Your SlideShare is downloading. ×
Cloud computing y UCS
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×

Introducing the official SlideShare app

Stunning, full-screen experience for iPhone and Android

Text the download link to your phone

Standard text messaging rates apply

Cloud computing y UCS

308
views

Published on

UTE Day - Presentación UCS - 14 de julio de 2011

UTE Day - Presentación UCS - 14 de julio de 2011

Published in: Technology, Business

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
308
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
40
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide
  • ~900 UCS customers since FCS June 2010  (based on Q3FY10 numbers) Especially related to cloud deployments- i.e. SAVVIS, SunGard, Telstra Nexus y/y Growth Rates disclosed Q3FY10 Earnings: Nexus 2000: Up 431% Nexus 5000: Up 315% Nexus 7000: Up 277%  UCS 168% Q/Q revenue Growth “ UCS gaining traction as 20% of CIOs have already evaluated the platform up from 10% in the Jan 2010 survey” -- Morgan Stanley, March 2010, CIO Survey reports
  • Q2 CY 09- NH EP 5500 Xeon Q1CY10- Westmere EP 5600 Xeon/NH EX Q3CY10- OOW/Vmworld Q4CY10- continuation Q1CY11- Westmere EP 5600 Refresh
  • Key points here: Simplified cabling Simplified architectural choices Scalability = 20GB/chassis or blade Balance = 40GB/chassis or blade Bandwidth = 80GB/chassis or blade Non-blocking, cut-through, lossless connectivity. (DCE, FCOE) From the perspective of the IT staff, the “California System” looks like a standard rack of servers. But the wiring of the components is much simpler and easier to manage, track and deploy. This approach creates a cleaner, simpler model for managing data center assets that stands in sharp contrast to antiquated PC architectures. As a result, IT organizations will find it easier to set and maintain hardware policies They will also be able to better physically secure the environment. The overall environment supports 63 percent more airflow than traditional servers, which leads to substantially less heating and cooling costs. Each component of the system can be managed by California Manager to the point where specific power and cooling thresholds can be set. The California System also provides easier access to disks residing on each server blade. Power supplies are hot-swapable. Each California System Should have two Fabric Extender to maximize availability. Taken all together, the California system greatly reduces all the points of management compared to traditional blade server environments.
  • Transcript : So in a nutshell, the pictorial representation of the unified computing system is this. California day one will be the ability for Cisco to deliver 320 compute nodes that are connected over a pair of upstream unified fabric switches that have embedded on them a single management domain for the compute element, the virtualization element and the networking element. So we will be able to deliver the customer stateless servers that have been custom-built using industry-standard x86 processors, industry-standard memory architectures and the ability to support virtualization functions, that virtualization functions day one have been optimized to run with VMware. But we're not stopping there. We're also collaborating with the likes of Microsoft on supporting their virtualization architecture. And the same we are doing with Oracle. The same we are doing with Red Hat from a Linux perspective.
  • Key points here: Simplified cabling Simplified architectural choices Scalability = 20GB/chassis or blade Balance = 40GB/chassis or blade Bandwidth = 80GB/chassis or blade Non-blocking, cut-through, lossless connectivity. (DCE, FCOE) From the perspective of the IT staff, the “California System” looks like a standard rack of servers. But the wiring of the components is much simpler and easier to manage, track and deploy. This approach creates a cleaner, simpler model for managing data center assets that stands in sharp contrast to antiquated PC architectures. As a result, IT organizations will find it easier to set and maintain hardware policies They will also be able to better physically secure the environment. The overall environment supports 63 percent more airflow than traditional servers, which leads to substantially less heating and cooling costs. Each component of the system can be managed by California Manager to the point where specific power and cooling thresholds can be set. The California System also provides easier access to disks residing on each server blade. Power supplies are hot-swapable. Each California System Should have two Fabric Extender to maximize availability. Taken all together, the California system greatly reduces all the points of management compared to traditional blade server environments.
