Difficulties in deployment of carrier SD-WAN, difficulties in sd-wan deployment

Difficulties in deployment of carrier SD-WAN, difficulties in sd-wan deployment

I. Introduction

Software defined wide area network (SD-WAN) is an important branch of software defined network (SDN). It is a hot field of SDN technology application. The main goal of SD-WAN is to achieve the network quality of leased line, reduce traffic cost and improve bandwidth utilization through a series of technical means such as software technology, application strategy deployment and overlay network; at the same time, multiple sites can automatically select the optimal path based on the current network conditions and configuration policies to achieve load balancing and ensure the network quality across the network.

SD-WAN is a kind of SDN application scenario, the related technology is still evolving, mature, many practitioners in China have put forward their own SD-WAN implementation scheme, for operators, pipeline is "Business killer", how to introduce SD-WAN, and make SD-WAN become the killer stone is the key research content of operators in recent years.

Through the Deployment Analysis of Google SDN, this paper summarizes the key elements of the successful deployment of its SD-WAN, and draws on the carrier network to help operators find the starting point of SD-WAN deployment.

2. Google Case Analysis

Google's network is divided into the internal network and WAN of the data center. The WAN consists of two Backbone Networks Based on the traffic direction, which are the interconnected network (G-scale network) between the data centers) an I-scale network (I-scale network) for Internet users to connect Google's data centers around the world. It is an internal network.

Google optimized the interconnection network between data centers for SDN transformation, and controlled the Border network devices of each data center through applications (software) to ensure the arrival of the highest-priority application traffic with the lowest latency, low-level traffic is filled with idle link bandwidth to make full use of its high submarine cable.

Google B4 Network

Compared with many other SDN practitioners, Google has achieved more results mainly due to the following:

1. simple and clear Network Structure

SDN-based transformation is available. Google has more data centers, but it is always limited. In addition, the WAN network is simpler than the operator's business. In fact, it is an internal Controllable network with a single device, which facilitates network transformation. Through the transformation of the control plane and forwarding plane, the network is more intelligent, the routing is more reasonable, and the network quality is better.

2. Powerful R & D capabilities

Google is an integrated company for development and operation. Its powerful R & D capabilities, or Google's own industrial ecosystem, are sufficient to support the creation of a huge cloud network integration system starting from the underlying hardware and software.

3. Clear business traffic flow direction and CoS level planning

In the traditional Internet era, Google strictly grades its internal data, such as what data is of high priority and what is medium or low priority. When SDN-based transformation is implemented, only when the traffic of each level is clear can the overall traffic of each level be summarized, and then the summarized CoS level plan can be specified.

4. Allowed errors to a certain extent

To maintain the high utilization of pipelines, Google uses medium-and low-priority data to fill the data with high-priority data. When a fault or a high-priority business has a burst of traffic, the low-priority business will be delayed or discarded. In this case, lower-level application errors must be allowed, or lower-level applications are required to have sufficient endurance.

Network Intelligence, data classification and application acceptance are the key elements of Google's success in SD-WAN, network intelligence is a kind of ability, application data classification is the true core, application acceptance is the criterion for selection. For the user, carry on the application "network transformation", make it match with the new network is an essential part of the SD-WAN application.

3. Analysis of difficulties in deployment of operator SD-WAN

Currently, three basic carriers are providing WAN services. "BAT" is currently gradually implementing self-built WAN. Taking China Telecom, China Telecom, China's largest basic carrier, as an example, has established a WAN network, including Chinanet, CN2, high-quality professional network, and MPLS/leased line networks deployed in various forms, in terms of hierarchy, WAN can also be divided into access, Metro, and backbone networks. From the perspective of SD-WAN, the operator can be the user of the SD-WAN, expect through the deployment of the SD-WAN, to provide better WAN services for "BAT" or directly to the customer to provide SD-WAN services; "BAT" will be a potential SD-WAN deployer and user who wants to build their own "Black Network" SD-WAN Based on the carrier's WAN network and bypass network operations. This section focuses on the Several Factors of Google SD-WAN success, combined with the actual situation of domestic carrier WAN network, analysis of carrier SD-WAN deployment difficulties.

1. Difficulties in intelligent network transformation

The core of network intelligence is to assign a "smart brain" to the network and assign different bearer forms and business forms to the network according to different needs. At this stage, it is difficult for operators to assign a smart brain to existing wide-area networks and transform them into "intelligent routing selection" like Google B4 networks.

Taking China Telecom's WAN network as an example, the network-wide control plane and the bearer plane are separated to unify the control plane to the SDN Controller, there will be more than 0.36 million vrouters of various brands and models for access networks, man networks, and backbone networks. The cost of how to coordinate the transformation of multiple brands and different batches of devices according to one standard will be daily and unacceptable.

A more feasible solution is to use the underlay method to transform the network, that is, to achieve some "smart network" functions in a unified network management mode. This solution can achieve the flexibility of network configuration to a certain extent, provides certain traffic scheduling capabilities to improve business response capabilities. But the biggest problem with this underlay method is that it limits "intelligence" to a limited range, which may be a device of the same brand, a man or a carrier. For WAN users, in recent years, with the increase in basic network investment by operators, the basic network traffic requirements have been basically solved, and the pipeline is already wide enough, quasi-intelligent networks based on a city and a region cannot meet their needs.

