Their backbone network will have to be simple to operate, agile to cater to unexpected changes in the business climate, and be high performing to meet new demands.
That means overcoming four key challenges facing many enterprises today:
First, the current legacy network uses numerous protocols and cannot scale up to new demands quickly. How can the architecture be simplified to be ready for the future?
Second, the current network focuses on stability and reliability but lacks sufficient agility. How can a backbone network deliver rapid service provisioning and fault rectification within seconds?
Third, the number of leased lines is fixed in a network topology, and the types of leased lines are limited. How can you precisely expand bandwidth using on-demand leased lines while also reducing costs?
Fourth, the network is prone to manual errors and misconfigurations, along with low operations and maintenance (O&M) efficiency. How can a new network add service intelligence and improve O&M efficiency?
The answer to these challenges is clear -- a stable, agile, simplified, and cost-effective backbone network. This calls for a transformation that will cut through various bottlenecks today, by unifying protocols, adjusting the network to cloud changes, using leased lines with flexible contracts, and adopting a new SRv6 (Segment Routing over IPv6 dataplane) policy.
Unifying protocols to simplify a network
In the past, the evolution from IP to MPLS (Multiprotocol Label Switching) had addressed various issues, such as those prevalent in VPN and TE. However, with MPLS introduced to the data plane, along with various protocols such as LDP, RSVP-TE, PCEP, and NETCONF being introduced to the control and management planes, interoperability can become a challenge.
Protocol unification means using SRv6 to replace MPLS on the data plane. On the control and management planes, this solution eliminates the use of protocols such as LDP, RSVP-TE, PCEP, and NETCONF, and uses EVPN to achieve VPN protocol unification.
It also extends IS-IS to achieve IGP unification and extends BGP to achieve EGP and control protocol unification. This way, the protocol unification solution delivers a simplified and efficient network along with enhanced scalability and interoperability.
This was one key change in a top tier bank's move to a next-generation backbone network. By using SRv6 instead of MPLS, its network reduced the number of required network protocols from more than 10 to only two. This simplified network configurations as the bank did not need to maintain the MPLS forwarding plane and control protocols. More importantly, SRv6 inherently supports SDN with functions encoded in the Function field, achieving flexible network programmability.
Adjusting your network to changes in the cloud
In the cloud era, an enterprise's infrastructure network is comprised of the intra-cloud interconnection network, cloud-to-cloud interconnection network, and cloud access network. The backbone network interconnects clouds and aggregates cloud access traffic to give full play to the value of cloud and deliver cloud services with good user experience. Agility is a core benefit of the cloud, but the current backbone network lacks sufficient agility as it is built more on stability.
The solution to this calls for the backbone node deployment to be centred on cloud data centres. It prioritises cloud-to-cloud interconnection traffic while also factoring in cloud access traffic aggregation. It focuses on stability for the underlay network and agility for the overlay network to logically decouple stability from agility, thereby providing a unifying framework that addresses the stability-agility trade-off and improving the value of both the network and cloud.
The bank's next-generation backbone network adopts a forwarding-control separation architecture to ensure high reliability of the entire network. Meanwhile, the control layer uses cluster-based controller deployment for high reliability and the forwarding layer uses a dual-plane architecture and various fast detection and switching technologies, such as FRR and BFD for millisecond-level convergence in fault scenarios.
In this way, a fault-triggered service switchover can be completed within 50 Ms without being detected by users. Moreover, SRv6 can work with In-situ Flow Information Telemetry (iFIT) to implement real-time visualisation of service quality and in-band measurement, shortening the fault locating time to minutes and improving service reliability.
Lowering leased line costs with flexible line leasing
Over the years, the backbone network has witnessed an evolution of transport technologies from X.25, frame relay (FR), Asynchronous Transfer Mode (ATM), and Packet over SONET/SDH (POS) to optical transport network (OTN), multi-service transmission platform (MSTP), and MPLS VPN.
Currently, OTN is the mainstream transport technology for a backbone network, supplemented by MSTP and MPLS VPN. As data centre deployments shift from employing, for example, three data centres in two cities to using multiple data centres in multiple cities, the number of core nodes on the backbone network is also constantly increasing. This drives up the costs of full-mesh interconnection to an unacceptable level.
A flexible line leasing solution allows OTN, MSTP, and MPLS VPN to be used separately or in combination, according to service requirements on latency, packet loss, and reliability.
