Extreme Networks: EMC VSPEX Integration for Data Centers A SOLUTION WHITE PAPER

Transcription

1 Extreme Networks: EMC VSPEX Integration for Data Centers A SOLUTION WHITE PAPER

2 Table of Contents Business Value...3 Extreme Networks VSPEX Converged Infrastructure Solution Overview...4 Key Components of the VSPEX Solution...6 VNX Customer Benefits...8 Key Components of the VSPEX Solution...8 Network Extreme Networks Summit X670 Series Extreme Networks 7100 Series ExtremeXOS Extreme Networks High Availability High Availability Features Software Defined Networking (SDN) Platform...17 Virtualization VMware vsphere VMware vcenter Compute Lenovo RD Outstanding Efficiency and Value Legendary Think Reliability Open Standards/Easy Serviceability...22 QLogic Conclusion...25 References...25 EMC VSPEX Integration for Data Centers White Paper 2

3 WHITE PAPER Converged Infrastructure EMC VSPEX Proven Infrastructure provides customer choice for best-in-breed components when optimizing for virtualizing business-critical applications in a high performing enterprise or data center infrastructure. VSPEX gives Value-Added Resellers (VARs), Channels, and Partners the ability to design and implement a proven solution that can be customized based on specific customer needs. The EMC VSPEX Private Cloud with Extreme Networks provides an all Ethernet pre-validated Private Cloud system that leverages the low cost curve of Ethernet and provides full hardware and software redundancy. This Proven Infrastructure solution leverages the VMware ESXi 5.1 Virtualization Layer, the highly available EMC VNX family of storage systems, and the Extreme Networks high performing 10/40/100G Ethernet family of Data Center Switches. This guide provides a solution overview and architecture, as well as high level deployment and steps to successfully deploy an EMC VSPEX with Extreme Networks Ethernet SAN solution. Business Value VSPEX accelerates IT transformation by enabling faster deployments and lowering risk while providing customer choice to pick best-in-breed validated partners and technologies, including Extreme Networks for high performance Ethernet systems. With Extreme Networks, the EMC VSPEX solution offers a high performance, high value and open standards approach that fits a diversity of business requirements. The Extreme Networks Data Center Switches offer high value through highdensity, low-power consuming and open standards platforms. This lowers both your CAPEX and OPEX since you have fewer physical switches to deploy that also have much lower power consumption to lower the variable power costs of the data center budget. Additionally, the open standards approach allows for seamless product integration with EMC, assuring the EMC VSPEX integration and future proofing for additional open standards requirements, including SDN with OpenFlow and OpenStack. EMC VSPEX solutions use the VNX unified storage platform that combines high performance I/O and flexible SAN options to meet a variety deployment requirements, including file, block, and object storage in a scalable, easy-to-use solution. This next-generation storage platform combines powerful and flexible hardware with advanced efficiency, management, and protection software to meet the demanding needs of today s enterprises. The VNX series is powered by Intel EMC VSPEX Integration for Data Centers White Paper 3

4 Xeon processors for intelligent storage that automatically and efficiently scales in performance, while ensuring data integrity and security. The operating systems deployed on the Extreme Networks switches ensures resilient and predictable features and performance for the EMC VSPEX Private Cloud users and workloads. Providing Best-in-Breed technology components and leveraging the low cost structure of Ethernet enables: High Performing Private Cloud Solutions All Ethernet eliminates topology inefficiencies, lowering CAPEX and OPEX Pre-validation Design and Testing provides solution assurance for maximum uptime and lower OPEX costs. Open Standards approach provides SDN ready infrastructure for OpenFlow and OpenStack. Extreme Networks VSPEX Converged Infrastructure Solution Overview The VSPEX solution uses the EMC VNX series storage array platform and VMware vsphere 5.1 to provide storage and server hardware consolidation in a private cloud while Extreme Networks provides the high speed, low latency network infrastructure. The virtualized infrastructure is centrally managed to provide efficient deployment and management of virtual machines and associated storage. The guiding principles for the solution design are: Function General purpose business applications that run in a virtual environment. Usability Simplified debug, maintenance, upgrade and monitoring with pre-determined network, storage and compute configurations, all validated and documented with procedures. Performance Targets a general-purpose workload across multiple virtual machines. Example workloads would be medium-to-light database, web application or mail server processing. Resilience/Reliability This design is intended to provide N+1 redundancy throughout the system to recover from typical single points of failure at each layer of the system. Compute, network and storage all provide fail over and fail back of services without impacting business applications. Ease of Deployment Channel partners trained on the platform can deploy a system ready for business application service enablement in days versus weeks or months for a similar solutions that are not pre-engineered Scalability Designed from the ground up to enable pay as you grow scaling. Compute and storage layers have assigned ratios that allow customers to incrementally grow each tier to match new business demands. For example, a preengineered server to disk ratio is designed into the architecture. EMC VSPEX Integration for Data Centers White Paper 4

