So what do hear from customers about their care-abouts for IT. Notice that we are not talking about just about Data center here.Customers are faced with increasing video, mobile devices, data deluge and increasing use cloud technologies.The care-abouts are for the IT infrasturcutre per se and not just networking.If you break it down to basic needs, there are 3 elements that is required of IT#1 is Agility - This requirements comes down how soon can I enable a service in their IT and how fast can I adopt to new market transitions. #2 is Simplication – give that I want agility, I want my IT to be simpler to deploy, maintain and operate so that I can get TCO and get to the agility goal as well#3 is Business value – At the end of IT is enabling the customer’s business. So IT should be able to help the customer differentiate themselves against competition and also be able to monetize new opportunities.
Cisco has a quite diverse set of customers and has been networking leader for the past almost 30 years. We have networking leadership in the campus, branch, data center, traditional and cloud service providers. As we asked around these varied set of customers on what they really want, the answer varied between the deloployment areas (whether it is campus versus data center) but if they are enterprise versus service providersIn campus/branch the use cases are around centralization of management of wired/wireless networks as well as ability to provide NAC or QoS for the end-users. So our Merakiacquision plays very well into requirement as well as recently annouced Catalyst 3850. In Enterprise Data centers, requirement is more about how do I get set up a self-service private cloud for dev/test or internal customers in a automated fashion across multiple hypervisor. Increasing customers are looking at Hyper-V as an alternate hypervisor in their enterprise DC. If you talk to our large search engine/social media type of customers, it is about ability to program directly to the devices or through agents in the devices – whether it is chef/puppet or OMI agents. Our cloud service provide is all about scale, how do I get thoustands of tenants quickly and in a automated fashion either via openstack or other custom portals. They need able to compete against public cloud provider like Amazon EC2.Overall trend in IT is towards simplicity, automation and manageability.
Transcript:\r\ntime for the last couple of days, and certainly the last couple of hours, the differentiation of our strategy with respect to programmable interfaces and orchestration in the network is based upon the fact that we're doing and focusing as much on extracting information out of the network as we are programming new state and new configuration into the network. So towards this end, as we're extracting information the type of information we're talking about is load, is the actual state of the topology. And pulling this out, analyzing it, and being able to program a policy-- could be spin up a new TELSP, find a new protect path, find a path that meets an SLA. Or, if you're in a cloud environment, the service that I'm offering-- whether it's a web server, whether it's a firewall, or whatever the case might be-- dependent upon the load of that particular service, cause a reaction such that I can spin up more of those particular services with elasticity. And so these are some of the different demos that you're going to see this afternoon.\r\n\r\nAuthor's Original Notes:Cisco’s view of where we need to focus, and how we should be building our products is based on three very simple, but powerful tenets.First, the network has been a significant investment over the last decade This network contains significant intelligence:learned dynamically – watching traffic patterns, error rates, congestion issues, routing decisions, etcStatically configured by customer’s network engineersIt knows what is connected, what is being accessed (internally and externally)It is one of the most telemetry rich resources in your businessHarvesting this intelligence is something from which we can benefit greatly, especially the temporal and dynamic dataSecond, data is meaningless without the ability to analyze the data, correlate it to meaningful/actionable resultsThe foundations of dynamic security, dynamic resource utilization, and overall orchestration lie in the ability to see patterns (old and new) and act on themThink of this as Business Intelligence (BI) for the networkThird is PolicyWhat your business rules are for acting on dataThe priorities and governance of your corporate resourceIn many cases compliance regulations, etc, etcPolicy is the rules of your business which determine how you respond to information, to users, to events, etcAll of this culminates in the ability to take action – Programming the networkYou don’t want this to be about packet paths – you want this to be about business optimizationYour network needs to provide the ability to take a rich set of actions in response to the wide range of events, policies, and business rules on which you run your businessYou will see Cisco is taking a comprehensive approach, which is what I am here to discuss with you today
Let us look at how industry is approaching SDN or in general network programmabilityToday you have a tightly coupled control and data plane but is using distributed control plane protocols to run your network. This works fairly well (atleast that is how the world is running today) . The way you access these devices is through network management platforms, CLI, Netflow etc. When I talk to customers, some are pretty happy with this approach and has built tools around it.Now there is emerging set of use cases which requires more programmatic approach to managing the network.One of the first is individual box level programmatic API. This is basically opening up the aPI to the huge install base of products that already exist. Advantage of this approach is that you are can write applications or programs that does not need to know the CLI differences for each device at hand. Now the boxes continue to have control and data plane tightly coupled and run distributed protocols. You are configuring/modifying/managing the deviecs through this API calls.Second approach is what industry knows as classic SDN – you take the control plan function and centralize it and use protocol to talk to the data plane. Most often this is openflow but that is not necessarily the case all the time. Applications sit on top of the controller and use the controller north bound API to talk to the controller. Today there is no standard for these APIs.Variation of second approach is the hybrid SDN model where we still have the control and data plane together at device level but use a controller to program these devices as well. Some example of this is things like I2RS, PCEP or openflow.Final model is about network overlays, this could be done as a pure software overlay or a fully integrated approach. The goal of this approach is keep the phsyical layer of network separate from the overlay networks. It may benefit in terms of simplicity but unless you have integration to some degree you may not really get consistent app performance between one day and next as the overlay network is aware of of the underlay network and also troubleshooting could become harder co-ordinating between the two separate networks.
