Build Your Exclusive Twilio in 48 Days - Talking about the Operator's Communication Center

Guide :

With the continuous increase of global mobile traffic and the continuous extension of communication cloud in vertical industries, Internet communication cloud has gradually become a mainstream communication service form. The new crown epidemic has stimulated the online needs of users. Whether it is personal life or corporate office, the transfer of communication activities from offline to online has become a trend, which in turn has spawned new consumption habits. In this context, my country's Internet communication cloud industry is facing new opportunities and challenges. In addition to meeting users' stricter interactive experience needs and exploring innovative scenarios, it is also necessary to quickly provide corporate customers with "thin and complete" solutions, allowing Communication capabilities and application scenarios are deeply integrated.

Text｜Yang Xianjun, NetEase Yunxin Architect and Technical Manager

foreword

Against a backdrop of high-profile videoconferencing, artificial intelligence and robotic automation, there is a silent, emerging disruptor: the humble phone call and SMS texting. 2022 is a year of rapid growth in the communications business, with calls and text messages already increasing significantly. Messaging is the most basic communication business, and it is also the entrance to many industry services and marketing. In the 5G era, SMS has been upgraded to 5G messaging, bringing new opportunities to various industries. The financial industry scenario will be one of the most typical application areas of 5G messaging. Through the mid-platformization of carrier communication infrastructure, voice calls, smart calls, text messages, video messages, smart messages, 5G messages, and integration with RTC and IM can be realized. With an open and agile structure, improve the efficiency of enterprise services and achieve integrated precision marketing. Combined with line/media management platform, intelligent marketing system, all-media intelligent customer service system, etc., to expand service applications, build a complete communication ecosystem and closed-loop communication chain. Realize comprehensive management of operator communication, reduce operator communication cost, and build operator communication ecosystem.

1. Seeing flowers in the fog: unveiling the veil of the operator's communication center

For the field of Internet communication, Zhongtai is more of a subject involved in the construction of 2B products, because of the abstract reuse of software systems, it is more of a problem faced in the construction of complex B-end systems. Therefore, the design of the middle platform is a topic that all B-end system architects should pay close attention to. How should the operator's communication center be designed? What are the characteristics? What is the essence of design? What are the challenges? This article will re-examine the construction of the communication middle station from a new perspective, so that you can have a deeper understanding of the essentials of the communication middle station.

There is a saying in the field of software development: " Any problems that arise in the process of software design and development can be solved by adding a layer". We will not discuss its right and wrong and its scope of application here, but what is certain is that the emergence of the middle platform is to solve the contradiction between the different operating efficiencies of the front and back offices. Differences in efficiency, in order to achieve the overall balance of the system.

The author very much agrees with the concept of Wang Yuan, vice president of NetEase: "All middle offices are business middle offices" . In a broad sense, the so-called middle and Taiwan are all for business services, so that enterprises can quickly respond to business needs and launch new products at lower costs and higher efficiency. The operator communication middle station is no exception. It belongs to a subset of the communication middle station. It is mainly aimed at operator communication, such as voice calls, VoLTE video calls, short messages, video messages, 5G messages and other basic operator services. process, and put the output to serve the business again. In essence, the communication capability of the operator is the source of the business, and each SKU developed by the platform in the operator communication is the carrier of the business, which allows the enterprise to quickly and efficiently replicate the solution, so that the case can be reused, the principle is the same but the form Each has its own merits, so as to achieve the improvement of production efficiency and achieve the purpose of reducing cost and increasing efficiency.

(1) What exactly is an operator's communication center?

Wang Jian from ThoughtWorks made a concise and clear definition: Zhongtai is an "enterprise-level capability reuse platform". When looking for various materials, the author found that there are many concepts related to "Zhongtai", but this definition is more inclusive, and clearly divides the scope and key values. "Enterprise-level platform" represents a top-level mechanism that covers multiple aspects and coordinates the overall situation, rather than a single system or service system; "Capability" is a highly abstract description of different technologies, business, technology, algorithm, data, AI. A variety of capabilities are combined into an enterprise multi-dimensional capability network; "reuse" is the core value of the middle and Taiwan, and the core essence of the enterprise's more agile and lower-cost operation.

The operator communication center, I think it is unprecedented, here I also define its birth: it originates from the voice, video and other media line resources provided by the basic operator, text messages such as SMS, RCS and rich media messages resources, combined with RTC real-time audio and video communication, IM instant message communication, plus the integration and scheduling of infrastructure and services, as well as the management of multi-tenant systems and distributed clusters, following the design concept of the domain-driven model, providing various business solutions The integration and precipitation of the solution ultimately serve various concrete needs of the enterprise, thereby forming a unified platform for reusable capability collections.

