日常前言

• 四月份花了一大部分時間去深入代碼，把我們的雙攝虛化流程解析了一遍。然后為了給組內(nèi)分享，又花了相當(dāng)一部分時間去做總結(jié)，畫思維導(dǎo)圖、作流程圖等等，這其中學(xué)到了挺多東西的，尤其是對高通 Camera HAL 層的數(shù)據(jù)流部分，Camera Post Process 的前后節(jié)點(diǎn)都有了一個比較大概的了解，在跟蹤數(shù)據(jù)流的時候沒那么頭暈了。
• 還有，總結(jié)、分享知識的時候，作圖真的非常重要，一份填滿大量文字的 PPT，可能講 3 個小時都講不完，最后聽眾也很難有所收獲。然而如果有 70% ~ 90% 的篇幅用圖片來直觀表述，剩下的文字用于精煉、簡潔地描述，這樣可能 1~2 個小時就能搞定，并且聽眾至少也能留下一個比較整體的印象。
• 好吧，扯遠(yuǎn)了，回歸這次的活動，這一期是物聯(lián)網(wǎng)的主題，又是我不熟悉的領(lǐng)域，只能找一些介紹性的文章來翻譯了。以及……上期又送來一個抱枕……這期要是再送公仔，那就把這些東西送給女盆友一宿舍當(dāng)畢業(yè)禮物吧hh
• 這期采納了四篇：
  
  • 十大業(yè)務(wù)步驟——確保物聯(lián)網(wǎng)生態(tài)系統(tǒng)的安全
  • 物聯(lián)網(wǎng)中的認(rèn)知科學(xué)與網(wǎng)絡(luò)監(jiān)督
  • 剖析物聯(lián)網(wǎng)解決方案
  • 在物聯(lián)網(wǎng)中應(yīng)用機(jī)器學(xué)習(xí)：使用 Android Things 與 TensorFlow

---

版權(quán)相關(guān)

翻譯人：StoneDemo，該成員來自云+社區(qū)翻譯社
原文鏈接：The Anatomy of an IoT Solution
原文作者：Hans De Visser

---

The Anatomy of an IoT Solution

題目：（剖析物聯(lián)網(wǎng)解決方案）

The Internet of Things (IoT) is a powerful, transformative force and the cornerstone for digital businesses taking advantage of the convergence of the physical and digital worlds.

物聯(lián)網(wǎng)（IoT，Internet of Things）是一個強(qiáng)大的變革力量，它是數(shù)字商業(yè)利用來融合物理與數(shù)字世界的基石。

McKinsey estimates that IoT could generate $11.1 trillion a year in economic value by 2025. Understandably, IT leaders face growing pressure from the business to deliver new IoT solutions that improve operational efficiency and grow revenue via connected products and services.

麥肯錫公司（McKinsey）估計(jì)，到 2025 年，物聯(lián)網(wǎng)每年可創(chuàng)造 11.1 萬億美元的經(jīng)濟(jì)價值?？梢岳斫獾氖?#xff0c;為交付新的物聯(lián)網(wǎng)解決方案，通過將產(chǎn)品與服務(wù)進(jìn)行連接來提高運(yùn)營效率并增加營收，IT 領(lǐng)導(dǎo)者隨之面臨著越來越多業(yè)務(wù)上的壓力。

I’ve been involved in quite a few discussions with customers lately on what it takes to develop IoT solutions and how to establish the right architecture. When diving into the subject, I hardly found materials online that would give me an end-to-end overview of IoT solutions architecture, or best practices for getting started with IoT application development.

最近，我與客戶進(jìn)行了相當(dāng)多的討論 —— 討論如何開發(fā)物聯(lián)網(wǎng)解決方案，以及如何建立正確的架構(gòu)。當(dāng)我們深入研究這個主題時，我發(fā)現(xiàn)，幾乎沒有任何在線材料能給我提供一個物聯(lián)網(wǎng)解決方案架構(gòu)的，端到端的概覽，或者關(guān)于物聯(lián)網(wǎng)應(yīng)用程序開發(fā)的最佳入門范例。

So, I’ve started to paint the picture, relate it to projects that Mendix customers are doing in this domain, and distil some best practices.