5 Manageability/Maintainability The best practices to configure each component in the system are documented to maximize the value of the tools used to manage a converged Infrastructure platform. Some examples are: BIOS settings, MTU size, onboard IPMI capabilities, power consumption and rack design. Economic and technical constraints, and tradeoffs N+1 design requires powered and connected compute, storage and network infrastructure to be lightly loaded during normal business production hours. This architecture includes an optional server to enable high availability at the compute layer, so if a server fails, the system has enough resources to maintain business operations. The design is an active/passive design aimed at availability and reduced complexity. MINIMUM SOLUTION REQUIREMENTS To qualify as an EMC VSPEX solution, specific baseline hardware and software criteria must be met. HARDWARE RESOURCES HARDWARE CONFIGURATION NOTES Servers (hypervisors) Network Infrastructure EMC VNX Servers (Infrastructure) Minimum Memory: 2GB RAM per virtual machine 2GB RAM reservation per hypervisor host CPU: One vcpu per virtual machine Maximum of four vcpu per physical core Network: Six 1GbE NIC ports per physical server or two 10GbE NIC ports partitioned to allow network isolation Minimum switching capacity: Two physical switches Sufficient ports for Hypervisor host connectivity 1GbEport per control station for managment (VNX) Minimum of four 1GbE ports per data mover for data (VNX) or appropriate 10GbE connectivity Two data movers (active/standby) Minimum of four 1GbE interfaces per data mover or two 10GbE interfaces per data mover Minimum number required: Two physical servers 16GB RAM per server Four processor cores per server Two 1GbE ports per server The hypervisor will be constructed as a single cluster containing all of the compute nodes. Redundant LAN Configuration N/A N/A SOFTWARE RESOURCES SOFTWARE CONFIGURATION VNX OE for file: Realese OE for block: Release 31 ( ) EMC VSI for WMware vsphere Version 5.1 VIRTUAL MACHINE Base Operating System VDBench Microsoft Windows Server 2008 R Notes: VDBench was used to validate this solution. It is not a required component for production. EMC VSPEX Integration for Data Centers White Paper 5

6 VSPEX SYSTEM PROFILE DATA COLLECTION COUNTER VALUE Virtual machine operating system Variable, minimum Windows 2008 Processors per virtual machine 1 Number of virtual processors per physical CPU core 4 RAM per virtual machine Average storage available for each virtual machine Average IOPS per virtual machines Number of datastores to store virtual machine disks Number of virtual machines per datastore Disk and RAID type for datastores 2GB 100GB 25 IOPS Variable Variable RAID 5, 600GB, 15k RPM, 3.5-inch SAS disks Key Components of the VSPEX Solution STORAGE The storage layer is critical for the implementation of the private cloud. With multiple hosts accessing shared data, many of the use cases defined in the private cloud can be implemented. The EMC VNX storage family used in this solution provides high-performing data storage while maintaining high availability. NETWORK The network layer provides the high performing, resilient and fault tolerant connectivity between the compute and storage systems. The VSPEX program defines the minimum number of required network ports, provides general guidance on network architecture, and enables the customer to implement the solution by using any network hardware that meets these requirements. VIRTUALIZATION The virtualization layer decouples the physical implementation of resources from the applications that use them. In other words, the application view of the available resources is no longer directly tied to the hardware. This enables many key features in the private cloud concept. COMPUTE The compute layer provides memory and processing resources for the virtualization layer software, and for the applications running in the private cloud. The VSPEX program defines the minimum amount of required compute layer resources, and enables the customer to implement the solution by using any server hardware that meets these requirements. EMC VSPEX Integration for Data Centers White Paper 6

7 EMC VSPEX Integration for Data Centers White Paper 7

8 The solution architecture was designed and tested based on the goals below to enable a robust compute platform for typical business applications without sacrificing performance or availability: Matched Compute Cores and Memory to disk spindles for typical business work loads Converged network adaptors equipped with hardware offload for storage processing, allowing more CPU cycles for business applications Simple growth rules to increase capacity In band VMware ESX management path over redundant interfaces; other components are managed out of band (network, storage, compute hardware) Reduced cabling requirements Converged layer 2 Ethernet fabric all storage, network and high availability traffic. Key Components of the VSPEX Solution STORAGE The storage layer is critical for the implementation of the private cloud. With multiple hosts accessing shared data, many of the use cases defined in the Private Cloud can be implemented. The EMC VNX storage family used in this solution provides high-performace data storage while maintaining high availability. EMC VNX SERIES VNX Customer Benefits EMC VNX Series is high performance unified storage with unsurpassed simplicity and efficiency, optimized for virtual applications. The EMC VNX family is optimized for virtual applications; and delivers industry leading innovation and enterprise capabilities for file, block, and object storage in a scalable, easy-to-use solution. This next-generation storage platform combines powerful and flexible hardware with advanced efficiency, management, and protection software to meet the demanding needs of today s enterprises. The VNX series is powered by Intel Xeon processors for intelligent storage that automatically and efficiently scales in performance, while ensuring data integrity and security. It is designed to meet the high performance, high-scalability requirements of midsize and large enterprises. Different software suites and packs are also available for the VNX series, which provide multiple features for enhanced protection and performance: FEATURE VALUE Next-generation unfied storage, optimized fr virtualized applications Capacity optimization features including compression, de-duplication, thin provisioning, and application centric copies High-availability, designed to deliver 5 9 s availability Automated tiering with FAST VP and FAST Cache that can be optimized for the highest system performance and lowest storage cost simultaneously Simplified management with EMC Unisphere for a single management interface for all NAS, SAN and replication needs. Up to three times improvement in performance with the latest Intel Xeon multi-core processor technology, optimized for flash X X X X X X EMC VSPEX Integration for Data Centers White Paper 8