(2) Characteristics of the operator's communication middle station

1. Reusability

\
What Zhongtai provides must be different products in the enterprise, or general functions that can be used by different enterprises. If it is a product-specific function, it is meaningless to reuse it as a function. In terms of reuse, comprehensive evaluation will also be carried out in terms of functional reuse, data reuse, and industry reuse. The existing solutions of the operator's communication center, such as the high-connection integration solution launched by the RTC, can be applied to entertainment and social scenarios. The cloud call center SKU can be provided to various customer service SaaS platforms as the core call center scenarios. The intelligent voice robot capability integrates ASR, TTS, NLP and other general intelligent technologies of major manufacturers, and can provide capability support for various marketing inbound and outbound voice robot SaaS platforms. Voice notifications, voice verification codes, 5G messages, and video text messages can provide enterprises with various access services and information interaction satisfaction evaluation services . Ordinary outbound/inbound calls can be integrated into SCRM, ERP, HIS, IT systems, etc. as a general value-added service for government and enterprise offices to help traditional enterprise Internet + upgrades. All cases can be copied and applied in batches in various scenarios.

2. Platformization

\
In order to be effective, Zhongtai needs to empower multiple product lines, so it emphasizes platformization and serves the entire product ecosystem. Therefore, in the design of mid-stage products, try to consider how many common needs your product can meet in a product ecosystem, and what common scenarios your product can meet for different industry ecosystems. At present, the operator communication center mainly negotiates the scenario of the capability open platform combining the core switching network of the three major operators with innovative services; explores scenarios such as seeking medical consultation, chronic disease tracking, patient return visit, and doctor-patient communication together with medical enterprises ; for the financial industry Provide customer service system capabilities and office system upgrades; provide high reach rates for the entertainment and social industry , and complete the last mile of communication; provide the e-commerce industry with scenarios such as business-consumer communication, marketing promotion, and customer relationship management . A series of common application scenarios have been extracted, standardized and platformized for government enterprises, medical care, finance, e-commerce and other industries.

3. Business

\
In the past, there has been extensive discussion on whether or not China-Taiwan products should highlight business, but in fact, the concept of China-Taiwan itself is proposed to reduce business costs, share business product resources, and communicate data . Therefore, the middle-end products serving the business must have their own business nature, that is, the business functions can realize the empowerment of different business scenarios, and the data can be analyzed based on different business dimensions. The operator's communication middle station is not only a communication capability, it will combine the above-mentioned four major industries, and carry out a layer of abstraction according to the actual customer needs, with certain common business capabilities. By separating the layered model, the aggregation capabilities of various industry-related application scenarios can be encapsulated without interference on top of the basic capabilities , which greatly simplifies the complexity of logic during enterprise access, and can be used directly for common capabilities. , so that enterprises only need to focus on business needs without mastering specific technical details, reducing business coupling and understanding costs, and achieving the goals of improving productivity, reducing costs and increasing efficiency .

4. Standardization

\
The middle station needs to provide standardized functions. If your communication module function can only serve a single industry, as a middle station function, there will be few reuse scenarios. This does not mean that your mid-office business module can only remain unchanged, but it can have custom expansion functions to meet the support of multiple front desks for such functions in different scenarios. The operator's communication center has extracted a set of standardized capabilities and refined a set of standardized methods.

• Access/connection interface standards: For the access of operators or secondary agents, a set of standard E1 lines, NGN core network, IMS core network and other operator backbone networks, SIP, H323, CMPP, SMGP, SGIP are provided. Access standards such as standard protocols and non-standard hot-pluggable component standards. Support OPUS, G711, iLBC, G729, H264, VP8, VP9 and other standard codecs. It supports standard push-pull streaming protocols such as RTMP, HLS, FLV, etc. For the access of components such as ASR and TTS, it supports standard custom pluggable component modules, and also supports standard MRCP protocol docking. For the forward open interface, also refer to the NetEase Yunxin Capability Open General Specification to output the standard RESTFUL-style network capability open interface. For the standard system, it also provides general operator interface standards such as SIP and CMPP, which can be seamlessly connected. A set of docking process specifications and standards are defined, which are followed by the implementers, which reduces the cost of front-to-back docking and improves docking efficiency.
• Business domain division standard: follow the vertical and horizontal division of capability service modules in accordance with the microservice ecology and DDD domain-driven design, formulate the model and principles of architecture layering, define the roles and responsibilities of functional modules, the hierarchical relationship and depth of dependencies, and reverse dependencies The decoupling strategy and the specification of microservice governance, infrastructure construction, application and expansion of technical components, etc.
• Standards in the R&D process: also align development collaboration specifications, deployment specifications, log specifications and existing standards, and conduct further exploration attempts. The main purpose is to solve many problems such as the contradiction between system stability and flexibility, the fuzzy business boundary between the middle office and the forward business, the distance, and the difference in the R&D life cycle. As a mediator of conflicts and contradictions to resolve problems, thereby increasing productivity.

(3) The essence of the operator's communication middle station

Enterprise-level, abstract, sinking, and multiplexing, these keywords represent the essence of operator communication middle-station construction . At the same time, it is also a problem that requires deep thinking in the design of enterprise application architecture. (The so-called enterprise application architecture refers to the form in which the various software systems within the enterprise should be constructed and combined to efficiently support the operation of the enterprise.) Therefore, if you want to think deeply about the enterprise-level abstraction and down-level of software products. The problem of Shen and reuse can be reviewed from the following three perspectives: the perspective based on abstract reuse, the perspective based on the rationality of the architecture, and the perspective based on the unified management of business. I won't go into details here, we will discuss it in the example.

