storage latest Intel Xeon processor E5/E7 based Best Choice for Virtualization The family of Intel Ethernet CNA X520 platforms. Intel leads the industry in virtualization adapters lowers your data center total by being the first to provide virtualiza - This card allows customers to move cost of ownership (TCO) by providing the tion for all the major operating systems to 40GbE today for high bandwidth ability to route LAN and storage traffic and working with the OEMs to implement application requirements such as content over a single fabric. By enabling IT to virtualization not only on the adapter but distribution, high-end virtualization using unify LAN and storage traffic in a single also on the platform. multiple CPUs, network appliances, and network, Intel helps reduce costs and Applications Delivery Controllers (ACD) complexity, and enhance the efficiency of used for content caching, load balancing, your virtual infrastructure.Unified networking is ideal for cloud us - the cost of expensive FC HBAs and FC PCIe v3.0 (8 GT/s) age. Discrete data center networks for switch ports. PCIe v3.0 (8 GT/s) support enables cus- specific components such as storage and tomers to take full advantage of four 10 iSCSI Acceleration servers are no longer appropriate. As GbE links by providing a maximum of 63 companies move to cloud computing, net- The adapters provide complete support Gbps throughput on a single quad-port works must be simpler to manage while for proven native OS iSCSI initiators as card. supporting large volumes of traffic. Con - well as iSCSI boot. Intel Ethernet server verged networks, with storage and server adapters include hardware-based iSCSI Designed for Multi-core Processors traffic running across a unified fabric, pro - acceleration features that do not require Support for technologies such as multiple vide that capability. The benefits of such offloading to a proprietary TCP/IP stack. queues, receive-side scaling, multiple a unified network are simplification, lower iSCSI acceleration uses large send offload, MSI-X vectors, and Low Latency Inter- TCO due to infrastructure consolidation, Receive Side Coalescing and transmit rupts enable the X520 family of adapters and greater flexibility. send offloads to help reduce latency provide high-performance, 4x10 gigabit and lower CPU utilization. To improve In todays dynamic data center environ- connectivity in multi-core server plat- efficiency, MSI-X, Receive-side Scaling ment, you need to simplify as much as forms. These technologies distribute and Intel Ethernet Flow Director scale I/O possible. Thanks to unified networking network processing across multiple CPU processing across multiple CPU cores. from Intel, you can combine all the traffic cores, improving overall performance. Direct memory access (DMA) and direct of multiple data center networkssuch cache access (DCA) improve network data as iSCSI, FC, TCP/IP and NASusing one Intel Ethernet Flow Director processing efficiency, and data center efficient network fabric. You can also get Todays data centers depend on the bridging (DCB) supports multiple traffic world-class iSCSI and FCoE performance multiprocessing of high performance classes that can be prioritized for iSCSI with native OS initiators and intelligent capability of servers to increase system traffic. CPU and networking hardware offloads. throughput, responsiveness, and reliabil- Such innovations add up to a simpler, ity through the introduction of additional Flexible Port Partitioning Support more efficient data center environment. hardware threads, CPUs, or cores. But Intel Flexible Port Partitioning (FPP) is the in a multiprocessing environment, it is Intel has lead broad OS and software eco- ability to use SR-IOV Virtual Functions, essential to ensure a coordinated affinity system engagements to enable effective (which have, up until now, been thought of protocol processing and network ap- protocol processing for storage and LAN of as a strictly virtualization technology) plications on the same target cores. This traffic through native software storage in a bare-metal (or mixed) Open Source affinity significantly reduces contention initiators. Native initiators are: OS. This provides a way to very flexibly for shared resources, minimizes software and efficiently configure your Ethernet Easy to useit has the same manage- synchronization overheads between ports. ment interface that is integrated into the cores, and enhances cache efficiency. operating system The Intel Ethernet CNA X520 family does Receive Side Scaling (RSS) resolves the it all: 4x10Gigabit LAN, FCoE, and iSCSI Reliablebacked by the reliability and single-processor bottleneck by allowing truly delivering on the promise of unified trust from operating system vendor the receive side network load from a net- networking. Cost Effectiveintegrated into the work adapter to be shared across multiple Kernel means customers can implement cores. RSS enables packet receive-pro- Reliable Performance over standard Ethernet connection and cessing to scale with the number of avail- The single port QSFP+ X520-QDA1 dont have to pay extra for the feature able cores. However, RSS has a limitation, adapter includes a number of advanced it cannot steer an incoming network flow With the use of intelligent offloads, Intel features that enable industry-leading to the same core where its application is able to target processing functions performance and reliability. process resides. RSS does not maintain enabled with hardware assists or bypass the Traffic Flow -> Core (Application) offloads to optimize performance, scal - Intelligent Offloads relationship. If an application is running ability, and flexibility. The adapters stateless offloads acceler - on one core, while RSS has scheduled ate TCP/IP traffic while preserving proces - receive traffic to be on another core, poor Support for Fibre Channel sor computing power for critical applica- cache efficiency and significant core- over Ethernet (FCoE) tions. By doing operations in hardware, to-core synchronization overheads will FCoE encapsulates Fibre Channel frames rather than a local micro-processor, the result. The overall system performance over standard Ethernet networks, en- X520-DA4 performance is able to scale can be significantly degraded. abling Fibre Channel to take advantage with the processor generational gains. Intel Ethernet Flow Director and the of 10 GbE networks while preserving Security Optimizations Application Target Routing (ATR) service its native protocol. The X520 adapt- found in Intel Ethernet Controllers, is ers offer FCoE hardware acceleration to The adapters support IPsec offload for an advanced network offload technol - provide performance comparable to FC Microsofts* Network Access Protec- ogy that provides the benefits of parallel HBAs. The adapters support Data Center tion (NAP), Active Directory,* and future receive processing in multiprocessing Bridging (DCB), also known as Converged security capabilities in Windows 7* and environments that automatically steer Enhanced Ethernet (CEE), which enables Windows 8*. An X520 adapter enables incoming network data to the same core customers to configure traffic classes and customers to run a secure network envi- on which its application process resides. priorities to deliver a lossless Ethernet ronment without sacrificing performance. Intel Ethernet Flow Director and ATR fabric. An Intel Ethernet Converged preserve the Traffic Flow -> Core (Ap - Network Adapter X520 reduces TCO by plication) relationship. As a result, Intel eliminating redundant fabrics and saves 2