  • When you take a step back, the ability to dynamically provision server, storage and networking assets is really the Holy Grail of enterprise computing It’s really the inability to dynamically provision these assets that creates all the inflexibility associated with enterprise IT today. But once you change the underlying architecture and, more importantly, the way it is managed, you create the opportunity to fundamentally change the way people think about enterprise computing. That in turn changes the way senior IT leaders can strategically think about IT and their role in the enterprise.
  • Transcript : So if you look at the picture on the left today, right. Once again not to pick on anybody, but we'll pick on HP. If you take a C-7000 Blade System, it looks something like this in terms of management, on the left hand side. So you have a Blade Manager, and you probably know this as iLO or iLO2, you have a Chassis Manager, which is called Onboard Administrator. You have a Connectivity Manager, which is called Virtual Connect Manager, and then you have a multi-chassis Connectivity Manager. Virtual Connect Enterprise Manager, that manages multiple domains of Virtual Connect. So you have four of these. HP does a pretty decent job of gluing these together with ICE or SIMM or whatever other glue they have. If you actually have a Systems Management product that you're deploying in your datacenter and you have to integrate with it, you will actually have to write to four different APIs. And it's a pain, believe me. With us, we have a single point of management. A single Embedded Device Manager with a single API. And all management is done through that. As a result, if you have an upstream Systems Management product that needs integration with us, all they need to do is write to a single API. And it's also XML, which is very easy to code to. As a result, so what does it all mean for your end customer? Reduced management costs. Believe it or not, deploying Systems Management Software is a time consuming and a costly process. If we can reduce costs there, it's huge cost savings. And the second one is, it's very easy to integrate with existing frameworks and that is really, really important. If we make it really hard to integrate with existing frameworks, it's going to be a problem trying to get a bigger footprint in the datacenter for us.
  • Contact slide author (M. Sean McGee; seanmcgee@cisco.com) for slide updates or explanation if needed.
  • The Unified Network Fabric sets the stage for automating large swaths of the enterprise computing experience. In truth, we have used IT to automate every process in existence except the process of IT. Cisco is creating a unified network fabric that substantially reduces the amount of time currently needed to maintain and manually configure systems In so doing, it creates the opportunity for IT organizations to concentrate more activities that add more value to the business as opposed to spending most of their time maintaining systems. And it has the potential to lift staff morale because it allows IT professionals to see themselves as something more than glorified digital maintenance workers.
  • Here’s what IT organizations deal with today Parallel LAN/SAN Infrastructure Inefficient use of Network Infrastructure 5+ connections per server – higher adapter and cabling costs Adds downstream port costs; cap-ex and op-ex Each connection adds additional points of failure in the fabric Longer lead time for server provisioning Multiple fault domains – complex diagnostics Management complexity – firmware, driver-patching, versioning
  • With these goals in mind, we’re now going to walk you through the individual components of the California system that Cisco has integrated to create a unified model for data center operations The components leverage industry standard hardware in combination with Cisco advanced management technology that is embedded across the entire system. Most importantly, Cisco is providing a unified approach to managing those assets that will increase collaboration across multiple IT disciplines while also working towards reducing the total cost of managing the data center.
  • Author’s Original Notes: Overall throughts – designed for enterprise “lights out” datacenter environments. Similarities to Nexus 5000 technology - but very different (details later). Dual PS - redundant Simple design – easy to install, replace Power Supply wattage – set the stage for next deck
  • Author’s Original Notes: 10G ports – for connecting Chassis’ Expansion Bays (cover the expansion cards shortly) Large Fans – lots of coverage on the rear of the system 2 Fabric Extenders for High Availability and redundancy Dual P
  • Notes should mention that the 9216i and the
  • UCS Blade Server Industry-standard architecture UCS Adapters Choice of multiple Converged Network Adapters and Virtual Adapters The “Unified Computing System” is based on a standard set of components that most IT staff are very familiar with. The intelligence for managing the overall system is based on a pentium-class processor that Cisco has embedded in the fabric interconnect The UCS Manager software that manages the entire system communicates with firmware embedded in every device in the system. It is important to note that there are three adapters. The first one is a standard 10GB Ethernet adapter and the second is a 10GB Fiber over Ethernet adapter. The most important one is the third one, which is referred to as Palo. That adapter support the virtualization of the network connections, which will be discussed more in depth later in the presentation.