2. Dreaming about: Interpretation of the design scheme of the operator's communication middle station

(1) System technical characteristics

1. Follow the microservice ecosystem architecture and provide an open and standard web application programming interface

• Its characteristic is that it enables third-party users to develop a variety of services without knowing the specific technologies and protocols adopted by the basic network , as long as they know what capabilities the basic network can provide and utilize these capabilities. The basic network provides a variety of services. For example, location-based services, call center services, switchboard services, etc., the business logic of these services is located on the application server.
• Provide an open API interface to facilitate the secondary development of services and the introduction of third-party application developers to provide services.
• Equipment suppliers and operators can cooperate to develop new services required by the market according to market demands.
• Operators can independently and quickly develop new services according to market demands.
• New services with market prospects can be developed jointly by operators and third parties.

2. A unified communication platform spanning PSTN, GSM, NGN, IMS and IP networks

• In order to meet the integration of traditional services and data services, the operator's communication middle station provides multiple protocol adapters such as SIP and SMPP, supports access to multiple network devices, and can build unified services across PSTN, GSM, NGN, IMS, and IP networks. The platform provides unified services for users of these networks.
• The operator's communication middle station shields the details of the underlying bearer network and provides a flexible networking method, which can keep the original access method unchanged, reduce operation and construction costs , and can easily carry out comprehensive services.
• The operator's communication middle platform is developed on the basis of Internet applications and IP networks . In the early stage of design, the seamless integration with Internet applications and the unified service platform that can perfectly connect with the operator's basic network are considered.

3. A richer business

• The operator's communication middle station makes full use of the characteristics of IP to users in the NGN network to provide value-added services that combine voice and data , which is more competitive than the pure voice service in the PSTN network, and more comprehensively supports voice, data and multimedia. business.
• The customers have been subdivided, providing not only point-to-point voice and SMS services for enterprise users, but also new services such as call centers, contact centers, voice robots, multi-party calls, conference calls, and enterprise switchboards that are integrated with the Internet, and voice verification codes. ,notification. Become an important guarantee for operators to attract customers and win profits.

4. Flexible and powerful billing functions

• With flexible and powerful billing functions, it can provide various billing packages to meet the needs of various consumer groups, and complete the functions of generating, storing and transmitting billing data. It can be charged according to the duration of the session and billed by the article, and the charges can be recorded on the calling number and the called number.
• It can complete the corresponding charging function according to the charging information and charging rules of the specific service, and generate the charging record of the fixed content according to the service requirements.
• It can also generate billing events or billing reconciliation files, and a third party can directly collect and reconcile bills or perform real-time billing.

5. High reliability and high performance

• The operator's communication center adopts an advanced micro-service ecosystem to achieve service discovery and automatic failover, fuse, downgrade, and cure.
• Infinitely scale-out performance without attenuation. Compared with hot backup and cold backup solutions, resources can be fully utilized, and the requirements for hardware hosts are lower, ensuring high reliability and high performance of the system.
• It is divided according to the business function domain and distributed on different application servers to reduce the fault coupling between systems. When one business fails or the call volume is too large, it will not affect other businesses.

6. The product is mature and reliable to meet the needs of operation

• The capabilities provided by the operator's communication middle station at this stage are all self-developed , stable in operation, and have obvious advantages in maintenance, modification, customization, and response time, which can fully guarantee the smooth operation of the business.
• Since July 2017, the system has been running online, carrying tens of billions of short messages sent online, billions of calls, millions of users online, tens of thousands of concurrent conversations, and thousands of enterprises operating normally.

(2) Interpretation of the overall structure

​

This block diagram covers the whole picture of the operator's communication platform. The mode is divided into five levels, and the vertical is divided into two latitudes for service management. The five horizontal layers are the access layer, the gateway layer, the application layer, the capability layer, and the base layer . A three-layer progressive aggregation was carried out.

The first layer is the access integration of operators’ basic resources , Yunxin RTC basic communication capabilities, AI-related providers and self-developed capabilities, etc. The resources are arranged and distributed through the line provider platform to form a resource set to solve the stability of a single resource, Concurrency capabilities, blind spots, optimal path selection, resource isolation, etc., make a unified management allocation and scheduling.

\
The second layer is to encapsulate these resources into discrete standard capabilities with the smallest granularity, the highest composability, and the weakest association relationship. It does not reflect the performance of upper-layer services and applications. The purpose is to hide operators, RTC, AI, etc. technical details, while providing the basic capabilities of atomicity.

\
The third layer relies on the atomic capability of the second layer, and aggregates packaging according to business needs . It is heavily involved in the design, extraction, and packaging of business applications. Generate application capability sets one by one, which can not interfere with each other, or can be linked to each other, supporting rapid trial and error and agile iteration of products.