9 SOFTWARE SUITES FAST Suite Automatically optimizes for the highest system performance and the lowest storage cost simultaneously. Local Protection Suite Practices safe data protection and repurposing. Remote Protection Suite Protects data against localized features, outages, and disasters. Application Protection Suite Automates application copies and proves compliance. Security and Compliance Suite Keeps data safe from changes, deletions, and malicious activity. SOFTWARE PACKS Total Efficiency Pack Includes all five software suites Total Protection Pack Includes local, remote, and application protection suites. VNX SNAPSHOTS VNX Snapshots is a new software feature introduced in VNX OE for Block Release 32, which creates point-in-time data copies. VNX Snapshots can be used for data backups, software development and testing, repurposing, data validation and local rapid restores. VNX Snapshots improves on the existing SnapView Snapshot functionality by integrating with storage pools. NOTE: LUN s created on physical RAID groups, also called RAID LUNs, support only SnapView Snapshots. This limitation exists because the VNX Snapshots require pool space as part of its technology. VNX Snapshots support 256 writeable snaps per pool LUN. It supports Branching, also called a Snap of a Snap, as long as the total number of snapshots for any primary LUN is less than 256, which is a hard limit. VNX Snapshots use redirect on write (ROW) technology. ROW redirects new writes destined for the primary LUN to a new location in the storage pool. Such an implementation is different from copy on first write (COFW) used in SnapView, which holds the rights to the primary LUN until the original data is copied to the reserved LUN pool to preserve a snapshot. This release (Block OE Release 32) introduces consistency groups (CGs). Several pool LUNs can be combined into a CG and snapped concurrently. When a snapshot of a CG is initiated, all writes to the member LUNs are held until snapshots have been created. Typically, CGs are used for LUNs that belong to the same application. VNX SNAPSURE VNX SnapSure is an EMC VNX Network Server software feature that enables you to create and manage checkpoints, which are point-in-time, logical images of a production file system (PFS). SnapSure uses a copy on first modify principle. A PFS consists of blocks. When a block within the PFS is modified, a copy containing the block s original contents is saved to a separate volume called the SavVol. Subsequent changes made to the same block in the PFS are not copied into the SavVol. The original blocks from the PFS in the SavVol and the unchanged PFS blocks remaining in the PFS are read by SnapSure according to a bitmap and block map data-tracking structure. These blocks combine to provide a complete point-in-time image called a checkpoint. A checkpoint reflects the state of a PFS at the time the checkpoint is created. SnapSure supports two types of checkpoints: Read-only checkpoint Read-only file system created from a PFS. Writeable checkpoint Read/write file system created from a read-only checkpoint. SnapSure can maintain a maximum of 96 read-only checkpoints and 16 writeable checkpoints per PFS while allowing PFS applications continued access to real-time data. VNX VIRTUAL PROVISIONING EMC VNX Virtual Provisioning enables organizations to reduce storage costs by increasing capacity utilization, simplifying storage management, and reducing application downtime. Virtual Provisioning also helps companies to reduce power and cooling requirements and reduce capital expenditures. Virtual Provisioning provides pool-based storage provisioning by implementing pool LUNs that can be either thin or thick. Thin EMC VSPEX Integration for Data Centers White Paper 9

10 LUNs provide on-demand storage that maximizes the utilization of your storage by allocating storage as needed. Thick LUNs provide high performance and predictable performance for your applications. Both types of LUNs benefit from the ease-of-use features of pool-based provisioning. Pools and pool LUNs are also the building blocks for advanced data services such as FAST VP, advanced snapshots, and compression. Pool LUNs also support a variety of additional features, such as LUN shrink, online expansion, and User Capacity Threshold setting. EMC VNS Virtual Provisioning allows you to expand the capacity of a storage pool from the Unisphere GUI after disks are physically attached to the system. VNX systems have the ability to rebalance allocated data elements across all member drives to use new drives after the pool is expanded. The rebalance function starts automatically and runs in the background after an expand action. Monitor the progress of a rebalance operation from the General tab of the Pool Properties window in Unisphere, as shown in the following figure. LUN EXPANSION Use Pool LUN expansion to increase the capacity of existing LUNs. It allows for provisioning larger capacity as business needs grow. The VNX family has the capability to expand a pool LUN without disrupting user access. Pool LUN expansion can be done with a few simple clicks and the expanded capacity is immediately available. However, you cannot expand a pool LUN if it is part of a data protection or LUN-migration workload. For example, snapshot LUNs or migrating LUNs cannot be expanded. For more detailed information of pool LUN expansion, refer to White Paper: EMC VNX Virtual Provisioning Applied Technology. LUN SHRINK Use LUN shrink of thin LUN to reduce the capacity of existing LUNs. VNX has the capability of shrinking a pool LUN. This capability is only available for LUNs served by Windows 2008 and later. The shrinking process has two steps: Shrink the file system from Windows Disk Management. Shrink the pool LUN using a command window and the DISKRAID utility. The utility is available through the VDS Provider, which is the part of the EMC solutions enablement package. The new LUN size appears as soon as the shrink process is complete. A background task reclaims the deleted or shrunk space and returns it to the storage pool. Once the task is complete, any other LUN in that pool can use the reclaimed space. For more detailed information on thin LUN expansion, refer to the White Paper: EMC VNX Provisioning Applied Technology. USER ALERTING THROUGH CAPACITY THRESHOLD SETTING Customers must configure proactive alerts when using file systems or storage pools based on thin pools. Monitor these resources so that storage is available to be provisioned when needed and capacity shortages can be avoided. EMC VSPEX Integration for Data Centers White Paper 10