  • Cisco offers a number of adapter options to give IT organizations a choice between standard configurations and new state-of-the-art adapters that support “virtual” connections. The Palo adapter ultimately serves optimize I/O performance in virtual environments. The single biggest stumbling block in the mainstream adoption of virtualization in production environments has been concerns about I/O performance. The ability to provide granular control over I/O performance across any number of virtual links mitigates those performance concerns. We will have three different adapter families and four different adapters. Cost: Intel Oplin based adapter, supports FCOE software driver. There is currently an Open Source FCOE driver available. Compatibility – 2 Adapters: Standard CAN architecture modified to fit our mezzanine form factor. Based on Emulex or Qlogic FC ASIC (LP11xx and QL2642). Ethernet is standard Intel 10GE Oplin Don't require specific CNA drivers. Palo – Virtualized Adapter Explained in detail in later slide We don't require the whole California to be one kind of adapter. They can be different on the double-width blade, but it isn’t recommended.
  • Transcript

    • 1. Cisco UCS2011Carlos Nicasiocarlos.nicasio@la.logicalis.com
    • 2. Contents- UCS Market Positioning- UCS Architecture2 Cisco Unified Computing System
    • 3. UCS Market Positioning
    • 4. Cisco UCS Results • 5400 UCS customers at the end of 2010 • 400%+ Growth year over year • 30 World-breaking benchmarksCisco Unified Computing System
    • 5. Clips from the “The Worst Predictions inHistory” Video Cisco Unified Computing System http://www.youtube.com/watch?v=O0oIffAjuqI&feature=youtu.be
    • 6. Market Share Data #3 vendor in x86 blade servers WW with 10.5% revenue share #3 vendor in total blade servers WW with 9.4% revenue share #2 vendor in US x86 blades with 19.7% revenue share Tied for #2 in North America x86 blades with 18.8% revenue share IDC – Press Release: http://www.idc.com/getdoc.jsp? containerId=prUS22841411 Customer facing slideshare: http://www.cisco.com/web/solutions/data_center/ucs _marketshare.html http://www.slideshare.net/MeredithSabye/ucs-impact- of-innovation http://www.flickr.com/photos/cisco_pics/5328636375 /in/set-72157626675863887/ Cisco Unified Computing System
    • 7. WW UCS momentum is fueledX86 Server Blade Market Share, Q1 CY 111 by game-changing innovation Cisco is quickly passing UCS #3 with established players in 10.5% fastest growing segment of x86 computing market 2 UCS After Two US Short Years UCS #2 with 19.7% 5400 UCS Customers WW Cisco Unified Computing System Source: 1 IDC Worldwide Quarterly Server Tracker, Q1 2011, May 2011 2 IDC Q4 CY10 Server Forecaster, 2010-2015 CAGR of x86 Blade Servers
    • 8. WW X86 Server Blade Market Share Customer adoption of UCS is changing the server industry landscape Cisco growth is out-pacing the market Market appetite for Innovation UCS #3 and fuels UCS climbing growth Customers have shifted over 10% of the global x86 blade server market to Cisco and nearly 20% in the USDemand for data center innovation has vaulted Cisco Unified Computing System (UCS) to the #3 leader in the fast growing segment of the x86 server market Cisco Unified Computing System Source: IDC Worldwide Quarterly Server Tracker, Q1 2011, May 2011