So this brings to the summary view of our strategy which I believe is the most comprehensive approach in the industry, spanning our entire portfolio of hardware/software and network/compute products.Our goal is to address programmability requirements in campus, branch, wan, data center, cloud, SP deploymentsHarvest Network Intelligence through deep “full-duplex” programmatic access to Cisco devices and softwareRich-set of APIs, development kits across IOS and NX-OSTie-in into analytics, policy engines and orchestrationExplore emerging technologies through joint partnerships with industry, academia and standards-bodiesEarly stage SDN/OpenFlow controllers and agentsOpenStack for cloud computingDeliver consistent operational experience for different use-cases Overlay network virtualization solutions, video, IaaS etc. Strong synergy between network and compute environmentsTraining, developer portals and advanced services Cisco’s Open Network Environment envisions a customizable framework to harness the entire value of the intelligent network delivering openness, programmability and abstraction across multiple layers in an evolutionary manner. It offers a choice of protocols, industry standards, use-case based deployment models and integration experiences while laying the foundation for a dynamic feedback loop of user, session or application analytics through policy programming.Benefits:Flexibility to use APIs, agents and controllersConsistency across both physical and virtual environmentsComplements traditional definitions of SDN (which focuses primarily on de-coupling control and data planes) -Encompass entire portfolio stack from transport to management and automation
1. © 2010 Cisco and/or its affiliates. All rights reserved. Cisco Confidential 1© 2010 Cisco and/or its affiliates. All rights reserved. Cisco Confidential 1Leading the WayCisco Open Network EnvironmentCisco’s Framework for Open, Programmable and App Aware NetworksBalaji SivasubramanianCo-Sponsored by Intel®
2. © 2010 Cisco and/or its affiliates. All rights reserved. Cisco Confidential 2IT InfrastructureAgilitySimplificationBusinessValueCloud Video Mobility Data Deluge• Rapid ServiceEnablement• Ability to Adopt NewMarket Transitions• Automation• ReduceOperationalComplexity• Better TCO• Differentiate /Monetize• Be a Enablerof Business
3. © 2010 Cisco and/or its affiliates. All rights reserved. Cisco Confidential 3EnterpriseCampus/Branch Centralizedmanagement ofwired/wirelessnetwork with userNAC/QoSMassively ScalableData Center Customize withprogrammaticAPIs to providedeep insight intonetwork trafficService Providers Policy-basedcontrol andanalytics tooptimize andmonetizeservice deliveryCloud Rapid scalableautomatedprovisioning oftenants viaOpenStack etcScalableMulti-TenancyNetwork FlowManagementCentralizedManagement andUser Access/QoSAgile ServiceDeliveryPrivate CloudAutomationEnterprise DC Multi-Hypervisor,orchestration ofsecurity profilesand services andself-serviceportalsDiverse Programmability Requirements Across SegmentsMost Requirements are for Simplicity, Automation & Manageability
4. © 2010 Cisco and/or its affiliates. All rights reserved. Cisco Confidential 4Looking beyond Day 0 Automated ProvisioningPolicy AnalyticsOrchestrationProgram forOptimizedExperienceHarvest NetworkIntelligenceNetworkNetworkCisco Approach: Build Highly Visible and Responsive End-to-End Networks
6. © 2010 Cisco and/or its affiliates. All rights reserved. Cisco Confidential 6Industry’s broadest approach for Network ProgrammabilityCisco Approach: Comprehensive—The Power of ―AND‖Approach 3AppsNetworkNetwork OverlaysNetworks OverlaysPhysicalandVirtualApproach 1AppsAPIsNetworkTightly-coupled HW & SWApproach 2AppsControllerOpenFlowDeviceTraditionalDevice withagents(Hybrid)Specific ProtocolDeviceAgentsLoosely-coupled HW & SWAgents
7. © 2010 Cisco and/or its affiliates. All rights reserved. Cisco Confidential 7aHardware + Software Physical + Virtual Network + ComputeNetworkPlatformAPIsControllers andAgentsNetworkOverlaysApplications:Cisco, Customer , ISV, Open SourceonePK – Comprehensive APIand developer programming kit Controller and Agents(e.g. OF,I2RS,PCEP, Chef,Puppet etc )IntegratedPhysical/VirtualNetwork and ServicesCampus, Branch, Data Center, Cloud
8. © 2010 Cisco and/or its affiliates. All rights reserved. Cisco Confidential 8• Twitter: @ciscoDC• Facebook.com/CiscoDC• Video: http://www.youtubecisco.com/datacenter• Cisco blog: http://blogs.cisco.com/datacenter• Slideshare: http://slideshare.com/CiscoDataCenterIn Collaboration with Intel®Intel and the Intel logo are trademarks of Intel Corporation in the U.S. and/or other countries.
9. © 2010 Cisco and/or its affiliates. All rights reserved. Cisco Confidential 9© 2010 Cisco and/or its affiliates. All rights reserved. Cisco Confidential 9Thank You