11 The figure to the left shows thin pool provision monitoring. Monitor the following values for think pool utilization: Total Capacity is the total physical capacity available to all LUNs in the pool. Total allocation is the total physical capacity currently designed to all pool LUNs. Subscribed Capacity is the total ost reported capacity supported by the pool Over-subscrbed Capacity is the amount of user capacity configured for LUNs tha exceeds the physical capacity in a pool. Total allocation may never exceed the total capacity, but if it nears that point, add storage to the pools proactively before reaching a hard limit. POOL PROPERTIES DIALOG BOX IN UNISPHERE Below is the Storage Pool Properties dialog box in Unisphere, which displays parameters such as Free Capacity, Percent Full, Total Allocation, Total Subscription, Percent Subscribed and Oversubscribed by Capacity. When storage pool capacity becomes exhausted, any requests for additional space allocation on thin provisioned LUNs fail. Applications attempting to write data to these LUNs usually fail as well, and an outage is the likely result. To avoid this situation, monitor pool utilization, and alert when thresholds are reached. Set the Percentage Full Threshold to allow enough buffer to make remediation before an outage situation occurs. Adjust this setting by clicking on the Advanced tab of the Storage Pool Properties dialog box, as seen in the below figure. This alert is only active if there are one or more thin LUNs in the pool, because thin LUNs are the only way to oversubscribe a pool. If the pool only contains thick LUNs, the alert is not active as there is no risk of running out of space due to oversubscription. You also can specify the value for Percent Full Threshold, which equals Total Allocation/ Total Capacity, when a pool is created. EMC VSPEX Integration for Data Centers White Paper 11

12 View alerts by using the Alert tab in Unisphere. The below figure shows the Unisphere Event Monitor Wizard, where you can also select the option of receiving alerts through , a paging service, or an SNMP trap. View alerts by using the Alert tab in Unisphere. The below figure shows the Unisphere Event Monitor Wizard, where you can also select the option of receiving alerts through , a paging service, or an SNMP trap. EMC VSPEX Integration for Data Centers White Paper 12

13 The table below displays information about thresholds and their settings under VNX OE Block 32. THRESHOLD TYPE THRESHOLD RANGE THRESHOLD DEFAULT ALERT SEVERITY SIDE EFFECT User Settable 1% 84% 70% Warning None Built-In N/A 85% Critical Clears user settable alert Allowing total allocation to exceed 90 percent of total capacity puts you at the risk of running out of space and impacting all applications that use thin LUNs in the pool. VNX FAST CACHE VNX FAST Cache, a part of the VNX FAST Suite, enables Flash drives to function as an expanded cache layer for the array. FAST Cache is an array-wide, non-disruptive cached, available for both file and block storage. Frequently accessed data is copied to the FAST Cache in 64 KB increments and subsequent reads and/ or writes to the data chunk are serviced by FAST Cache. This enables immediate promotion of highly active data to flash drives. This dramatically improves the response time for the active data and reduces data hot spots that can occur with a LUN. The Fast Cache feature is an optional component of this solution. VNX FAST VP VNX FAST VP, a part of the VNX FAST Suite, can automatically tier data across multiple types of drives to use differences in performance and capacity. FAST VP is applied at the block storage pool level and automatically adjusts where data is stored based on how frequently it is accessed. Frequently accessed data is promoted to higher tiers of storage in 1GB increments, while infrequently accessed data can be migrated to a lower tier for cost efficiency. This rebalancing of 1GB data units, or slices, is part of a regularly scheduled maintenance operation. Network The network infrastructure provides high performance, fault tolerance and investment protection. The EMC VSPEX solution requires redundancy through the use of both multi-path topologies and fast recovery optimizations. For instance, both multiple paths and multiple switches are required within a VSPEX design to assure physically separate paths exist to maximize uptime. This validated solution uses virtual local area networks (VLANs) to segregate network traffic of various types to improve the throughput, manageability, application separation, highavailability, and security. For block, EMC unified storage platforms provide network highavailability or redundancy by two ports per Storage Processor (SP). If a link is lost on the SP front end port, the link fails over to another port. All network traffic is distributed across the active links. In the side topology, all links and paths are redundant. The Storage Processors each have a path to each primary and secondary switch. Extreme Networks switches have dual links between each other and those links are configured as a Link Aggregation Group. Finally, the ESXi hosts have redundant paths to both switches as well. VLAN s are configured for both redundant paths to ensure that a VLAN group is able to forward in the event of a physical link failure. Extreme Networks provides a family of switching products to meet the demands of the Data Center. The X670 is VSPEXcertified by EMC and has been validated in a VSPEX testlab by an independent third party. The 7100-series switch has also been validated in the same VSPEX testlab though doesn t carry the EMC certification. Other switches that share a common ExtremeXOS include X770, X480, X440, X430. Customers can choose between them depending on the requirements of the data center. For example, data centers may deploy Summit X670 if the requirements include support for latency-sensitive applications, or require co-deployment with the BlackDiamond X8 for the 100G capabilities or high density. Or data centers may deploy 7100 series if the requirements include deeper packet buffers, or require co-deployment with an S-series switch for the CoreFlow switching technology. EMC VSPEX Integration for Data Centers White Paper 13