    • 9.  Cisco Unified Computing System (UCS) has moved the industry forward by unifying compute, network, storage access and virtualization into one cohesive system Cisco data center customers report tangible business results due to transformative improvements in IT efficiency and agility UCS is designed to solve key customer challenges in the data center  Manual solution assembly  Inflexible infrastructure  Operational friction  Virtualization complexity  Inefficient scaling  Compliance and audit control Cisco Unified Computing System
    • 10. # 1 SPECint_rate200 6 Cisco UCS C460 M2 SPECjbb2005 SPECOMPL VMmark 2.1 X86/64 2- base2001 socket B200 M2 4-socket C2460 M1 C460 M2 SPECOMPL SPECOMPM SPECjbb2005 VMmark 2.1 SPECompLbas base2001 2- base2001 4- X86/64 2-socket 2 -socket-Blade e2001 socket B200 M2 socket C2460 M1 B230 M1 B200 M2 C460M2 SPECOMPM SPECjAPPServ SPECint_rate_base SPECompMba LS-Dyna 4- 2006 X86/64 2- base2001 2- er 2004 2-node se2001 socket C460M1 B230 M1 socket B200 M2 socket B200 M2 C460 M2 SPECfp_rate_bas SPECjbb2005 Oracle E-business SPECjEnterprise SPECompLbas Suite Medium e 2006 X86/64 2- X86/64 4-socket Model Order to 2010 Overall e2001 socket B200 M2 C460 M1 Cash B200M2 B440 M1 B230 M2 SPECfp_rate_bas Oracle E-business SPECOMPL SPECompMbas SPECint_rate_base Suite Medium 2006 X86/64 2- e 2006 X86/64 4- Model payroll Batch base2001 2- e2001 socket B200 M2 socket C460 M1 socket B200 M2 B230M2 B200M2 SPECfp_rate_bas SPECjAPPServer SPECint_rate_bas Oracle E-business SPECOMPM SPECijbb2005 e2006 X86/64 2- Suite Ex-large Model base2001 2004 single node e 2006 X86/64 4- payroll Batch C260 M2 Socket B200 M1 2-socket C250 M2 socket C460 M1 2-socket B200 M2 B200M2 2-socket VMmark VMmark 1- Oracle E-business SPECint_rate_ SPECint_rate_bas VMmark 1.1 Suite Medium e2006 X86/64 2- VMmark B250 Overall Blade C460 2 -socket-Blade Model payroll base2006 Socket B200 M1 M2 C460 M1 M1 B230 M1 C260 M2 Batch B200M2 2-socket server LInPack VMmark 1- SPECjbb2005 VMmark 2.0 SPECfp_rate_b 2-socket VMmark B200 M1 VMmark B200 2-socket Blade B440 X86/64 2-socket Overall ase2006 recapture B200 M2 M1 B230 M1 B200 M2 C260 M2 Cisco Unified Computing System Q2 CY09 Q1 CY10 Q3 CY10 Q4 CY10 Q4 CY10 Q2 CY11http://www.cisco.com/en/US/prod/ps10265/industry_benchmarks.html#~application_performance_reports
    • 11. “Fabric computing is a fixture on the radar screen of many IT groups, driven by the increased penetration of virtualization and prospects for cloud computing.” —Gartner Fabric computing has emerged as the preferred infrastructure for data center virtualization and cloud computing, and Cisco is the market leader in this industry transition Cisco Unified Computing SystemGartner report: Fabric Computing Poised as a Preferred Infrastructure for Virtualization and CloudComputing, February 11, 2011, George J. Weiss and Andrew Butler Report. ID number:G00210438.