所以，我開始描繪這幅藍(lán)圖，將其與客戶（Mendix）在該領(lǐng)域所做的項(xiàng)目聯(lián)系起來，并提煉出一些最佳范例。

I’ll cover the story in three blog posts, starting with describing the anatomy of IoT Solutions, followed by one about the role of platforms in simplifying IoT Solution Development. The last post will cover recommendations on architecture and best practices for defining an incremental IoT strategy.

我將用三篇博文講述這個故事，先從描述對物聯(lián)網(wǎng)解決方案的解剖開始，然后就是關(guān)于平臺在簡化物聯(lián)網(wǎng)解決方案開發(fā)中的作用。最后一篇文章將會介紹一些關(guān)于架構(gòu)的建議，以及定義增量物聯(lián)網(wǎng)戰(zhàn)略的最佳范例。

What does it take to develop IoT solutions, and how do you establish the right architecture? IoT solution design is quite different from typical IT solutions in that it bridges the physical world of Operations Technology (OT) with sensors, actuators and communication devices, and the digital world of Information Technology (IT) with data, analytics, workflows, and applications.

開發(fā)物聯(lián)網(wǎng)解決方案需要什么？如何建立起正確的架構(gòu)？物聯(lián)網(wǎng)解決方案的設(shè)計(jì)與典型的 IT 解決方案大相徑庭，因?yàn)樗鼘⑦\(yùn)營技術(shù)（OT，Operations Technology）的物理世界（傳感器、執(zhí)行器和通信設(shè)備），與信息技術(shù)（IT，Information Technology）的數(shù)字世界（數(shù)據(jù)，分析，工作流，以及應(yīng)用程序）連接了起來。

The diversity of use cases and operational requirements creates an array of IoT Endpoints, communication protocols, data management, and analytics technologies, as well as corresponding deployment topologies.

用例和運(yùn)營需求的多樣性創(chuàng)造了一系列物聯(lián)網(wǎng)終端，通信協(xié)議，數(shù)據(jù)管理和分析技術(shù)，以及相應(yīng)的部署拓?fù)?#xff08;Deployment topologies）。

And that’s just for establishing the foundation. The real value of IoT comes from turning data into insight, and making it actionable to drive smarter operations or launch new products and services. This fuels the need for IoT apps that empower users to act upon insight, by combining sensor data, data residing in enterprise systems such as ERP, CRM, and PLM, and even third-party services such as weather and traffic data.

而這些僅僅是為了打下基礎(chǔ)。物聯(lián)網(wǎng)的真正價值源自將數(shù)據(jù)轉(zhuǎn)化為洞察力（Insight），并使其能夠付諸實(shí)施，以推動更智能的運(yùn)營或推出新產(chǎn)品和服務(wù)。通過結(jié)合傳感器數(shù)據(jù)、保存在企業(yè)系統(tǒng)（如 ERP，CRM 和 PLM）中的數(shù)據(jù)，甚至是第三方服務(wù)（如天氣和交通）數(shù)據(jù)，使得用戶可以根據(jù)自己的想法行事，這就激發(fā)了用戶對物聯(lián)網(wǎng)應(yīng)用程序的需求。

What an IoT Solution Looks Like

（物聯(lián)網(wǎng)解決方案看上去是何種模樣？）

Despite the diversity, there is a level of commonality across use cases that can illustrate the anatomy of IoT solutions. Taking a layered approach in describing the anatomy helps identify relevant services and technologies from the things-level all the way up to IoT apps.

盡管多樣化，但還是有一些跨用例的共性，可以用于圖解說明物聯(lián)網(wǎng)解決方案的剖析結(jié)構(gòu)。采用分層方法描述剖析結(jié)構(gòu)有助于從事物層面直到物聯(lián)網(wǎng)應(yīng)用，即從各個層面識別相關(guān)的服務(wù)和技術(shù)。

IoT Endpoints

（物聯(lián)網(wǎng)終端）

This layer covers the physical world and operational technology required to connect things and communicate:

• Things: The real endpoint for IoT is obviously the thing that should be connected, whether physical products like cars, jet engines, and lighting systems, or other ‘things’ like livestock, crops, human beings, or spatial areas like rooms or outdoor space.
• Sensors: collect and report data on the actual status of things to which they’re connected. Sensors could be mounted on, or embedded in, things to monitor temperature, pressure, light, motion, location, etc.
• Actuators: control the physical or logical state of a product through signals they get from IoT apps or other systems, like opening a valve, or turning a camera, motor or light on/off. This includes commands sent to embedded software e.g. to reboot or update configurations.
• Agents: components that mediate between a set of IoT devices and act as a bridge between the sensors/actuators and the cloud, deciding what data to send and when. In reverse, they also process commands and updates coming from the cloud.
• Edge computing device: a distributed architecture in which IoT data is processed at the edge of the network. Transmitting massive amounts of raw data over a network puts tremendous load on network resources. In some cases, it is much more efficient to process data near its source and send only the data that has value over the network to the cloud.
• Communication: For IoT device communication, the physical layer and communication protocols are distinguished. As far as the physical layer is concerned, gateways, mobile devices, mesh networks, and direct- or broadcast device communication are alternatives that may or may not be suitable depending on the use case. The choice for the physical layer will determine which communication protocols are most suitable (e.g. MQTT, COAP, HPPT(S), AMQP, ZigBee, Z-Wave, etc.)

這一層涵蓋了連接事物和進(jìn)行通信所需的物理世界和操作技術(shù)：

• 事物：真正的物聯(lián)網(wǎng)終端顯然是那些應(yīng)該連接的東西，無論是汽車，噴氣發(fā)動機(jī)和照明系統(tǒng)這樣的物質(zhì)產(chǎn)品，還是其他諸如家畜，農(nóng)作物，人類或空間區(qū)域（如房間或室外空間）此類 ”事物“。
• 傳感器：收集并報(bào)告他們所連接的事物的實(shí)際狀態(tài)數(shù)據(jù)。傳感器可以進(jìn)行安裝，或嵌入到物體中，從而監(jiān)測溫度，壓力，光線，運(yùn)動，位置等數(shù)據(jù)。
• 執(zhí)行器：通過從物聯(lián)網(wǎng)應(yīng)用程序或其他系統(tǒng)所獲得的信號（如打開閥門，打開/關(guān)閉相機(jī)、電機(jī)或燈）來控制產(chǎn)品的物理或邏輯狀態(tài)。這包括發(fā)送到嵌入式軟件的命令，例如重啟或更新配置。
• 代理：在一組物聯(lián)網(wǎng)設(shè)備之間進(jìn)行調(diào)解的組件，它們同時還作為傳感器/執(zhí)行器和云端之間的橋梁，決定發(fā)送什么數(shù)據(jù)以及何時發(fā)送。反過來，它們也處理來自云端的命令和更新。
• 邊緣計(jì)算設(shè)備：在網(wǎng)絡(luò)邊緣處理物聯(lián)網(wǎng)數(shù)據(jù)的分布式體系結(jié)構(gòu)。通過網(wǎng)絡(luò)傳輸大量的原始數(shù)據(jù)會給網(wǎng)絡(luò)資源帶來巨大的負(fù)擔(dān)。在某些情況下，在數(shù)據(jù)源附近處理數(shù)據(jù)，并僅通過網(wǎng)絡(luò)將有價值的數(shù)據(jù)發(fā)送到云端會更高效。
• 通信：對于物聯(lián)網(wǎng)設(shè)備之間的通信，物理層與通信協(xié)議之間是有區(qū)別的。就物理層而言，網(wǎng)關(guān)，移動設(shè)備，網(wǎng)狀網(wǎng)絡(luò)（Mesh network）以及直接設(shè)備或廣播設(shè)備的通信是可選方案，根據(jù)實(shí)際用例可能適用或不適用。物理層的選擇將決定哪種通信協(xié)議最為合適（例如 MQTT，COAP，HPPT(S)，AMQP，ZigBee，Z-Wave 等）。

IoT Software

（物聯(lián)網(wǎng)軟件）

The next layer is the (cloud) platform that brings essential IoT software services together to manage the IoT endpoints securely, represent the ‘digital twin’ of connected things, process and analyze data, and provide APIs to consume and expose services:

• Device management: simplifies the process of configuring, provisioning, and operating the endpoint devices. It supports monitoring, testing, updating software, and troubleshooting connected devices.
• Digital twin management: For many IoT use cases, particularly in industrial IoT, it’s valuable to define a digital twin of the connected thing. This could be as simple as a 1:1 mapping of the physical things to logical identifiers in the IoT Platform, or as sophisticated as mapping an engineering view of an asset with a hierarchical structure of components/systems to the physical devices representing that asset on an instance and class level.
• Event and data processing: Event Processing deals with event streams coming from connected devices, filtering, and monitoring. In addition, services for data aggregation, data storage, and management are required.
• Analytics/machine learning: Analytics services perform statistical analysis and apply machine learning to detect patterns on a device instance or class level for predictive maintenance, making recommendations, triggering engineering changes, etc.
• API management: provides openness on all layers in the IoT platform for device communication, data-, service-, and backend integration, and application development.
• Security management: ensures that IoT endpoints do not expose security threats due to the increased attack surface IoT creates. IoT devices generate sensitive information about operations transmitted over the internet. Also, devices themselves are vulnerable to hacks that could cause serious business damage. Security services should include (certificate-based) device attestation, network connectivity, software upgrades, authentication, identity and access management, and data loss prevention.

往下一層就是（云端）平臺，它們將基本的物聯(lián)網(wǎng)軟件服務(wù)整合到一起，從而安全地管理物聯(lián)網(wǎng)終端、表示連接物的 “數(shù)字雙胞胎”、處理和分析數(shù)據(jù)，以及提供消費(fèi)和公開服務(wù)的 API：

• 設(shè)備管理：簡化配置、服務(wù)開通（Provisioning）和操作終端設(shè)備的步驟。它支持監(jiān)視、測試、更新軟件，以及對連接設(shè)備進(jìn)行故障排除。
• 數(shù)字雙胞胎（Digital twin）管理：對許多物聯(lián)網(wǎng)用例來說（尤其是在工業(yè)物聯(lián)網(wǎng)中），定義連接物的數(shù)字雙胞胎是很有價值的。這可能與物聯(lián)網(wǎng)平臺中，從物理事物到邏輯標(biāo)識符的一一映射那樣簡單，或者像將具有組件/系統(tǒng)分層結(jié)構(gòu)的資產(chǎn)工程視圖映射到一個實(shí)例以及類級別（Class level）的，代表該資產(chǎn)的物理設(shè)備一樣復(fù)雜。
• 事件與數(shù)據(jù)處理：事件處理涉及到來自連接設(shè)備的事件流，過濾以及監(jiān)視。另外，還需要用于數(shù)據(jù)聚合（Data aggregation），數(shù)據(jù)存儲和管理的服務(wù)。
• 分析/機(jī)器學(xué)習(xí)：分析服務(wù)（Analytics services）執(zhí)行統(tǒng)計(jì)分析，并應(yīng)用機(jī)器學(xué)習(xí)來檢測一個設(shè)備實(shí)例（或類級別）上的模式，從而進(jìn)行預(yù)見性維護(hù)（Predictive maintenance）、提出建議，和觸發(fā)工程變更等等。
• API（Application Programming Interface，應(yīng)用編程接口） 管理：在物聯(lián)網(wǎng)平臺的所有層面上，為設(shè)備通信，數(shù)據(jù)，服務(wù)和后端集成以及應(yīng)用程序開發(fā)提供開放性。
• 安全管理：確保物聯(lián)網(wǎng)終端不會因物聯(lián)網(wǎng)創(chuàng)造的攻擊面的增加而暴露在安全威脅之下。物聯(lián)網(wǎng)設(shè)備會生成有關(guān)操作的敏感信息，并通過互聯(lián)網(wǎng)進(jìn)行傳輸。并且設(shè)備本身易受黑客攻擊，這可能會導(dǎo)致嚴(yán)重的商業(yè)損失。安全服務(wù)應(yīng)包括（基于證書的）設(shè)備認(rèn)證、網(wǎng)絡(luò)連接、軟件升級、身份認(rèn)證、身份與訪問權(quán)限管理，以及數(shù)據(jù)丟失防護(hù)。

