Introducing compelling services--and driving technical advancements to deliver those services--is a way of life in the telecom and networking world. Over the past 20 years alone, the networking industry has launched many exciting services, such as IPTV, Video on Demand, IMS voice and video telephony, 4G services and of course, cloud based applications like Netflix and WhatsApp.
The business models that drove these new services, however, were quite different from each other. During IPTV, IMS and 4G introductions, carriers had direct business relationships with paying subscribers. To ensure a great user experience, carriers carefully managed all aspects of the service delivery – from end user devices such as set top boxes/CPEs, to optimized application servers to optimized metro networks, BNGs and mobile core gateways.
In the cloud era, however, the business dynamics have changed. For cloud-delivered applications such as Netflix, Skype and WhatsApp, carriers must ensure a seamless service experience with little control over the application itself, and with minimal business relationship with the end customer. Enterprise markets have followed very similar trends, where enterprise applications such as large databases and ERP systems have moved from server rooms to private DCs in COLOs and ultimately to public clouds.
To succeed in this changed business landscape, networks must be as agile and efficient as the cloud applications themselves. SPs and enterprises must adopt modern network architectures that incorporate cloud design principles.
Architectural Simplicity is Critical to Success in Cloud Era
The fundamental business objectives of the cloud era are service agility and enhanced user experience with service customization, all while controlling costs. Modern network architectures achieve these business objectives with the following architectural simplification pillars.
CLEAN SEPARATION BETWEEN NETWORK LAYERS
The most important architectural enhancement that improves network simplicity and service agility is better decoupling of network services from underlying network transport itself. Such separation enables different network layers to evolve independently. Operators should be able to utilize all available transport options – MPLS, SPRING, or IP Fabrics – and allow service nodes to negotiate separate service tunneling with each service peer based on the capabilities of that peer. Such clean separation not only simplifies new service rollouts, but also enables operators to utilize 3rd party networks - even the Internet - to improve their market reach.
The separation of different network layers extends even further to include controllers. Controllers enable path computation and traffic engineering across different network domains and across different transport technologies like MPLS and SPRING, thereby simplifying service rollouts and accommodating increasingly disperse application workloads running on the network.
Importantly, this separation of controller, service and transport layers also allows operators to select best in class products at each layer.
AGILE SERVICES DEPLOYMENT BY SERVICE FOOTPRINT REDUCTION
For better service agility, localizing the service configurations to only devices where a customer attaches to the network is critical. The operator needs to configure a service only where a customer is present, thereby minimizing service footprint, thus simplifying service provisioning.
Operators have efficiently localized service footprint for L3 workloads for many years. However, L2 services remained beholden to hop-by-hop bridging, which led to VLAN stacking, VPLS with full mesh O(n2) pseudo-wires, and H-VPLS technologies, thereby increasing L2 service provisioning complexity. Many times operators had to configure multiple devices along the service path to connect a customer. Modern EVPN based L2 service architectures have brought in major simplification for service provisioning, since now L2 services provisioning is limited only to network edges that connect customers and all other nodes remain pure transport. EVPN based L2 services also allow fine grained customization and hence are driving emergence of new agile cloud connect products serving a growing market need for agile public cloud on-ramps to enterprises.
ADOPTING SIMPLE ‘SCALE-OUT’ ARCHITECTURES
Modern cloud grade routing architectures improve network economics by increasing network utilization and service availability. They offer end-to-end entropy friendly traffic load balancing - from multi-homed service edges to much simpler ECMP friendly SPRING and IP fabric cores. Traffic load balancing across all available paths improves network utilization and simplifies network capacity planning by easy scale out, without requiring traffic re-engineering. Additionally, multi-pathing architectures improve service availability and reduce failure domains since traffic can reroute to alternate path within milliseconds of a failure. Even better, multi-pathing architectures improve capital efficiency and network economics by allowing operators to run their networks ‘hotter,’ without compromising service SLAs.
Routing Technologies That Evolve Your Network for the Cloud Era
Juniper has made significant investments in innovative open standards based routing technologies that deliver new cloud era network architectures, in and across all networks domains – from metro to edge to core to DCs. These technologies include
Transport Layer (Underlay):
IP Fabric protocols (VxLAN, MPLSoUDP, IPoUDP): The latest IP fabric innovations allow operators to offer all services – L3VPNs, L2VPNs, Internet – over simple IP transport, dramatically simplifying network underlay and allowing service delivery over 3rd party network including Internet. Additionally, operators can encrypt IP fabrics to offer a secure transport service.
SPRING (aka Segment Routing): SPRING simplifies the transport layer by eliminating the need for running MPLS protocols. Additionally, with help of a controller, SPRING allows each service to determine its own path (using label stack) through the network as per its SLA requirements without maintaining any state in the network.
Routing in Fat Trees (RIFT): RIFT is poised to revolutionize DC architectures by cleverly combining link state and distance vector protocols to allow 90%+ CLOS utilization and taking into account available link bandwidths, without relying on ECMP or complex traffic engineering solutions. RIFT is another example of Juniper bringing optimized, purpose built best in class solutions to market, rather than tenuously stretching existing protocols to fit the needs of DC CLOS architectures.
Application (Service) Layer:
EVPN: EVPN revolutionized L2 services by liberating them from 30 year old bridging-based technologies by using the control plane to distribute MAC/IP reachability. By breaking dependency on hop-by-hop bridging, EVPN decouples L2 services from the transport technologies. Additionally, EVPN’s all-active multi-pathing improves service scale and availability, on top of any transport technology available.
NorthStar SPRING Controller: Northstar determines the optimal path for an application based on monitored network operating parameters, and encodes that path as label stack on the service routers. With embedded Junos routing protocols, Northstar is always in complete sync with the network it controls. Most importantly, with a programmable interface, Northstar allows operators to make customized path selection decisions based on business logic.
Ensuring Business Continuity through Network Transformation
Ultimately, the great advantage of cloud grade networking is architectural simplicity that improves service agility and efficiency. With Juniper, deploying IP fabrics, EVPN, SPRING, RIFT and the Northstar Controller complement current network operations and architectures, and provide a graceful network transformation to modern, cloud era architectures. Juniper brings decades of deployment experience to the table to help you make the best choices for your unique requirements and profit from the transformation.