Transcription
Composable Infrastructure:Treat Your Infrastructure as Code
AgendaEvolving Infrastructure Models Richard Fichera, VP and principal analyst, Forrester ResearchCisco’s View of Composable Infrastructure Jim Leach, Director of Platform Strategy, Cisco Todd Brannon, Director of UCS Marketing, CiscoRoundtable Discussion Richard, Jim and Todd
Evolving Infrastructure ModelsForrester ResearchRichard FicheraVP & Principal AnalystJanuary 21, 2016
At the Heart of the Problem: Complexity,Scale and Shifting Architectural Models 2016 Forrester Research, Inc. Reproduction Prohibited4
Growth: Inexorable Simple, Ubiquitous, PainfulYou’ve seen this before – this is our worldThis is not new – what is new is the number of decimal places.And it is more than just volume, it is also underlying complexity, forcinga change in the way we engineer our infrastructure, incorporating newrelationships between data and processing 2016 Forrester Research, Inc. Reproduction Prohibited5
Explosive Growth Is Changing Storage and ProcessingRequirementsVELOCITY, EFFICIENCY & SCALABILITY ARE KEY CHALLENGES, AND MATCHING STORAGETO CPU BECOMES A CHALLENGEOur view: Data growth combined with thedesire for infrastructure flexibility are forcingenterprise and midsized companies to rethinktheir both their data management and theirnetwork architectures. Storing more data for every businessoperation,. Performance requirements continue toincrease Storage to compute ratios are highlyvariableSTORAGE MUST BE FASTER, SMARTER ANDEASIER TO PROVISION! 2016 Forrester Research, Inc. Reproduction Prohibited6
But Architectural Diversity Drives Storage andProcessing RequirementsCPU10OLTP86Network - L4MemoryMemCache2Search0AnalyticsNetwork -ThruDiskLatency and throughput are stronglylinked to network characteristics 2016 Forrester Research, Inc. Reproduction ProhibitedVMInfrastructure7
The Architectural Dilemma - Cheap or Optimized?Budgetpressuresforce mostof us hereButincreasinglywe want tobe hereLower TCOOptimalMuch of the I&O innovation in the past few years has been focused on trying to“cheat” this seemingly inviolable rule of operations and infrastructure engineering 2016 Forrester Research, Inc. Reproduction Prohibited8
New problems spawned new platformsDensity actionRack systemsBlade SystemsExternal Array IntegrationSoftware-Defined StorageComposable Systems 2016 Forrester Research, Inc. Reproduction Prohibited9
The Reality Today is Most Organizations Have aComplex Set of InfrastructureEnterprise rackValue yOptimized/HPCBladeTower 2016 Forrester Research, Inc. Reproduction Prohibited10
Configuration Complexity Costs› Increased complexity in systems management› Complexity in vendor management – from both sides. Configuration diversitymakes forecasting, inventory management less precise, leading either torequirements to carry extra inventory, or delays in acquisition and provisioning.› Pushing the inventory issue off onto a distributor or the vendor provides illusorysavings – they are smart enough to build it into their costs› Acquisition delays can impact productivity and time to solution – this is a bigloss› Cost and management issues can force us into sub-optimal configurations 2016 Forrester Research, Inc. Reproduction Prohibited11
Can we cheat?Lower TCONowYouCan HaveBOTHOptimalThe new wave of composable systems gives us the chance to have both highlyefficient and near-optimal systems in our infrastructure 2016 Forrester Research, Inc. Reproduction Prohibited12
Composable Systems – The Future of EnterpriseInfrastructureDefinition – A composable system allows individual servers to be “composed” frompools of CPU/memory, disks and shared network connections.Pooled Disaggregated ElementsMagicA “composed” server 2016 Forrester Research, Inc. Reproduction Prohibited13
Essential Characteristics of Composable SystemsUsable Composition Requires More Than Just A Box of Components› The composed systems must be indistinguishable from a standard physicalserver with the same configuration (CPU, memory, disk and network)› Composition must be independent of and have no effect on other composedsystems in the same management domain Systems must be capable of being “decomposed” and resources returned to pools› Management functions must be accessible via API as well as CLI/GUI› Subject to physical limitations imposed by packaging, elements must be as locationindependent as possible 2016 Forrester Research, Inc. Reproduction Prohibited14
Closing ThoughtsComposable Systems Represent the Missing Link in Software-Defined Data Centers› Previous approaches to the implementation of SDDC have fallen short due tolack of a flexible physical layer At some point the fluidity of the VM layer ran into the inflexibility of the physical server layer› Composable systems will allow a flexible coupling of the virtual layer with amore fluid physical layer› Vendor differentiation will be along several dimensions Number and selection of compute nodes and size of shared storage pools Network flexibility, configuration and integration Integration with higher levels of the SDDC, including OS, application and complete serviceimages 2016 Forrester Research, Inc. Reproduction Prohibited15
A Final Thought – Simple and Uniform Building BlocksDo not Mean Mediocre Outcomes 2016 Forrester Research, Inc. Reproduction Prohibited16
Cisco’s View ofComposable Infrastructure
Cisco Composable Infrastructure yMemoryLocal StorageLocal StorageLocal StorageNetwork I/ONetwork I/ONetwork I/OSAN I/OSAN I/OPhysical, Fixed RatioMonolithicNexusUCSLocal StorageNetwork I/OSAN I/OComposableInfrastructureM-Series
Composable Makes the Resources Fluid Composable systems allow flexible coupling of virtual layer with amore fluid physical layer Supports VMs, containers and bare metal applicationsBare MetalOperating SystemOperating SystemX86 ArchitectureMemoryNICDiskSSDCPUVirtualization LayerIntelligent Automation SoftwareX86 ArchitectureResource PoolMemoryNICDiskSSDContainerContainer cation
UCS Management: “Infrastructure as Code” Cisco innovation treats hardware as code Software object model: hardware in not configured manually API centricity: a unified system control plane Virtual Interface card (VIC): all network and SAN adapters are software definedXML APIUCS ManagementResource Pool
Service Profiles Optimize Resources Determine which resources to allocate Compose infrastructure in a set that isoptimized for the workloadProfile AProfile BUCS ManagementABAResource PoolBAB
UCS M-Series: Composable InfrastructureShared Local ResourcesPCIeCompute CartridgesUCS M Series
UCS M-Series: Composable InfrastructureShared Local ResourcesImproved utilization of resourcesShared Local ResourcesBased on Cisco System Link TechnologyPCIeEnables disaggregated, composable infrastructureCompute CartridgesModular DesignImproved subsystem lifecycle managementLean ComponentryCost and power optimizationUCS M Series
UCS Management and M-SeriesSystem Link connects disaggregatedresource elementsUCS ManagementCreate and modifyservice profilesModel abstracts ofrequired resourcesResources allocatedfrom ID poolFabric Interconnectprovides networkconnectivity andmanagementNetworkResourcesPower &CoolingResourcesStorageResourcesComputeResources
UCS M-Series: FlexibilitySupports multiple types of cartridges and different processor generations Preserves investment in systemSimplified management using service profiles
For more information aboutComposable InfrastructureIncluding videos and papers from other analystsGo towww.cisco.com/go/composable
Thank Youwww.cisco.com/composable
Cisco’s View of Composable Infrastructure Jim Leach, Director of Platform Strategy, Cisco Todd Brannon, Director of UCS Marketing, Cisco Roundtable Discussion Richard, Jim and Todd Agenda . Evolving Infrastructure Models Forr