• access layer

  1. It mainly refers to the SDK of each platform. At present, it is mainly the SDK for terminals developed based on WebRTC.
  2. The other is the aPaaS-based call toolbar, which can be easily embedded to reduce the amount of development.

  3. The resources of the three directly connected operators are also converted into the SIP standard through the access layer for unified access.

  4. The front and back ends of the enterprise can also access the gateway directly to receive CC events or initiate calls or short message processes.

<!---->

• Gateway layer

  OpenAPI is mainly an open interface for external capabilities , with strategies such as downgrading and fusing frequency control. OpenAPI can be encapsulated in accordance with Yunxin open specifications and standards.

  *-GatewayRestApi This is a series of application gateways, mainly encapsulating the internal Dubbo microservice interface, presenting the rest interface externally, with the function of authentication and authentication, and the function of interface aggregation.

  WssLink is a long connection of WebSocket . The main function is to push tenant-related notifications or signaling to the SDK, and to improve server performance through push-pull combination.

  Zuul and Eureka are portal managers of gateway services . They manage the routing distribution of various Gateways, achieve visual configuration, reduce configuration on nginx, and dynamically distribute subdomain routing. The focus is on fusing, current limiting and frequency control. action, and unified authentication can also be performed on it.

  Another part is the access layer from SIP to the operator, part of this layer is the proxy forwarding of SIP signaling, and the other part is the proxy of the media (selected according to the actual situation of the network architecture).

<!---->

• application layer

  1. Mainly, various application fields are divided by domain , so that they are logically isolated from each other. At present, there are five domains: cloud call capability domain, general voice service domain, AI-related intelligent call domain, SMS service, and 5G message.

  2. The cloud call service domain provides common inbound, outbound, call control, interactive voice response, automatic call distribution, call queue management, call routing and other related capabilities, which can be expanded according to the characteristics of the cloud service platform.

  3. General voice service domain: This domain mainly implements simple IVR broadcast, VoIP point-to-point call connects the telecommunication network and the Internet, and two-way callback is sent by the server to control the operator-to-operator callback mode, through the number pool, etc. Relevant functions are encapsulated with hidden sign logic, business logic encapsulation of voice verification through processes such as verification code issuance and reporting, and other types of simple point-to-point business capability encapsulation.

  4. Intelligent call service domain: notification robot, single round, multiple rounds, mainly relying on NLP, ASR and TTS capabilities in the communication capability domain. ASR is currently connected to Alibaba Cloud ASR and NetEase Hangyan ASR; TTS is connected to Alibaba Cloud TTS , Netease Hangyan TTS, self-developed local TTS, and business functions such as human-machine assistance and intelligent quality inspection have been realized.

  5. Other business capability domains: The parallel domains are independent and decoupled from each other, and the communication capabilities of each other are shared. Emerging capabilities can be dynamically expanded according to the development of business capabilities. It is conceivable that there are IoT-related, teleconference-related, and video telephony. Or three-party capability integration and other related functions.

• Capability layer

  Mainly to provide a complete set of operator communication capabilities . From the perspective of application, it can be divided into two layers, one is the basic capabilities, and the other is the line provider resource integration platform. The design of the line provider platform will be discussed in detail here. Talk about basic skills.

  From the perspective of functional modules , it can be divided into: tenant-related account system and rights management, various media scheduling load balancing in the scheduling center, multi-region scheduling, etc. The media center is mainly edge media links, signaling routing, relay Proxy distribution, the basic capabilities mainly include various system-level intermediate components and plug-ins or modules on media services.

  - Tenant management: This is mainly the construction of the SIP account system. For SIP calls or linkage operations, you must have an account that can log in and register and have a heartbeat keep-alive link to be unblocked. Therefore, a set of your own communication will be established. Account system, two-way callback or simple point-to-point calls do not need to enter the account system. The tenant management account allocation here is the concept of logical tenants, and media allocation supports logical isolation and media service physical location isolation. It also includes the management of trunk numbers and the management of black and white lists for part of the risk control processing. The online billing system is mainly used for settlement and reconciliation with line operators or operators, and also uses the tenant enterprise prepaid model.

  - Scheduling center: It mainly includes resource scheduling of media services. There are two cases. One: scheduling between media groups used in different business fields, and scheduling between application clusters according to the relationship between tenants and application clusters . Second: load balancing is performed according to the survival or health status of the cluster according to the same business capability. The capability signaling agent layer is the connection layer between the capability domain and the application domain, and is mainly responsible for the packaging, conversion, delivery or callback of signaling; the media signaling management and control service mainly translates the capability signaling to the softswitch signaling to the events of the softswitch service. Assembly of notifications and signalling generation and aggregation to softswitch services.

  - Media center: The current architecture of the media center is what NetEase Qiyu has been using. It is divided into tenants, application functions, and clusters, and can be supported to the tenant application level. Media, routing and relay services need to develop a visual cluster management function interface to manage various local configuration files, and the main design remains unchanged.

  - Basic capabilities: including configuration center, distributed tasks, distributed timers, various self-developed AI capability-related modules of softswitch and other components or basic capability services, providing support for the basic capability layer or each capability domain and each application domain.