    • 12. Which vendor would you perceive to be the most competent to deliver on a fabric–based strategy in your enterprise? Cisco Dell Egener a HP IBM VMwar e Other Don’t Know/Not Sure 0 5 10 15 20 25 30 35 40 45 % of RespondentsGartner report: Fabric Computing Poised as a Preferred Infrastructure for Virtualization andCloud Computing, February 11, 2011, George J. Weiss and Andrew Butler Report. IDnumber: G00210438. Cisco Unified Computing SystemYou can read the full Gartner report here: http://www.gartner.com/technology/media-
    • 13. “After several years of being a highly consolidated market where the top 3 vendor accounted for over 80% of blade revenue, the recent entry of Cisco has introduced a viable new competitor to the market.” — IDC: Jed Scaramella IDC Worldwide Quarterly Server Tracker Press Release, May 24, 2011“The Cisco Unified Computing System is a high-end, high-density, highly scalable, awesomely powerful network, compute, virtualization, and management backbone that re-architects the notion of the blade chassis ….” — Windows IT Pro Tech Ed 2011 Best of Show, May 2011“According to VARs, Cisco’s UCS is scaring the heck out of all of Cisco’s data center rivals, even if they put on a good face in public and scoff at UCS viability.” — Computer Reseller News, December 15, 2010“Its a paradigm shift in datacenter infrastructure whose time has come.” —InfoWorld Technology of the Year Award, June 2010 Cisco Unified Computing System
    • 14. The market has Choosing Cisco UCS is a affirmed that as a trusted proven, reliableCisco has truly partner for platform for changed the computing is enterprise and game and is proving to be cloud leading an the right path computing; industry for many as the rapid UCS transition that data center market traction was long evolves is driven by overdue! customers seeking better solutions to IT challenges Cisco Unified Computing System
    • 15. UCS Architecture
    • 16. The legacy blade solution is a Chassis Level “Mini-Rack” Design • Limited or no Unified I/O • More hardware to manage • Larger energy footprint Multi-Chassis Manager Multi-Chassis Manager • More management software and licenses Interconnect Managers Interconnect Managers • Proprietary interconnects (VC) Chassis Manager Chassis Manager • More NICs and HBAs per server Blade Manager (iLO) Blade Manager (iLO) • No network policy engine for VMs Multiple Network, Fabric Modules Multiple Network, Fabric Modules • No network QoS • OS based agents and CMS required Multiple NICs, HBAs Multiple NICs, HBAs • Clustered CMS required for mgmt HA Complex and fragile Life Cycle mgmt Complex and fragile Life Cycle mgmt • Must buy switches and mgmt devices for every chassis. • Complexity and cost is amplified as you(3 Chassis, Rear scaleView) Cisco Unified Computing System
    • 17. Simple Building Block Building Block: 2 x FI + chassis Uplinks2 x Unified Fabric to Existing LAN/SAN Interconnects (Embedded mgmt,access layer for multiple UCS chassis) BLADE CHASSIS 2 x Fabric REAR VIEW Extenders (I/O MUX, CMC) 6U chassis (one or more) BLADE CHASSIS FRONT VIEW Stateless compute blades w/ CNA, opt. HDDsCisco Unified Computing System
    • 18. Unified Computing System Architecture Single Management Point for EntireUnified Fabric Domain - Scale Up to 320 servers - Large memory blade - VN-Link - Stateless compute engines - VIC - QoS - Single high BW network access for all chassis - Unified fabric within domain Industry Standard Components, open XML API Cisco Unified Computing System