14 Extreme Networks Summit X670 Series The Summit X670 series switches are purpose-built top-of-rack switches designed to support emerging 10 Gigabit Ethernetenabled servers in enterprise and cloud data centers. The Summit X670 helps optimize new server deployments with its optional, future-proofing 40 GbE uplink for 10 GbE-based servers to start the transition to the new virtualized environment. The Summit X670 series provides intelligent switching and routing with high-performance stacking technology and dedicated 40 Gigabit Ethernet uplink capabilities. The Summit X670, like other Extreme Networks switches, is powered by ExtremeXOS, a robust, modular OS. The Summit X670 offers 48-port 10 Gigabit Ethernet nonblocking switching with 10GBASE-X SFP+ interfaces. The Summit X670 is capable of Layer 2 and Layer 3 forwarding at up to 714 million packets per second in a compact 1RU form factor, enabling next-generation high-performance server deployment in data centers. The Summit X670V model can support an additional four QSFP+ ports of 40 GbE with the optional VIM4-40G4X module. With this option, you can maximize the number of interfaces for servers up to 48 ports while using the dedicated four-port 40 GbE module for uplink connectivity. Each 40 Gigabit Ethernet port can be independently configured as 40 Gigabit Ethernet or 4 x 10 Gigabit Ethernet; thus with the VIM4-40G4X module, Summit X670V can support up to 64 ports of 10 Gigabit Ethernet in a 1RU form factor and is capable of Layer 2 and Layer 3 forwarding at up to 952 million packets per second forwarding rate. This configuration provides 3:1 oversubscription from front ports to uplink ports. SUMMIT X670 SPECIFICATIONS High Performance Switching and Routing Up to 1.28Tbps switching performance with 48- or optional 64-port 10 Gigabit Ethernet and 4-port 40GbE (Summit X670V) all running at wire rate. High-performance SummitStack -V, SummitStack-V160, or SummitStack-V320 stacking support for virtualized chassis configuration supporting mix and match with Gigabit Ethernet switches Sub one microsecond low latency for Layer 2 switching and Layer 3 switching, and supports cut-through switching Green design with very low power consumption and high efficiency power supply Data Center Switching SFP+10 GbE ports with support dual speed 10 Gigabit Ethernet and Gigabit Ethernet providing smooth migration from Gigabit Ethernet to 10 GbE Software configurable 40 GbE QSFP+ uplink ports for native 40 GbE as well as 4 x 10 GbE to provide compatibility with non 40 GbE capable systems Data Center Bridging (DCB) support with Priority Flow Control (PFC), Enhanced Transmission Selection (ETS) and Data Center Bridging exchange (DCBX) Scalable Virtual Machine (VM) deployments with Direct Attach (VEPA) for inter-vm switching and XNV (ExtremeXOS Network Virtualization) for VM mobility Reversible air flow option to provide separation of cold and hot aisle in data center installation High Availability ExtremeXOS modular OS for highly available network operation Extends high availability across switches with Multi-Switch Link Aggregation (M-LAG) Carrier-grade packet ring protocol, Ethernet Automatic Protection Switching (EAPS) Internal redundant AC/DC power supply and 2+1 fan tray EMC VSPEX Integration for Data Centers White Paper 14

15 Comprehensive Security Robust MAC and IP security framework Threat detection and response with CLEAR-Flow security rules engine Motion sensor detection for physical security Extreme Networks 7100 Series The Extreme Networks 7100-Series is a family of high density, high-performance 10Gigabit Ethernet switches that are ideally suited for the demands of today s Enterprise data center. The 7100-Series 10 Gigabit Ethernet switches are available in the following configurations: ports 1/10Gb SFP+ with 4 10/40Gb QSFP+ ports ports 1/10Gb SFP+ with 4 10/40Gb QSFP+ ports 7148T - 48 ports 1/10GBASE-T with 4 10/40Gb QSFP+ ports 7124T - 24 ports 1/10GBASE-T with 4 10/40Gb QSFP+ ports 7184 SPECIFICATIONS High Performance Switching and Routing Up to 1.28Tbps switching performance with 48- or optional 64-port 10 Gigabit Ethernet and 4-port 40GbE all running at wire rate. Low latency for Layer 2 switching and Layer 3 switching Green design with very low power consumption and high efficiency power supply ExtremeXOS ExtremeXOS is designed to provide the robust performance and capabilities needed for everything from large cloud and data center applications to intelligent, converged enterprise edge networks. A modular architecture, deeply-integrated identity aware security, ready integration with VOIP systems as well as control and automation features for highly virtualized data centers are key features of ExtremeXOS. The ExtremeXOS network operating system consists of the following: hardware abstraction layer, a kernel, kernel-loadable modules, application modules and a configuration management module. HARDWARE ABSTRACTION LAYER The Hardware Abstraction Layer (HAL) decouples all the hardware/ ASIC control code from the underlying hardware through a message layer. A HAL allows an OS to interact with the hardware at a general level rather than at a detailed, hardware specific level. To use an analogy from the PC world, Windows 8, BSD, and Linux are OSs that include a HAL. These OSs have different subsystems for particular functions (e.g. sound and video) that run on a general hardware platform. The HAL makes it possible to add support for new devices and new ways of connecting devices to the computer, without modifying every application that uses the device. The same concept applies to the HAL in ExtremeXOS. The HAL allows ExtremeXOS to run on different Extreme Networks hardware switching platforms, CPUs, and ASICs. DATA CENTER SWITCHING SFP+10 GbE ports with support dual speed 10 Gigabit Ethernet and Gigabit Ethernet providing smooth migration from Gigabit Ethernet to 10 GbE Software configurable 40 GbE QSFP+ uplink ports for native 40 GbE as well as 4 x 10 GbE to provide compatibility with non 40 GbE capable systems Data Center Bridging (DCB) support with Priority Flow Control (PFC), Enhanced Transmission Selection (ETS) and Data Center Bridging exchange (DCBX) Reversible air flow option to provide separation of cold and hot aisle in data center installation HIGH AVAILABILITY Extends high availability across switches with Virtual Switching Bonding (VSB) Redundant modular power supplies and fan modules EMC VSPEX Integration for Data Centers White Paper 15