  - Line management: It is mainly the scheduling management of line resources, which will be explained below.

<!---->

• base layer

  It mainly includes data access layer, relational data storage, KV data storage, object storage, data retrieval, etc. In fact, the basic layer also includes some line resources, and the hardware gateway devices of the hybrid cloud together form a basic resource network for the line provider platform to perform resource scheduling and data storage at the capability layer.

<!---->

• operational domain

  This area manages application functions and capabilities from a macro perspective. It is mainly a variety of operation support systems that are supported by colleagues in forward departments, such as: operator management platform, billing and settlement platform, and capability openness management platform. The data analysis platform makes analysis and statistics of relevant business data for the forward, which is the reference data for decision-making.

<!---->

• Operation and maintenance domain

  This area manages the service cluster availability and O&M convenience from a micro level. It is mainly convenient for R&D, technical support, testing and other students to carry out visual system management, online problem troubleshooting, performance optimization to provide monitoring or performance index data, service quality monitoring and other functions to monitor, manage, manage and maintain operations. It also has a built-in set of microservice governance management. Tools, including service governance, configuration management, data query, alarm management, link tracking, indicator center, etc.

(3) Interpretation of network structure

The following figure mainly shows the network level distribution of the operator's communication center in a privatization project. The combination of the three networks is as follows:

• Carrier communication backbone network: refers to the access point of the carrier's core network. This layer of network basically belongs to the carrier's internal private network that is different from the Internet, generally referring to GSM, NGN, and IMS. The end office is basically not external. Through DC1 and DC2, the resources must be opened to the outside world after strict network security control by the SBC edge control node. Smaller customers open resources to the outside world through analog lines, digital E1, IMS lines, and VoIP lines.
• Enterprise intranet communication area: generally refers to the enterprise internal network, which mainly stores related service clusters such as the application layer, capability layer, gateway layer, etc. If there is a network requirement, the media layer will also be placed in this layer, and the SBC will be used as a media agent to go out to the public network.
• Enterprise DMZ mixed area: generally put the application cluster related to the access layer. If the enterprise has no requirements, the media layer can also be directly placed in this layer for performance considerations, omitting the media proxy process of the SBC.

Through multiple computer rooms and multiple activities, the overall failover of the computer room can be achieved. The global distribution of media endpoints can be achieved through the WE-CAN global communication network. Through the weak network confrontation ability of Yunxin RTC, each application scenario can achieve 40% higher voice call quality than the traditional architecture model. The weaker the network, the more obvious it is.

By layering applications and formulating application levels, the importance and importance level of applications are defined for products, R&D, and testing, thereby formulating application invocation dependencies and concurrent performance requirements.

For details on how the three networks are connected for media exchange, you can continue to see below.

​

(4) Interpretation of the media switching layer

Based on the perspective of abstract multiplexing, the construction of media layer, signaling, and media scheduling management is the core of this middle station. Rely on repetitive functions and modules to abstract and sink one layer down. What is abstraction? What is sinking? You can refer to the figure below.

​

There are multiple subsystems, and multiple subsystems will access, connect, and edit media. Switching has dependent operations. After analysis, it is found that the processing functions of media are highly similar to repetition, which can be abstracted and merged to avoid repeated construction. Therefore, there is a high-availability cluster for core signaling and media processing, and then subsets are divided according to different services. , there is a unified routing scheduling strategy, which not only achieves the unified abstraction of merging , but also can do moderate business isolation.

• The media layer constructed in the communication capability platform is divided into five levels for transmission, including access terminal, WE-CAN communication network, converged communication gateway, audio and video media exchange capability domain, and line provider platform.
• The access terminal mainly provides an operator communication capability SDK . This SDK performs a layer of service encapsulation on the existing Yunxin RTC-SDK. The media transmission adopts the existing method, and the transmission is packaged based on the existing interface. Ability, business signaling uses an independent link to connect to it to encapsulate call center login, online status management, session management, call origination management, called event management, status event management, sending DTMF, call hold, mute, call forwarding , multi-party conference and other related basic modules, these modules are directly connected to the relevant application services in the back-end call capability domain to drive the overall operation of the application layer and the media layer.
• WE-CAN communication network is mainly responsible for network quality optimization of RTC audio and video . Including weak network confrontation correlation, packet loss retransmission mechanism, QoS dynamic detection and optimization adjustment, support for point-to-point media negotiation and transmission. Focus on the control of media quality, and at the same time do signaling bridging with the converged gateway layer through wss+jsonrpc.
• The layer of the converged communication gateway is the key to convergence. It mainly plays the role of undertaking and compatibility, and it is the bridge connecting the WE-CAN communication network and the calling capability domain. It is only responsible for point-to-point signaling negotiation , and some specific interaction processes will be discussed in the detailed design later. For stability and capacity considerations, the converged communication gateway does not do media processing and media proxying. It only does signaling negotiation and conversion, and then exchanges the bilaterally negotiated RTP port to the other party to start media transmission. It is used to address and route the WE-CAN communication network to discover the available MediaServer.