    • 19. Cisco Unified Computing System LAN SAN A Mgmt Any IEEE Compliant LAN SAN B Any ANSI T11 Compliant SAN Any ANSI T11 Compliant SAN One Logical Chassis to Manage* LAN Connectivity SAN Networking Blade Chassis’ Server Blades Rack Servers Server Identity Management Monitoring, Troubleshooting etc. *architectural limit of Cisco Unified Computing System 320 servers with 160 servers supported as of 1.4(1)
    • 20. UCS External Connectivity LAN CLOUDNAS STORAGE FC Storage LAN Switch LAN Switch SAN Switch SAN Switch UCS Fabric Interconnect Access Layer LAN & SAN Access Layer LAN & SAN Unified Fabric (FCoE) Unified Fabric (FCoE) Cisco Unified Computing System
    • 21. Cisco UCS:A single, logical, expandable blade server chassis Cisco Unified Computing System
    • 22. System Interconnect Choices (rear) Uplinks FEX Scalability Balance Bandwidth (Up to 40 chassis; (Up to 20 chassis; (Up to 10 chassis; 20 Gbps) 40 Gbps) 80 Gbps)Cisco Unified Computing System
    • 23. UCS Components and RelationshipsUCS ManagerManagement resides in the Fabric InterconnectUCS 6100 - Fabric InterconnectUnified access layer interconnect and management20 or 40 ports and option uplink modulesUCS 2104 – I/O ModuleInserts into Blade ChassisCMC, port aggregator, I/O MUX, extension of FI n ai UCS 5108 – Blade Chassis om D Blade inserts into the Chassis t Blades are a logical par of the chassis en Up to 40 chassis per environment emUCS Blade Server n agIndustry Standard components a M2s, 4s Intel Nehalem, Westmere, eDDR3 RAM, SAS, SSD n gl Si UCS Mezzanine Adapters VIC, Menlo (Q & E), Oplin Cisco Unified Computing System
    • 24. UCS Manager:Simplified Compute Managementand Operations Cisco Unified Computing System
    • 25. Simplified Management Minimize errors, reduce risk, lower cost, minimize complexityLegacy Blade Architecture Cisco UCS Solution Single Embedded Device Manager Single Embedded Device Manager Mgmt ServerMgmt Server+Plug-ins Multi-Chassis Multi-Chassis Manager Manager Multiple switch Multiple switch Managers Managers Chassis Chassis Manager Manager Blade Manager Blade Manager (iLO) (iLO) Multiple Network, Multiple Network, Fabric Modules Fabric Modules Multiple Multiple NICs, HBAs NICs, HBAs Complex and fragile Complex and fragile Life Cycle mgmt • Embedded device manager for family of UCS components Life Cycle mgmt • Enables stateless computing via Service Profiles • Efficient scale: Same effort for 1 to 320 blades OS based agents OS based agents per blade per blade • Clustered Fabric Interconnects for HA Cisco Unified Computing System
    • 26. Real Customer Results Agility is the Biggest Value in Virtualization The draw to virtualization is saving money, but after a year or so, surveys show that adopters believe the key value is agility. Thomas Bittman Gartner UCS amplifies the value of agility by virtualizing the compute hardware 100%!Cisco Unified Computing System
    • 27. Compute as a Service Software & Hardware Virtualization VM VM VM VM VM Software-Based Virtualization HYPERVISOR HYPERVISOR (OS and application layer virtualized) Hardware State Cisco Hardware State Virtualized  Cisco UCS virtualizes the Server Hardware State 100%  Hypervisor (or OS) is unaware of underlying hardware state abstractionCisco Unified Computing System