Therefore, the carrier's WAN network will only be a network with a lack of overall brain for a long time in the future and will not meet the intelligent requirements of the entire network.

2. Data Classification and fault tolerance

The definition of data level is determined by the characteristics and requirements of the application. In different scenarios, the data requirements for network QoS vary. For example, in applications that adopt dual-active deployment, the dual-active synchronization data of applications requires a high level, and the data backed up by daily systems can be relatively low.

For enterprises that require multi-site interconnection, the Data Classification of their systems is seldom performed with operators. First, for existing systems, data classification involves transformation of existing applications, the existing private line bandwidth price of an enterprise is sufficient to support the one-size-fits-all bearer mode. Second, data classification involves data management within the enterprise, management-related problems are often not solved by IT personnel through design. Therefore, we can see that most of the enterprises in reality on the SD-WAN "multi-path selection by business level" function does not have the enthusiasm of imagination, the main reason is that enterprises cannot reasonably distinguish Application Data levels.

The application data level is tagged at the egress of the server, so that the network can know the importance of application data and facilitate the hierarchical network carrying and differentiated operation. It is the core of the operator's "cloud network convergence, is the main driver for implementing SD-WAN. However, most of the current business users are not aware of the importance of data classification, making the operator in the implementation of "multi-path selection by service level" and other typical business characteristics of SD-WAN network transformation is difficult. For WAN operators, if users do not classify the data, the "Application Data becomes customer data" will occur. In the same price system, all the customers of the operators are equally important, errors are not allowed. In this case, the operator can only increase the "pipeline bandwidth" by making annual investment.

Iv. Current Application of SD-WAN

In reality, in the absence of network and application convergence, operators can design networks with different network bandwidths and OoS at the network egress to provide Differentiated Services.

In most carrier WAN networking cases, large enterprise customers will divide their applications into a high-level network and adopt a high-QoS network bearer mode, small and medium-sized enterprises will select a medium-and low-QoS-guaranteed network. The carrier will tag different customer levels at the network egress (for example, using VLAN tags) for different customers for differentiated bearing.

Before SDN, the carrier will connect enterprises to the telecom data center through optical fiber for the above design, and implement "hard edge" devices in the telecom data center to carry customers of different levels in a hierarchical manner. The problem is that the customer needs to bear the heavy professional leasing price of the enterprise and cannot flexibly change according to the business format. The carrier cannot flexibly choose intelligent routing and fill in traffic within the WAN.

Combined with SDN ideas, drawing on Google's SD-WAN design, at present combined with the popularization and application of cloud computing virtualization technology, through the softening of the enterprise boundary, to build a "Soft Edge" approach to introduce the simplified version of the SD-WAN, 2.

Simplified SD-WAN

There are three leased lines. The red dotted line is the MPLS leased line with high QoS, and the green dotted line is the virtual leased line provided by tunnel with high QoS, the yellow dotted line is an Internet leased line with low QoS. All three methods are built on CPE devices controlled by SD-WAN controllers. CPE devices can use white box devices or virtual machines. The SD-WAN controller can configure the routing mode according to the designed data information, and deliver the configuration to CPE, CPE select different physical network for "most reasonable" Data routing.

In this way, the IDC internal network does not need to be transformed. Instead, you only need to introduce the SDN concept at the network egress/portal, which simplifies implementation and avoids various problems of large network devices. First, through the SD-WAN controller, realize the centralized management and automatic configuration of CPE equipment, realize the configuration management of various access methods, improve the quality of network service, increase the flexibility and mobility of the network. Second, the operator can carry on the traffic fill operation through the SD-WAN control, to realize the maximization of network utilization, realize the maximization of network value. Third, it brings network access interfaces to competent customers, and opens large network route scheduling conditions and scopes to enterprise customers. It also allows small and medium-sized enterprises to save costs when they are tolerable.

V. Conclusion

At present, most of the SD-WAN high-quality network solution is actually a kind of overlay solution that carries on the original WAN network, by overlapping the network "Tunnel", taking advantage of the characteristics of high-quality network light load, local improvement of network quality does not fully meet the customer's expectations for SD-WAN. Based on the current situation of the carrier network, this paper analyzes the existing carrier network to the Internet SDN transformation process, the WAN into SD-WAN difficulties and limitations, according to the analysis of Google network, the starting point of introducing SD-WAN to operators is put forward to maximize the benefits.

This paper discusses the difficulties of carrier-based SD-WAN construction, focuses on the analysis of the difficulties and feasibility of introducing SD-WAN to carrier-wide network at present, and puts forward the deployment direction based on IDC application. However, due to the limited time and capability, the actual application effect has not been further tracked and analyzed, I hope that the work will continue to follow the study of this project, and grasp the practical significance of SD-WAN for the deployment of operators, SDN, a major and stronger operator.