• The audio and video media exchange capability domain is the main body of the operator's communication center. This set of architecture is composed of a distributed, high-availability, high-concurrency media cluster + scalable business layer.

  -The media cluster includes a three-layer structure of access layer, routing layer, and media processing layer. It acts as a SIP signaling proxy through Sip-Proxy, and mainly receives related events or messages such as Register and Invite sent by the converged communication gateway. The server discovers and registers the available nodes of the routing cluster, finds the available media nodes through addressing through the policy server, exchanges the RTP port of the WE-CAN communication network and the RTP port of the media node, and then conducts point-to-point audio and video calls on both sides.

  • All events generated in the entire access process and the intermediate control process of the call will be converted and transmitted to the upper-layer application business domain through the event control service . The business domain will process its own state, record the data it cares about, change its own state, or initiate Controls on the platform side.

  - Connect the specific available nodes of the media layer with the line provider platform through the signaling connection of the SIP-TRUNK service, carry out the media landing of outbound calls, and at the same time receive the incoming call requests from the line provider platform 400, etc. and connect with the media node through application services The domain controller performs agent addressing, and finally contacts an RTC terminal for the called call.

• The main responsibilities of the line provider platform are: to connect trunk lines and the backbone network directly connected to the operator and the enterprise's self-pulled line, management of number risk control, line side billing, and bridging negotiation of incoming and outgoing calls, the line provider platform It will manage and control the concurrency of lines and optimal path selection, as well as detection of connection rate and avoidance of abnormal lines and blind spots. The maintenance and management of the resource pool can solve the single point of failure and line bottleneck through various scheduling strategies, so as to meet the performance and stability requirements of a single access operator.

(5) Interpretation of terminal access layer

As shown in the figure below, the NECCC SDK architecture is also a good layered model design. It is mainly connected to the signaling service and media service of the operator's communication center . The encapsulation reduces the detailed process and provides a closed functional interface to the outside world. It is divided into six layers: base layer, engine layer, cross-platform encapsulation layer, SDK interface layer, platform terminal layer, and easy-to-use system. Through the assembly of these layers, a lightweight, extensible, application terminal SDK library is finally formed. It is mainly responsible for the collection, editing, optimization, transmission of audio and video, negotiation and interaction of signaling and processing of some core business logic.

​

• The basic layer is mainly the basic network library, the encryption and decryption algorithm library, the concurrent processing of threads, the management modules such as logs, and the resource layer unrelated to the business.
• Cross-platform encapsulation mainly includes some frameworks and component layers such as Uniapp, Ajax, WebSocket, Signalling, etc., to enhance the compatibility of different platforms.
• The SDK interface layer mainly includes the end-side capability interface and signaling interface encapsulation of the operator's communication middle station, the establishment and destruction of the user agent layer, the session management of RTC, and some asynchronous event handlers, etc., which are the upper platform terminals. Standard interfaces are provided.
• The platform terminal is mainly adapted to the application of the relevant presentation layer such as Web and applet, and more other terminals can be expanded in the future.
• The easy-to-use system includes sample code, common components, and solution demos for reference by enterprise developers, enabling easier access and function demonstrations.
• The engine layer is shown in the figure below. The core of the SDK mainly includes audio engine, video engine, and QoS engine. It integrates a variety of audio and video algorithms to fight against weak networks such as network packet loss, delay, and jitter. It also expands some AI technologies to collect sample data from the SDK to build a learning model to achieve the best effect of network fluctuation self-regulation .

​

(6) Interpretation of the line resource layer

Based on the perspective of unified business management , the line operator platform mainly plays a central role in resource management and scheduling, to avoid the physical failure of a single line, improve stability, and improve concurrency capabilities to break the concurrency and regional restrictions of a single operator. As many line resources as possible for backward access. Forward to provide internal business lines, externally to agents and individual enterprises to provide optimal line resources , and more importantly, to provide line resources to the open platform for forward operator capability applications to construct applications that fit the needs of enterprises and enterprises. solution, so how does it do it? Let's continue reading.

​

• The input and output of this platform use UDP/SIP/RTP protocol as the standard protocol. Whether it is IMS or VoIP line between operators, it is basically SIP standard. It performs well between the computer room and the computer room network. For analog lines and E1 digital lines, they are currently converted to standard SIP in the remote computer room through a hardware trunk gateway, and then connected to the line provider platform on the cloud to achieve unity.
• This platform mainly includes a management configuration visualization platform, a softswitch self-developed module to realize the synchronous application of management configuration, and a proxy service for load balancing and high-availability cluster management.
• Load balancing supports forwarding traffic by SIP302 hopping, supports traffic allocation according to actual load, supports quantitative weighted random allocation of a single softswitch host, supports random allocation and sequential allocation, and the specific implementation details refer to the mode of microservice discovery.
• Softswitch service self-developed module to synchronize and apply dial plan, number change, creation and use of access gateway and landing gateway. Supports active push synchronization after configuration changes, supports configuration hot loading, supports two modes of media and media transparent transmission, and supports configurable SIP header customization parameters.
• The configuration management service mainly includes visual display, editing of various configurations, billing logic, tenant management, number pool management and intelligent routing rule management. As shown in the figure below, packet capture analysis, data statistics and other functions are being improved in the future.