    • 28. Service Profile: Virtual Server Hardware  Policy Driven Virtualized Server Hardware FW,boot device, MAC, WWN, vLan, vSAN, UUID, and QoS managed through policies, profiles, templates  Dynamic and Consistent Provisioning Databas e Easily deploy in minutes, not days or weeks Rapid HW deploy, repair, change = maximum agility ESX RBAC, multiple levels of administration Consistent server builds = minimized risk & errors WWW Service Profile: DataBase Network1: DB_vlan1 Service Profile: ESX-Host Network1 QoS: Platinum DataBas Network1: esx_prod MAC : 08:00:69:02:01:FC e Network1 QoS: Gold Boot Order: SAN, LAN MAC : 08:00:69:11:19:EQ FW: DataBaseSanBundle WWN: Boot Order: SAN, LAN FW: ESXHostBundle Service Profile: WebServer Network1: www_prod Network1 QoS: Gold MAC : 08:00:69:10:78:ED Boot Order: LOCAL FW: WebServerBundle Cisco Unified Computing System
    • 29. Stateless Computing: UCS Service Profiles Configurable HW State Examples •RAID settings •Disk scrub actions •Number of vHBAs •HBA WWN assignmentsServer hardware state •FC Boot Parametersis fully configurable •HBA firmwarePreserved in softwareknown as a Service •FC Fabric assignments for HBAsProfile. •QoS settingsService Profiles can •Border port assignment per vNIC •NIC Transmit/Receive Rate SANthen be dynamically Limitingassigned to specificblade hardware forruntime. •VLAN assignments for NICs •VLAN tagging config for NICs •Number of vNICs LAN •PXE settings •NIC firmware •Advanced feature settings •Remote KVM IP settings •Call Home behavior •Remote KVM firmware •Server UUID •Serial over LAN settings •Boot order •IPMI settings •BIOS scrub actions Cisco Unified Computing System •BIOS firmware •BIOS Settings
    • 30. Server Identity Management Comparison Multi Chassis Access Layer Unified FabricFC Chassis ModulesEnet Unified FabricFC AdaptersEnet Unified Fabric Server Blades Cisco Unified Computing System
    • 31. Unified Computing System The Right Solution at the Right Time Next Gen Market Direction: Legacy (HP, IBM, Dell) UCS Management and Control Primary SAN A Networ k SAN B Secondar y Network Server = Application Server = ResourceInefficient Complex High Cost Fragile Efficient Agile Transformative Cisco Unified Computing System
    • 32. DifferentiatingTechnologiesEmbedded in UCS Cisco Unified Computing System
    • 33. I/O Consolidation – Unified Fabric in UCS (Universal I/O, Mixed Workloads, Wire Once, minimal access points) Legacy Server Access Connectivity I/O Consolidation with FCoE LAN SAN A SAN B LAN SAN A SAN B Ethernet FC FC Traditional Traditional FC EthernetEthernet FC FCoE (at server access layer where most (Nth Server) savings for I/O consolidation resides) Cisco Unified Computing System
    • 34. VN-Link: VM Level Network Transparency Problems: VMotion • VMotion may move VMs across physical ports—policy must follow • Impossible to view or apply policy to internally switched traffic • Cannot correlate traffic on physical links—from multiple VLAN 101 VMs VN-Link (Problems Solved): •Extends network to the VM •Consistent services •Coordinated, coherent management Cisco Unified Computing System
    • 35. Virtual Interface Card (VIC)• True wire once architecture – highly dynamic• Network policy and visibility to VM (VN-Link in hardware)• Hypervisor bypass support – increases performance• Reduce NIC and HBA cards (get up to 58 virtual PCI devices) Switch Legacy Server With Virtualization Adapter Hypervisor Hypervisor Soft Switch Virtual Virtual Virtual Virtual Virtual Virtual Machine Machine Machine Machine Machine Machine Cisco Unified Computing System 35