IoT Apps

（物聯(lián)網(wǎng)應(yīng)用程序）

The Apps layer is where IoT solutions are brought to life, turning data into actionable insight, putting it in the hands of business users, customers and partners. This is the layer where integrations with existing back-ends and 3rd party services are established and workflows are defined to act upon insight. Core services in the apps layer include:

• Integrated development environment (IDE): A design time environment is required to develop IoT apps. This could be a traditional IDE for coding in a specific language or a model-driven environment for collaborative, visual development of IoT apps. In addition, core services for software configuration management and branching & merging are needed for development teams to commit their work, and create builds and application packages. Finally, the IDE should guide developers to apply the right patterns and best practices for IoT app development.
• Multi-channel apps: In today’s world of web and mobile apps, the IDE ideally supports development of cross-platform, responsive and multi-channel apps, optimized for specific form factors, using device features and supporting gestures with minimal overhead.
• Integration: The lifeblood of IoT apps. Apps should have access to IoT endpoints (via the digital twin) for reading the full history of a ‘thing’ after receiving an alert, or triggering an actuator. They should be able to leverage various IoT software services (e.g. time series data and machine learning algorithms) and weave these services into IoT apps. Last but not least, integration with enterprise back-ends and 3rd party services is needed for managing workflows and making IoT apps contextual e.g. by creating a dashboard for a service engineer, enhanced with engineering and customer support data.
• Testing: Testing & quality assurance are essential disciplines in IoT app development projects. Test automation on various levels (unit test, integration, functional test) helps minimize the test burden relative to (iterative) development cycles.
• Deployment: Staged deployment to target environments and automated provisioning of application resources (web server, OS, database, file storage) helps DevOps engineers to efficiently manage IoT apps. Ideally, there’s flexibility to deploy on a cloud of choice—for instance, close to where core IoT services that the application uses are running.
• Management: User management, application management, monitoring, and self-service options for horizontal/vertical scaling and configuring high-availability are important to manage IoT apps. Specifically, support for elasticity backed by a stateless application architecture is essential to deal with variable load and volume.
• Security: Like for the IoT platform layer, security on an app level is vital. This concerns both the application runtime environment and the security settings for the apps themselves (e.g. access and authentication).

應(yīng)用層是物聯(lián)網(wǎng)解決方案得以實(shí)現(xiàn)的地方，在這一層，數(shù)據(jù)轉(zhuǎn)化為可操作的洞察力，并將其交付給企業(yè)用戶、客戶與合作伙伴。該層集成了現(xiàn)有的那些后端，建立了第三方服務(wù)，并定義了工作流以根據(jù)洞察力行事。應(yīng)用層的核心服務(wù)包括：