​

Such a centralized management platform in the industry is relatively rare, and may be more professional, and the command line of people who pay attention to this area is sufficient, so it has not been developed. At present, only the VOS system has a relatively comprehensive visual line management docking platform. In the future, Yunxin will also be regarded as a family. Welcome to consult and build together.

(7) Interpretation of the application capability layer

From the perspective of architectural rationality, the previous part of the theory of the application capability layer has been mentioned, which is mainly divided into atomic basic function encapsulation, general basic application capabilities, and partial business solutions . A lot of space is spent on the part of audio and video media exchange, and the part of line resource management. Above these two parts is the application capability layer. The application capability layer follows the microservice ecosystem. The relevant internal structure of microservices is here. Without further discussion, here we focus on the relevant applications of the 5G messaging platform for an interpretation, as shown in the following figure:

​

NetEase Zhiqi has obtained the 5G messaging CSP qualification of the three major operators in 2021. The 5G message system architecture mainly includes 5G message center, MaaP platform and 5G message interworking gateway, and is connected with user data management (HSS/UDM), SMS center, security management and control system, business support system, etc.

The 5G message center is responsible for processing 5G messages and connecting with the MaaP platform to provide industry message functions. The MaaP platform is connected to the 5G message center to provide industry message functions, and is responsible for the access management, authentication, and multimedia content upload and storage of industry customers' Chatbots and messages. The 5G message interworking gateway provides cross-operator 5G message interworking service functions.

The core function points are as follows in the function map. There are two main blocks: CSP and RCS, related to contract customer management, rich media message template editing, material management, and message sending. More detailed information is not expanded here.

​

The difficulties of the 5G messaging platform are mainly the various audit mechanisms, agent operation mechanisms, terminal equipment support and message downgrade strategies , which have been complicatedly defined in the early stage, and the state flow control needs to be handled according to the standard. According to the user's operations on non-direct customers and Chatbots, we change the status of the corresponding entities inside our platform while synchronizing to the operator to help users understand the current information more intuitively. After the user performs the operation, he needs to wait for the platform to review, and judge the operation that can be performed according to the result of the review. The detailed process will not be repeated here. The main purpose of this is to intercept all possible harassment and chaos in the early stage of message sending . The following figure shows the internal state flow.

​

​

5G messaging has very powerful application advantages. First, it has the functions of chat, group chat, applet, and various service apps of OTT applications . Second, it is the traffic portal . ; The third is the native application , no need to download, register, login and other pre-operations, no need to bind the mobile phone number, collect the verification code and other actions, it can be used immediately; the fourth is safe and reliable , based on the strong association of mobile phone number and real-name authentication, Allow personal data to be interconnected between different applications to ensure data security and reliability; Fifth, cross-platform connection , users do not need to switch multiple APPs, the same interface, to achieve cross-application interaction and interconnection between applications.

3. Ukiyo Qinghuan: A brief taste of the practical case of operator communication in the middle and Taiwan

In terms of space, the examples here only list a few innovations related to integrated communication. The specific implementation details of the cases are not expanded here. If you are interested in continuing to pay attention to NetEase Zhiqi Technology + Official Account, we will share each case from the source to the thinking in the future. , then to the design details, and finally to the interpretation of the whole process of practice.

Cloud call center application

Achieved results: intelligent routing and distribution, accurate service; service process monitoring, efficient management; insight into customer needs, active marketing; system-assisted labor to improve efficiency; sales process monitoring, mastery.

Application scenarios: Mainly used in customer service scenarios and telemarketing scenarios.

Refer to NetEase Seven Fish:

https://qi.163.com/callcenter

Outbound robot application

Achieved effect: Coping with massive scenarios, replacing manual calls for multiple rounds of interaction, enabling enterprises to accurately identify target users, achieving new customer growth, improving retention, and recovering losses.

\
Application scenarios: shoes and clothing, mother and baby, e-commerce, medical care, government affairs, public safety. \
Refer to NetEase Seven Fish:

https://qi.163.com/excallrobot

Converged Call High Connectivity Application

\
Realized effect: When making a call, the user can choose the RTC voice call in the application, can choose to call the mobile phone number directly, and can choose to transfer the mobile phone number when the RTC call cannot be answered.

\
Application scenarios: Mainly used in entertainment industry lines, as the last road call to improve the reach rate.

\
Reference Demo:

https://res.qiyukf.net/voip/7b6734f6d3c15a736611e50e1336894e https://res.qiyukf.net/voip/7b6734f6d3c15a736611e50e1336894e

Converged Call Voice Notification Application

\
Realized effect: users can receive voice notifications, and can listen to voice short messages after answering the phone.

\
Application scenario: It can be used in medical, financial, enterprise administration, telemarketing and other industries, as a single-point one-time notification application, easy to access and easy to use.