    • 36. Optimizing Memory for Intel EP processors Typical Memory Cisco UCS Memory Each DIMM Fixed the CPU number of sees Xeon 5500 DIMMs Xeon 5500 is made of 4 can be standard addressed DIMMs by the CPU Typical System Either Intel Xeon 5500/5600 CPUs • 12 DIMMs @ • Max 384GB per Blade 1066MHz - 1333 mhz in all configurations • Max 96GB in M2 models! Or • 18 DIMMs @ Benefit 800MHz •4x capacity • Max 144GB •Lower costs at lower •Standards DIMMs, CPUs, OS performance Cisco Unified Computing System
    • 37. QoS – Key Element for Multi-Tenancy SA N LA N MGMT SA N  QoS parameters can be GG G G S S G G G G configured at a per system Fabric FabricAA Interconnect Interconnect Fabric Fabric A A Interconnect Interconnect class level, or a per vNIC level. G G GG G G GG G G G G  All traffic belongs to 1 of 6 System Compute Chassis Compute Chassis Fabric Fabric R I C C I R Fabric Fabric Classes - Four are user configurable while Extender R R Extender x8 x8 x8 x8 Extender Extender the other two are for FCoE and standard Ethernet  No packet drops - Priority Flow MM B PP B PP Control (PFC) uses per priority pause Adapter Adapter Adapter Adapter Adapter Adapter mechanism to guarantee no frames are X X X X X X dropped in lossless priorities. x86 Computer x86 Computer x86 Computer x86 Computer  Segregation of resources into pools and organizations with Compute Blade Compute Blade (Half slot) (Half slot) Compute Blade Compute Blade (Full slot) (Full slot) their own policies  Allows bandwidth allocation forNo other compute vendors have different classes of traffic anything like this… Cisco Unified Computing System
    • 38. Industry Standard ServerComponents Cisco Unified Computing System
    • 39. Enclosure, Interconnect, & Blades (Front)Redundant, Hot Swap Power Redundant, Hot SwapSupply Fan 1U or 2U Fabric Interconnect Half width server blade Up to eight per enclosure Hot Swap SAS drive (Optional) Full width server blade Up to four per enclosure Mix blade types 6U Enclosure Cisco Unified Computing System Ejector Handles Redundant, Hot Swap Power Supply
    • 40. Rear View of Enclosure and Interconnect 10GigE Ports Expansion Bay (server or upstream GigE (for uplinks to networks existing GigE or FC networks) Redundant Hot Swap Fan Module Redundant Fabric Extender Fan Handle Cisco Unified Computing System
    • 41. UCS 6100 Series Fabric Interconnects Fabric FabricInterconnects Interconnects 20-Port L2 Interconnect 40-Port L2 Interconnect • 20 fixed ports 10GE/FCoE, fixed • 40 fixed ports 10GE/FCoE, fixed • 1 Expansion Module • 2 Expansion Modules Expansion Expansion Modules Modules FC + Ethernet Ethernet Fibre Channel • 4 Ports 10GbE/FCoE • 6 Ports 10GE/FCoE • 8 Ports 1/2/4G FC • 4 Ports 1/2/4G FC • 6 ports 1/2/4/8G FC Cisco Unified Computing System
    • 42. Compute Options B200 M2 2 Socket Intel 5600, 2 SFF Disk, 12 DIMM B250 M2 2 Socket Intel 5600, 2 SFF Disk, 48 DIMM B230 M2 2 Socket Intel E7-2800, 2 SSD (7MM), 32 DIMM B440 M2 4 Socket Intel E7-4800, 4 SFF Disk, 32 DIMM C200 M2 2 Socket Intel 5600, 4 Disks, 12 DIMM, 2 PCIe 1U C210 M2 2 Socket Intel 5600, 16 Disks, 12 DIMM, 5 PCIe 2U C250 M2 2 Socket Intel 5600, 8 Disks, 48 DIMM, 5 PCIe 2U C260 M2 2 Socket Intel E7-2800, 16 Disks, 64 DIMM, 6 PCIe 2U C460 M2 4 Socket Intel E7-4800, 12 Disks, 64 DIMM, 10 PCIe 4U Cisco Unified Computing System
    • 43. Adapter Offerings Ethernet Only Standard CNA VIC 10 Gbe Ethernet Existing Driver VM I/O Stacks Virtualization and Consolidation Emulex 10GbE/FCoE 10 GbFCoE Eth FC FC Eth Intel Niantic User Broadcom Definable 57711 (iSCSI vNICs 10GbE FC offload) 0 1 2 3 127 PCIe Bus PCIe x16 Cisco Unified Computing System
    • 44. Thank youCarlos Nicasiocarlos.nicasio@la.logicalis.com