In terms of bandwidth expansion, this solution applies traffic collection and big data technologies to continuously optimise the traffic forecast model in an iterative way, so that it can accurately forecast the growth trend of application traffic to provide decision-making support for business departments.
In terms of network resource allocation, this solution adopts network slice+DSCP-based QoS, instead of traditional VPN+EXP-based QoS. This allows network resources to be hard isolated and guaranteed, and provides more service levels (DSCP-based QoS provides 64 service levels, while EXP-based QoS provides 8 service levels).
In terms of network plane construction, this solution adopts a multi-plane architecture (three or four planes) at the physical layer and a single-plane architecture at the logical layer. Traditionally, the physical and logical layers each correspond to one plane. This solution allows a multi-plane physical network featuring a partial full-mesh topology to be constructed at the initial stage, based on service traffic requirements. In addition, it places backbone network link resources on all planes into a logical single-plane resource pool and uses intelligent path computation to maximise backbone network link utilisation.
A comparison between the traditional dual-plane architecture and the SDN three-plane architecture shows that the maximum average utilisation of lines is about 45% in the dual-plane architecture (if one line is faulty, the utilisation of the other line is 90%) and around 60% in the three-plane architecture (if one line is faulty, the utilisation of the other two lines is 90%). As such, the three-plane architecture delivers a 15% improvement in line utilisation compared with the dual-plane architecture.
With fine-grained, high-frequency detection at the application channel level, another top-10 bank in China can now detect line quality changes and service quality status, as well as automatically switch network paths in an agile manner. For example, the bank uses three physical backbone network links between the headquarters and each branch, while many other banks use only two such links (active/standby). Although the use of two lines is generally considered more economical, using three lines provides support for more scheduling policies, better redundancy protection, and finer-grained control. Furthermore, it improves bandwidth utilisation, lowers the requirements on total bandwidth, and facilitates fast traffic optimisation.
Implementing SRv6 Policy and boosting O&M Efficiency
Due to the high costs of bandwidth resource leasing, government and enterprise customers have urgently demanded traffic engineering on backbone networks. To address this demand, RSVP-TE and then SR-TE were introduced. While these traffic engineering technologies achieve load balancing, they also complicate network O&M because it is difficult for O&M personnel to maintain a large number of tunnels.
A SRv6 Policy solution decouples resource invoking from tunnels by shielding the internal implementation of tunnels and encapsulating all network functions into SRv6 Policies. Moreover, it can be used with the SRv6-BE solution to effectively reduce the number of required SRv6 Policies.
As an important enabler of the network programming era, the SRv6 Policy solution helps implement network programming from endpoints to the access network, backbone network, data centre, and applications, enabling applications to schedule network resources and allowing networks to understand applications, thereby achieving intent-driven networks.
The Chinese bank's intelligent network management centre, which functions as its intelligent scheduling brain, obtains real-time path information using telemetry and other methods. By tapping into big data covering historical traffic, service requirements, and topology/tunnel information, this centre uses AI technology to conduct continuous optimisation and training, gradually freeing network management personnel from laborious manual operations. This helps build an intent-driven intelligent backbone network, and greatly improves O&M efficiency.
Backbone Network Transformation Is Now
With rapid developments in 5G, IoT, cloud-edge-device collaboration, distributed architecture, and AI, enterprise are discovering that their backbone networks will be increasingly important. These networks will have to be transformed through protocol unification, network adjustment upon cloud changes, flexible line leasing, and SRv6 Policy.
At this moment in time, relevant standards and technologies are already mature, the required ecosystem has been built, and relevant practices have started. Government and enterprise customers can seize the opportunities of backbone network transformation to construct stable, agile, simplified, and cost-effective backbone networks, thereby maximizing the service value of these networks.
As a promoter and pioneer in the SRv6 field, Huawei will implement continuous technical innovation to help partners accelerate digital transformation and service development.
From September 23 to 26, 2020, Huawei will hold the fifth Huawei Connect in Shanghai. For this year's event, Huawei is joining hands with global thought leaders, business elites, technical experts, pioneering enterprises, ecosystem partners, application service providers, and developers, to discuss the key trends in industrial digital transformation. Huawei will demonstrate cutting-edge technologies, products, and solutions in the ICT field as well as explore successful transformation practices, building an open and mutually beneficial industry ecosystem to create new value together with industrial customers.