\
Reference Demo:

https://res.qiyukf.net/voip/ea4469d81d3fb35ea7a05b50204c8e17 https://res.qiyukf.net/voip/ea4469d81d3fb35ea7a05b50204c8e17

Converged Call VoLTE Application

\
Realization effect: When a user initiates a video call, he can choose the RTC video call in the application, and he can choose VoLTE video call mobile phone. If the called mobile phone supports VoLTE, he can directly use the video to answer.

\
Application scenario: It can provide a new possibility as a G2 audio and video call reaching endpoint, and it is currently in the experimental stage.

\
Reference Demo:

https://res.qiyukf.net/voip/2e672c4056e670745577d2726a651ea0 https://res.qiyukf.net/voip/2e672c4056e670745577d2726a651ea0

VoLTE + Virtual Human Application

\
Realized effect: After the RTC to the mobile phone user's video is connected, the motion of the RTC user can be directly captured and displayed to the mobile phone user through the virtual image.

Application scenario: It can be used as a virtual reception or interview scene when the anchor such as virtual customer service cannot appear on the screen.

\
Reference Demo:

https://res.qiyukf.net/voip/359245ca2522154abe7538af0d6728f1 https://res.qiyukf.net/voip/359245ca2522154abe7538af0d6728f1

4. Avenue Buddha: Imagining the development of operators' communication middle-end applications

The operator's communication center is open to the outside world as a cloud-based PaaS communication capability, and also serves as a long-term strategic partner of the operator, dedicated to the research and development of new technologies, aiming to help global operators meet new challenges and opportunities.

The operator's communication middle station is not only the construction of capacity but also the construction of ecology. With the aim of openness, cooperation and win-win, we will work together with operators and partners to actively participate in the construction of enterprise Internet + transformation. The development of Zhongtai is not achieved overnight, it is a gradual process and a process of value embodiment. As Chen Li, general manager of Yunxin Business, said: Any system or solution we make is to extract requirements, translate, organize, and abstract from the enterprise, and then carry out functional figurative packaging and integration, and finally Landed in the enterprise to reflect its value. With the development of the system and the in-depth understanding of enterprise applications, the future will be further discussed as follows:

• Refined communication capabilities: The expansion and application of refined capabilities such as risk control, VoLTE video, IVVR, ringback detection, etc. In today's informatization, information security will become the most important link. The text risk control system is relatively mature, and we have a good direction for the exploration of risk control in the audio and video field, but operators will continue to try to reduce costs and improve recognition efficiency in the construction of risk control for voice calls. Self-service interaction such as IVVR is introduced from the traditional call center IVR and applied to the video field to achieve the ability to edit streaming media online. And so on, there are many other function points, which can make the capabilities more refined and closer to the business form of the enterprise.
• Omni-channel contact center: Of course, we are not building a SaaS application for a contact center. We mainly provide endpoint resources and application solutions required by an omni-channel contact center. Weibo, WeChat, video, IM messages, and 5G messages are globally connected, with one touch and omni-channel integration.
• Cognitive intelligence, human-machine collaboration: customer service assistance, ASR, TTS, NLP access standardization, process engine and other AI-related basic capabilities integration, providing a platform-based general capability set for artificial intelligence application scenarios.
• Auxiliary G2 capability expansion: Provide reachability for audio and video calls or conference functions, etc. In the future, IVVR, ACD, ROUTER and other functions can be directly multiplexed on G2 audio and video media editing to expand G2 application capabilities.
• Componentization of applications to reduce enterprise access costs: Recently, there has been some substantial progress in the development of low-code Easily build a medium-complexity application system? There are still challenges. The following figure is an idea of ​​Netease's Qingzhou low-code platform, which can be based on components and modular programming. All functional points can be abstracted into components, which users can choose to assemble at will. After completion, an available communication capability platform will be formed. , which can be directly operated or used by abstracting the interface twice in the way of PaaS capability. It can be very convenient to choose on-demand, which reduces the purchase cost of enterprises. In addition, the application capability can be accessed with a lower coding amount, and the access efficiency is improved. Future solutions can be built directly on low-code platforms.
• ​

• Converged communication middle station: The operator communication middle station is a subset of the communication middle station. We will use it to practice in the early stage. It can generate a variety of unrelated component libraries like building blocks, which can be assembled at will to become large or small. controllable platform. Enterprise customers can assemble and use public cloud capabilities on demand, or they can purchase required modules to form a customized privatized platform and deploy them inside the enterprise. In the later stage, more modules will be integrated in the same mode, and finally integrated into a converged communication middle station.

references

• "Winning B-side"
• "The Way of Enterprise IT Architecture Transformation"
• "2021 Global Internet Communication Cloud Industry Research Report"
• "Phoenix Architecture - Building Reliable Large-Scale Distributed Systems"
• "5G Messaging White Paper"

Related reading recommendation

• Technical dry goods | Speech de-reverberation algorithm practice in real-time communication services
• In-depth analysis of the "Circle Group" message system design | "Circle Group" technical series of articles
• 通信