16 KERNEL-LOADABLE MODULES Kernel-loadable modules allow modular extensibility of the OS. The use of kernel loadable modules avoids direct modification of the ExtremeXOS kernel. When new functionality is required, a module can be loaded extending the ExtremeXOS kernel with new capabilities. A kernel-loadable module can include such things as VLAN mappings, QoS mappings, multicast cache handling and ACL processing, for example. APPLICATION MODULES ExtremeXOS application modules are individual applications that have a specific task. For example, services such as virtualization, Network Login, 802.1x, EAPS, ESRP, VRRP, SSH2, and protocols associated with IPv4 and IPv6 are individual ExtremeXOS applications that operate as separate processes. Using a PC analogy again, ExtremeXOS loadable application modules are similar to user applications such as Microsoft Excel, or Adobe Photoshop. Each PC application interfaces directly to the OS kernel and the user can load and unload either application without impacting the other. The same is true for ExtremeXOS application modules. Each ExtremeXOS loadable application module has a distinct function and operates independently. For example, one can dynamically load and unload an SSH2 module without impact to the core kernel and other ExtremeXOS application modules, for example BGP operations. Each ExtremeXOS application module runs in its own memory protected process so that a malfunction in a single application module does not impact the rest of the system. CONFIGURATION MANAGEMENT MODULES ExtremeXOS configuration management modules are responsible for interfacing with the users or management agents and relaying the configuration-related information to the ExtremeXOS applications (e.g. BGP, OSPFv3, etc.). ExtremeXOS configuration management includes the following major functionality: Multiple concurrent configuration access methods (Console CLI, XML, SNMP, etc.) Saving and retrieving configuration and statistics for the switching platform Management of multiple users and their associated security clearanceextremexos brings the latest operating system technology to the networking market and enables various features and benefits that are described in the next section. Extreme Networks High Availability The Highly Available Extreme Networks Ethernet infrastructure for the EMC VSPEX Proven Architecture is achieved with the resilient and modular Extreme Operating System, EXOS, and a redundant design philosophy to ensure and maximum uptime in the event of an unplanned failure event. MODULAR SOFTWARE ARCHITECTURE The Summit X670 series switches allow each of many applications such as Open Shortest Path First (OSPF) and Spanning Tree Protocol (STP) to run as separate OS processes that are protected from each other. This drives increased system integrity and inherently protects against DoS attacks. PROCESS MONITORING AND RESTART ExtremeXOS increases network availability using process monitoring and restart. Each independent OS process is monitored in real time. If a process becomes unresponsive or stops running, it can be automatically restarted. LOADABLE SOFTWARE MODULES The modular design of the ExtremeXOS OS allows the upgrading of individual software modules, should this be necessary, leading to higher availability in the network. High Availability Features SPANNING TREE AND RAPID SPANNING TREE PROTOCOLS Summit X670 supports Spanning Tree (802.1D), Per VLAN Spanning Tree (PVST+), Rapid Spanning Tree (802.1w) and Multiple Instances of Spanning Tree (802.1s) protocols for Layer 2 resiliency SOFTWARE-ENHANCED AVAILABILITY Software-enhanced availability allows users to remain connected to the network even if part of the network infrastructure is down. Summit X670 continuously checks for problems in the uplink connections using advanced Layer 3 protocols such as OSPF, VRRP and Extreme Standby Router Protocol (ESRP, supported in Layer 2 or Layer 3), and dynamically routes traffic around the problem. EQUAL COST MULTIPATH Equal Cost Multipath (ECMP) routing allows uplinks to be load balanced for performance and cost savings while also supporting redundant failover. If an uplink fails, traffic is automatically routed to the remaining uplinks and connectivity is maintained. LINK AGGREGATION (802.3AD) Link aggregation allows trunking of up to eight links on a single logical connection, for up to 80 Gbps of redundant bandwidth per logical connection. MULTI-SWITCH LAG (M-LAG) M-LAG can address bandwidth limitations and improve network resiliency, in part by routing network traffic around bottlenecks, reducing the risks of a single point of failure, and allowing load balancing across multiple switches. EMC VSPEX Integration for Data Centers White Paper 16

