Juniper Networks is delivering a new network that enables businesses and consumers to maximize the potential of their products and services and meets the infrastructure needs of the next decade.T: And many companies are already taking advantage of the potential the new network provides.
Cleanly separate networking software into four layers:Management, Services, Control, and ForwardingOptimizes each element of the networkCentralize appropriate parts of the Management, Services, & Control layersSimplifies network design and lowers operating costsUse the Cloud for elastic scale, usage-based pricing, and flexible deploymentReduces time to service and correlates cost based on valueCreate a platform for network applications and management integrationEnables new business solutionsStandardize protocols for interoperable, heterogeneous support across vendorsProvides choice and lowers costBroadly apply this approach to :Networking & Network Services including SecurityEnterprise & Service ProviderMobile & WirelineEdge, Datacenter, Access & Aggregation, Campus & Branch, WAN, and Core Domains
Centralize management extract services that were formerly on network devices but would now run on a common platform as virtual machines, leverage a centralized controller enabling a new broad range of network capabilitiessimplify the network device. Step 1:We have a 4-step approach to the path of SDN. Step 1: Centralize management. What does it mean when you centralize management?Historically, networking vendors have relied solely on a CLI or on SNMP agents for this capability, and have not excelled in this regard due to the inefficient nature of CLI and SNMP interfaces. The industry consequently is hungry for a platform that provides proper traffic monitoring, configuration management and useful network visibility and that results in reduced operational costs. Now you connect your DC orchestration systems to your single source of truth – management system. There is no longer any CLI or manual configuration. The configuration information is now brought in from the various devices into the single source of truth – mgmt plane. Administration, configuration is now automated and centrally managed. We propose that you use cloud techniques. It is an interesting paradigm truth in the center, devices no longer the master, management runs in a private cloud or virtualized environment.Step 2:Step 2: Extract the services. A lot of the key services are trapped within the network devices today. Different services require different amounts of processing speeds. Let me take for example a stateful firewall. A stateful firewall takes limited amount of processing power compared to IDP. When you turn on an IDP on top of the firewall, the capacity or throughput of the firewall goes down, because there is only so much computing power that can be allowed on the networking device. If you pull out these services like Stateful firewall, IDP and run them as VM on x86 hardware just like other apps or services are run, now you can elastically scale. You can dynamically scale up or down on these services. Step 3:Next step. Is where you centralize the controller. Control plane is the brains of the network. It communicates with other devices and learns the state of the network. It keeps track of the state of the network, Today, the control plane is very distributed and is in each device. We propose to centralize some of the aspects of the controller and have a local control plane copy to run distributed on the networking device. The centralized controller provides a birds eye view of the entire network topology, its state within a DC or campus or a geographical area and also giving you the ability to connect key services together.One of the most innovative and key components of SDN is what we have branded as SDN Service Chaining. It is this centralized SDN controller that enables this Service chaining. In a nutshell, SDN service chaining is about how you can extract services from network and security devices, run them as VMs and chain them together in a logical flow. The concept of service chaining – physical service chain is not new. You build these service chains all the time, but you do it physically.Taking a simple example, in DC edge, you will take a router, connect it to a firewall, you will have Ethernet cables connecting both, you then put another appliance or ADC box right next to it and again cable it together. Then, you would need to configure the router, the firewall, the ADC separately. That is a service chain, a physical service chain, This is a very cumbersome and very Opex intensive. Every time there is a change, in any of the elements within your physical service chain, you would need to re-configure the entire setup, test it before you deploy it. We think there is a much better way, less capex and opex intensive approach. We have what is called SDN service chainStep 4:So where does this leave us. The last step to SDN is Optimize Hardware. So where does hardware sit in this. Hardware will continue to play a key role. Specialized hardware will always exist in the SDN world and we will never be in a situation where everything will run outside the box. The forwarding plane will continue to be in the router or switch or security device and we will build innovative silicon ASICS which will provide the capacity and speed needed to deliver the packets. X86 will never be able to replace the innovation provided by our hardware. Like I mentioned, it makes sense to move some of the elements from the control plane, services into software. The services can run on general purpose hardware, but there are several functions like forwarding functions that can only be provided by the ASICS and silicon built within our hardware. We have 2 ASICs families today – TRIO and express. As we look into the future and see how we will innovate with our ASICs, we see we have a huge opportunity. We can provide tremendous optimization for our customers. So, what we have provided is a set of guiding principles and path to SDN we have established. In terms of the hardware, we see our customers adopting any one of the 4 deployment listed here.Customers will still choose to run everything in a vertical approach layered inside a box as it exists today. All of the network is self contained in a box and this is definitely a viable option and we will support this approach.Another option is to take a set o x86 servers and your networking devices and combine them. It certainly makes sense running management on the x86 servers rather than have it inside the box. Customers will want to run some or all of their key services outside on x86 boxes in the cloud with service chaining and tether it to the Mx or SRX. This is the JVAE approach and we will support that. It is a combination of using the hardware and getting the benefit of the running services in the cloud and getting the benefits of the capacity and speed you get with our ASICs.Last option is where everything moves into the cloud. A classic example of this is a MSP who wants to put a dumb CPE on customer premises and move al of the functionality into the cloud. This I think is less prevalent on the Enterprise side.So, these are the 4 steps to SDN.