• 集成開發(fā)環(huán)境（IDE）：開發(fā)物聯(lián)網(wǎng)應(yīng)用程序需要有一個設(shè)計(jì)期環(huán)境（Design-time environment）。這個環(huán)境可以是一個傳統(tǒng)的 IDE，它能夠用于以特定語言進(jìn)行編程，或者是用于協(xié)作、可視化開發(fā)物聯(lián)網(wǎng)應(yīng)用程序的模型驅(qū)動環(huán)境（Model-driven environment）。另外，開發(fā)團(tuán)隊(duì)需要軟件配置管理（Software configuration management）以及分支與合并（Branching & Merging）的核心服務(wù)，以便提交他們的工作、創(chuàng)建構(gòu)建和應(yīng)用程序包。最后，IDE 應(yīng)引導(dǎo)開發(fā)人員應(yīng)用物聯(lián)網(wǎng)應(yīng)用程序開發(fā)的正確模式和最佳范例。
• 多通道應(yīng)用程序：當(dāng)今世界上的 Web 與移動應(yīng)用程序中，IDE 理想地支持跨平臺、響應(yīng)式和多通道應(yīng)用程序的開發(fā)，針對特定外形因素進(jìn)行優(yōu)化，并且以最小的開銷使用設(shè)備特性與其所支持的姿態(tài)。
• 集成：這是物聯(lián)網(wǎng)應(yīng)用的命根子（Lifeblood）。應(yīng)用程序應(yīng)該能夠（通過數(shù)字雙胞胎）訪問物聯(lián)網(wǎng)終端，以便在收到警報(bào)或觸發(fā)執(zhí)行器后能夠讀出 “事物” 的完整歷史信息。它們應(yīng)該能夠利用各種物聯(lián)網(wǎng)軟件服務(wù)（例如時間序列數(shù)據(jù)和機(jī)器學(xué)習(xí)算法）并將這些服務(wù)編排到物聯(lián)網(wǎng)應(yīng)用程序中。最后一點(diǎn)（但并非最不重要的一點(diǎn)）是，集成企業(yè)后端與第三方服務(wù)，以管理工作流并使物聯(lián)網(wǎng)應(yīng)用程序具有上下文（例如，通過為維護(hù)工程師創(chuàng)建一個儀表盤，增強(qiáng)了工程和客戶支持?jǐn)?shù)據(jù)）。
• 測試：測試（Testing）與品質(zhì)保證（Quality assurance）是物聯(lián)網(wǎng)應(yīng)用程序開發(fā)項(xiàng)目中的重要科目。不同級別上的測試自動化（單元測試，集成，功能測試）有助于最大限度地減少相對于（迭代）開發(fā)周期的測試負(fù)擔(dān)。
• 部署：分階段部署到目標(biāo)環(huán)境，以及自動化配置應(yīng)用程序資源（Web 服務(wù)器，操作系統(tǒng)，數(shù)據(jù)庫，文件存儲）可幫助開發(fā)運(yùn)維工程師（DevOps）有效地管理物聯(lián)網(wǎng)應(yīng)用程序。理想情況下，在所選擇的云端上可以靈活地部署 —— 比如說，接近應(yīng)用程序使用的核心物聯(lián)網(wǎng)服務(wù)的運(yùn)行狀態(tài)。
• 管理：水平/垂直擴(kuò)展與配置高可用性的用戶管理、應(yīng)用程序管理、監(jiān)控和自助服務(wù)選項(xiàng)，這對于管理物聯(lián)網(wǎng)應(yīng)用程序來說非常重要。具體說來，通過無狀態(tài)應(yīng)用架構(gòu)（Stateless application architecture）提供的彈性支持對于處理可變負(fù)載與可變?nèi)莘e來說是很有必要的。
• 安全：就像物聯(lián)網(wǎng)平臺層一樣，安全性對一個應(yīng)用來說至關(guān)重要。這同時涉及到應(yīng)用程序的運(yùn)行時環(huán)境，及其本身的安全設(shè)置（例如訪問和認(rèn)證）。

The elements that define the anatomy of an IoT solution may come across as overwhelming. As mentioned before, the type and level of sophistication of the IoT solution will determine how many of the elements and services described are needed to create an end-to-end solution.

定義一份物聯(lián)網(wǎng)解決方案解析的元素可能會遇到巨大的挑戰(zhàn)。如上所述，物聯(lián)網(wǎng)解決方案的復(fù)雜性類型與級別將決定需要多少基礎(chǔ)和服務(wù)來創(chuàng)建端到端解決方案。

Nevertheless, it’s clear that the diverse set of endpoints, network technologies, protocols, IoT software, and application development services pose a challenge for enterprises planning to adopt IoT to transform their business operations. The question is: How do you make IoT solution development manageable? The answer lies in adopting a platform approach.

盡管如此，很明顯，各種各樣的終端、網(wǎng)絡(luò)技術(shù)、協(xié)議、物聯(lián)網(wǎng)軟件以及應(yīng)用程序開發(fā)服務(wù)，這些對于計(jì)劃采用物聯(lián)網(wǎng)來轉(zhuǎn)變業(yè)務(wù)運(yùn)營的企業(yè)來說都是一個挑戰(zhàn)。那么問題來了：如何使得物聯(lián)網(wǎng)解決方案開發(fā)是易于管理的？其答案就在于采用一個平臺方案。

總結(jié)

以上是生活随笔為你收集整理的[物联网文章之其三] 剖析物联网解决方案的全部內(nèi)容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯，歡迎將生活随笔推薦給好友。