17 DATA CENTER BRIDGING Data Center Bridging (DCB) is a set of IEEE 802.1Q extensions to standard Ethernet, that provide an operational framework for unifying Local Area Networks (LAN), Storage Area Networks (SAN) and Inter-Process Communication (IPC) traffic between switches and endpoints onto a single transport layer. Data Center Bridging provides the lossless Ethernet experience required when developing an Ethernet storage network. This ensures high performance and storage networking resiliency for any Ethernet based, block-storage networking protocol. DATA CENTER BRIDGING EXCHANGE PROTOCOL The Data Center Bridging exchange (DCBX) protocol is used by Data Center Bridging (DCB) devices to exchange DCB configuration information with directly connected peers. In an ExtremeXOS enabled switch, the switch uses DCBX to advertise its DCB configuration to end stations. The end stations can then configure themselves to use the switch DCB services. If the peers do not support a common configuration for one or more features, the switch generates messages to alert network management. The switch does not accept configuration change requests from end stations. The DCBX protocol advertises the following types of information: DCBX version information, so that the peers can negotiate a common version to use. Enhanced Transmission Selection (ETS) information for QoS parameters such as bandwidth allocation per traffic class (802.1p COS), priority for each traffic class, and the algorithm used for servicing traffic classes. Priority-based Flow Control (PFC) information for managing flow control between peers. Application priority information for prioritizing traffic for special applications that do not map directly to an established traffic class. After you enable DCBX, the protocol collects most of the information to be advertised from other switch services such as QoS and PFC. The only DCBX feature that needs configuration is the application priority feature. DCBX uses the Link Layer Discovery Protocol (LLDP) (IEEE 802.1AB) to exchange attributes between two link peers. DCBX attributes are packaged into organizationally specific TLVs, which are different for the Baseline and IEEE 802.1Qaz versions. CUSTOM APPLICATION SUPPORT The DCBX custom application support feature allows you to prioritize and manage traffic flow through the switch based on the application type. This feature allows you to configure DCBX handling of the following applications: Fiber Channel Over Ethernet (FCoE) FCoE Initiation Protocol (FIP) Internet Small Computer System Interface (iscsi) And any application that can be defined by: Ethertype value Layer 4 port number TCP port number UDP port number When you configure a custom application, you define a priority number that applies to traffic related to that application. DCBX advertises this priority to end stations in an application TLV. End stations that support this feature use the priority number for communications with the switch. The\ priority number maps to an 802.1p value, which determines which QoS profile in the switch manages the application traffic. The software supports a maximum of 8 application configurations. Software Defined Networking (SDN) Platform Extreme Networks simple, fast and smart SDN platform tightly integrates with existing and multi-vendor hardware and software network environments, preserving customer investments and avoiding vendor lock-in. Backward compatibility is maintained with multi-vendor network infrastructure that conforms to the OpenFlow standard and other open APIs. Extreme Networks OpenDaylight-based API, software development kit (SDK) and developer community will enable customers to evolve the network to keep pace with emerging security, wireless, and converged SDN infrastructure. SIMPLE FAST AND SMART Simple: End-to-end network automation and orchestration for simplicity and opex reduction SDN networks are complex, but Extreme Networks makes them simpler and easier to manage. We do this by leveraging one unified operating and management system (EXOS) across all platforms as well as the SDN controller. This makes deployment across multiple platforms much simpler. We are also fully integrating our solution with OpenDaylight by developing plug-ins for our management tools (such as NetSight) and hardening the overall OpenDaylight offering for increased reliability. As well we are adding additional applications (NAC, Purview, wireless, session management, etc.) that extend the benefits of SDN beyond the basic functions offered by most providers. This allows us to bring Unified Network Management as well as network access control, network optimization and advanced application analytics to the same SDN solution. This is all managed through a simple, single-pane-of-glass GUI front-end for the whole network including the wireless edge, campus and data center. Finally, orchestration tools like OpenStack can be simply deployed through open APIs down to the network infrastructure. EMC VSPEX Integration for Data Centers White Paper 17

18 Fast: Speed and Flexibility in provisioning the network for any application The second benefit provided is both speed and flexibility when deploying applications and solutions. An effective SDN solution must by definition be programmable and integrative with nearly every other solution and application on the market. We are leveraging both open and standards-based protocols, application programming interfaces (APIs) and NFV, such as OpenDaylight and OpenFlow. We are also combining them with proven, industry-leading Extreme Networks solutions including EXOS and OneFabric Connect SDN, as well as many predefined applications that are already available and deployed. The result is that Extreme Networks customers have choices. They can choose to develop their own applications or take advantage of the entire SDN development community. This not only puts power in the hands of the customer to easily create and deploy new services, it also makes these deployments faster. An Extreme Networks customer can choose the exact application or solution that best meets their needs, budget and timeline without worrying about interoperability. Smart: Investment protection with orchestration supporting backward compatibility Trialing or deploying a new SDN application with an existing network is often a complex undertaking. Extreme Network s open and standards-based SDN solution allows our customers to easily trial and deploy new services regardless of who develops or provides them. Our broad SDN partner and developer ecosystem includes not only proven, tested and deployed integrations, but also opens the door to any new applications that are developed using the OpenDaylight standard. Few network operators have the luxury of a greenfield network and even fewer appreciate being locked-in to one vendor. Our integrative solutions work with your existing network in order to protect your investment. At the same time, our modular, standards-based architecture reduces uncertainty, variability, unpredictability and overall cost thus enabling choice and letting our customers select a truly best-in-breed solution. DATA CENTER MANAGER Extreme provides open webservice APIs through OneFabric Solution Architecture which enables organizations to benefit from centralized management and configuration of both wired and wireless networks, network-wide visibility and control of new applications and services, secure and granular access control to network resources, and a centralized programming API to automate and orchestrate network provisioning through other IT systems. Different network and infrastructure elements can t behave in silos, and neither can the administrators who manage them. Evolving business applications require a transparent, crossfunctional service provisioning process that bridges the divide among the server, networking and storage teams, ensuring each has an integrated view of virtual server and network environments. Coordinating the automated assignment of virtual machines (VMs) within both the virtual and physical network fabrics, the Data Center Manager (DCM) ensures that proper network resources are allocated when a VM is provisioned, no matter where it is on the network. DCM, a module leveraging the OneFabric Connect architecture, has a unique vendor-agnostic approach supporting a variety of virtualization, storage and server platforms, enabling the unification of the physical and virtual network and ensuring networks will have the high availability necessary for critical applications and business data. DCM provides a multi-user authentication framework that can authenticate and profile multiple VMs on a single physical ports, for different VLANs. Traffic can be untagged or already encapsulated using the 802.1q tagging. VLANs can be dynamically created and policies can be applied per VM. OPENFLOW OpenFlow is a technology that allows for external devices to directly program Extreme Networks switches for flow-based forwarding. This allows a more custom and granular level of control for end users, applications, and management systems. ExtremeXOS supports hybrid functionality, allowing both network programmability and flow-based forwarding with OpenFlow, as well as more commonly used and traditional CLI and NMS-based provisioning with classic Ethernet-based forwarding decisions. The hybrid mode control is given to end-users based on initial provisioning of the system itself. Extreme Networks switch ports are either classic Ethernet or are specifically defined via CLI to be included into the OpenFlow SDN domain for flow-based forwarding. Extreme Networks switches support hybrid mode on a per VLAN basis. A single port can support both OpenFlow controlled VLANS and VLANS with traditional networking services. Virtualization The virtualization layer is a key component of any server virtualization or private cloud solution. It decouples the application resource requirements from the underlying physical resources that serve them. This enables greater flexibility in the application layer by eliminating hardware downtime for maintenance, and the physical capability of the system to change without affecting the hosted applications. In a server virtualization or private cloud use case, it enables multiple independent virtual machines to share the same physical hardware, rather than being directly implemented on dedicated hardware. The virtualization layer enables the physical implementation of resources to be decoupled from the applications that use them. EMC VSPEX Integration for Data Centers White Paper 18

19 In other words, the application view of the available resources is no longer directly tied to the hardware. This enables many key features in the Private Cloud concept. VMware vsphere is the leading virtualization platform in the industry. For years, it has provided flexibility and cost savings to the end users by enabling the consolidation of large, inefficient server farms into nimble, reliable cloud infrastructures. The core VMware vsphere components are the VMware vsphere Hypervisor and the VMware vcenter Server for system management. The VMware hypervisor runs on a dedicated server and allows multiple operating systems to run on the system at one time as virtual machines. These hypervisor systems can be connected to operate in a clustered configuration. These clustered configurations are then managed as a larger resource pool through the vcenter product, and allow for dynamic allocation of CPU, memory and storage across the cluster. Features like vmotion, which allows a virtual machine to move between different servers with no disruption to the operating system, and Distributed Resource Scheduler (DRS) which performs vmotion automatically to balance load, make vsphere a solid business choice. With the release of vsphere 5.1, a VMware-virtualized environment can host virtual machines with up to 64 virtual CPUs and 1TB of virtual RAM. Traditional Architecture Virtual Architecture VMware vsphere 5.1 VMware vsphere 5.1 transforms the physical resources of a computer by virtualizing the CPU, RAM, hard disk, and network controller. This transformation creates fully functional virtual machines that run isolated and encapsulated operating systems and applications just like physical computers. The high-availability features of VMware vsphere 5.1 such as vmotion and Storage vmotion enable seamless migration of virtual machines and stored files from one vsphere server to another, or from one data storage area to another, with minimal or no performance impact. Coupled with vsphere DRS and Storage DRS, virtual machines have access to the appropriate resources at any point in time through load balancing of compute and storage resources. VMware vcenter VMware vcenter is a centralized management platform for the VMware Virtual Infrastructure. It provides administrators with a single interface for all aspects of monitoring, managing, and maintaining the virtual infrastructure, which can be accessed from multiple devices. VMware vcenter also manages some advanced features of the VMware virtual infrastructure such as VMware vsphere High Availability and Distributed Resource Scheduling (DRS), along with vmotion and Update Manager. VMWARE VSPHERE HIGH AVAILABILITY The VMware vsphere High Availability feature enables the virtualization layer to automatically restart virtual machines in various failure conditions. EMC VSPEX Integration for Data Centers White Paper 19

20 If the virtual machine operating system has an error, the virtual machine can be automatically restarted on the same hardware. If the physical hardware has an error, the impacted virtual machines can be automatically restarted on other servers in the cluster. Note In order to restart virtual machines on different hardware, the servers need to have available resources. Compute provides detailed information to enable this function. With VMware vsphere High Availability, you can configure policies to determine which machines are automatically restarted, and under what conditions these operations should be attempted. Compute The compute layer provides memory and processing resources for the virtualization layer software, and for the applications running in the private cloud. The VSPEX program defines the minimum amount of required compute layer resources, and enables the customer to implement the solution by using any server hardware that meets these requirements. VSPEX provides the flexibility to design and implement your choice of server components. The infrastructure must conform to the following attributes: Sufficient RAM, cores and memory to support the required number and types of virtual machines Sufficient network connections to enable redundant connectivity to the system switches Excess capacity to withstand a server failure and failover in the environment Use several identical, or, at least compatible, servers. VSPEX implements hypervisor level high-availability technologies, which may require similar instruction sets on the underlying physical hardware. By implementing VSPEX on identical server units, you can minimize compatibility problems in this area. If you implement high availability at the hypervisor layer, the largest virtual machine you can create is constrained by the smallest physical server in the environment. Implement the available high-availability features in the virtualization layer, and ensure that the compute layer has sufficient resources to accommodate at least single server failures. This enables the implementation of minimaldowntime upgrades, and tolerance for single unit failures. Within the boundaries of these recommendations and best practices, the compute layer for EMC VSPEX can be flexible to meet your specific needs. Ensure that there are sufficient processor cores and RAM per core to meet the needs of the target environment. Lenovo RD630 ThinkServer RD630 is a highly scalable, powerful 2U rack server solution ideal for business process or decision support deployments. The RD630 features sophisticated hardware RAID standard on every model providing your managed networks and user groups vault-safe data protection and lightning-fast data performance. With outstanding remote manageability, an array of network adapter choices, and rich I /O, the RD630 server handles heavy workloads and robust bandwidth with energy efficient design innovation. Additionally, the VSPEX compute platform should conform to the following best practices: EMC VSPEX Integration for Data Centers White Paper 20