{"type":"doc","content":[{"type":"heading","attrs":{"align":null,"level":1}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2021 亞馬遜雲科技 re:Invent 剛剛落下帷幕，許多開發者都在討論今年亞馬遜雲科技 re:Invent 的重磅發佈。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這並不奇怪，每年亞馬遜雲科技 re:Invent 都會有數十項新產品、新功能發佈，其中一部分，可能代表了日後雲計算及整個 IT 界基礎設施的發展方向，比如曾經的 Amazon RedShift、Amazon Lambda，前者引導了業內雲原生數倉的發展，後者則把無服務器帶進了業內開發者的視線。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"但今年的亞馬遜雲科技 re:Invent 有些不一樣，除了在產品性能上的常規迭代，更重要的是體現了雲計算的服務在概念上的延伸。最爲典型的，便是 Amazon IoT TwinMaker 與 Amazon IoT FleetWise。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"元宇宙與物聯網：世界真的在雲化"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"據官方介紹，Amazon IoT TwinMaker 是一款可以讓開發人員更加輕鬆、快捷地創建現實世界的數字孿生，如樓宇、工廠、工業設備和生產線。用戶可以通過 Amazon IoT TwinMaker，使用數字孿生來構建反映現實世界的應用程序，提高運營效率並減少停機時間。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"數字孿生是物理系統的虛擬映射，可根據其所代表的現實世界對象的結構、狀態和行爲定期更新。Amazon IoT TwinMaker 讓開發人員可以輕鬆彙集來自多個來源（如設備傳感器、攝像機和業務應用程序）的數據，並將這些數據結合起來創建一個知識圖譜，對現實世界環境進行建模。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"數字孿生技術最早是用於航空航天飛行器的健康維護與保障，屬於冷門技術。但隨着“元宇宙”概念的興起，數字孿生技術越來越爲人所熟知，因爲數字孿生的本質特徵是在信息世界對物理世界進行等價映射，因此成爲元宇宙概念的重要支撐技術之一，尤其是工業元宇宙，應用更是廣泛。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"亞馬遜雲科技 re:Invent 這次發佈 Amazon IoT TwinMaker 可以說，既與元宇宙相關，也與工業互聯網有關。從前，雲計算的服務範圍，聚焦在互聯網產業，面向所謂的傳統產業，主要提供轉型服務。但現在，雲計算的服務外延正在飛速擴展，通過元宇宙這座橋，已將範疇擴大至整個物理世界的虛擬化映射。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"與此緊密相關的是 Amazon IoT FleetWise。Amazon IoT FleetWise 使汽車製造商可以輕鬆地收集和管理汽車中任何格式的數據（無論品牌、車型或配置），並將數據格式標準化，方便在雲上輕鬆進行數據分析。當數據進入雲端後，汽車製造商就可以將數據應用於車輛的遠程診斷程序，分析車隊的健康狀況，幫助汽車製造商預防潛在的召回或安全問題，或通過數據分析和機器學習來改進自動駕駛和高級輔助駕駛等技術。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果說 Amazon IoT TwinMaker 提供的是從現實世界到虛擬世界的映射服務，那麼 Amazon IoT FleetWise 就是聚焦車聯網領域，解決的是車聯網長期以來的發展問題。車聯網概念最早出現於 20 世紀 60 年代，但 60 年來，一直有點“瘸腿發展”的意思——大部分人對車聯網的理解是，在車內提供網絡服務，而不是將車輛數據上傳分析。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"直到 2012 年特斯拉 Model S 出世，才把車聯網作爲必選功能，納入汽車生產製造的流水線裏。現在 Amazon IoT FleetWise 發佈，把車聯網相關服務全面引上雲端。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Amazon IoT FleetWise 的一個重要特徵是，可以在雲中構建車輛的虛擬表示，並應用通用數據格式來構建和標記車輛屬性、傳感器和信號。Amazon IoT FleetWise 使用車輛信號規範 (VSS) 對車輛建模進行標準化，以便“燃料壓力”等信號始終表示爲燃料壓力，並以每平方英寸磅力 (PSI) 和千帕 (kPa) 爲單位進行測量。車輛建模後，上傳標準 CAN 數據庫 (DBC) 或 AUTOSAR XML (ARXML) 文件，以便 Amazon IoT FleetWise 可以讀取通過車輛控制器局域網總線 (CAN 總線) 發送的獨特的專有數據信號。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"看懂了吧，其實 Amazon IoT TwinMaker 和 Amazon IoT FleetWise 的底層理念如出一轍，都是在雲端構建虛擬映射，不過一個是針對工業領域，一個是針對汽車行業。可以說，這個世界正在虛擬化，同時也正在雲化。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"Amazon SageMaker Canvas：無代碼創建 ML 模型"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果說 Amazon IoT TwinMaker 和 Amazon IoT FleetWise 體現了雲服務在橫向上的概念延伸，那麼 Amazon SageMaker Canvas 則是在縱向上的概念延伸。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"大家都知道 Amazon SageMaker ，作爲一個已發佈四年的全託管機器學習服務。Amazon SageMaker 爲開發者提供了一套完備的“中央廚房”，使用 Amazon SageMaker 開發者只需準備好“食材”（數據）就可以直接開始做菜（訓練模型），大大提升了開發人員和數據科學家構建、訓練和部署機器學習模型的效率，開啓了全新的智能時代。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"但 AI 領域長期受到人才短缺問題的限制，而 AI 的應用領域正在增多，機器學習服務的門檻需要進一步降低。亞馬遜雲科技此次發佈 Amazon SageMaker Canvas 的目的即在於此 —— 用無代碼理念構建機器學習模型，做模型預測，保證在脫離數據工程團隊的情況下，依然可以提供服務。它利用與 Amazon SageMaker 相同的技術自動清理和組合您的數據，在幕後創建數百個模型，選擇性能最佳的模型，並生成新的單個或批量預測。支持二元分類、多類分類、數值迴歸、時間序列預測等多種問題類型。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/b0\/b0d72c06ee0096338f12780095197b68.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"此前業內還有許多關於低代碼、無代碼的爭議，但現在看來，這不是概念之爭，而是產業內有確實的需要。Amazon SageMaker Canvas 的發佈即是驗證了這一情況。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"站在整個 AI 的宏觀層面而言，無論是基於 AI 提供的預測服務還是分析服務，也脫離了單純對更高級別人工智能的追求，而兼顧 AI 能力對產業的賦能。這是雲計算對服務理念的進一步擴大和貫徹。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"Amazon Private 5G：用專有 5G 鏈接 IoT 設備"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"而在這種情況下， Amazon Private 5G 的發佈，自然也被許多人所關注，因爲它是支撐服務外擴的重要和必要嘗試。可以說，Amazon Private 5G 是本次 re:Invent 最重要的發佈之一。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在移動端，我們早就用上了 5G 通信服務，但企業需要的是專有 5G 服務網絡。Amazon Private 5G 可自動設置和部署網絡，並按需擴展容量以支持更多設備和網絡流量。亞馬遜雲科技 EC2 副總裁 David Brown 說：“藉助亞馬遜雲科技私有 5G，我們將混合基礎設施擴展到客戶的 5G 網絡，以簡化、快速且廉價地建立私有 5G 網絡。客戶可以從小規模開始，按需擴展，按需付費，並從亞馬遜雲科技控制檯監控和管理他們的網絡。”"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"而 Amazon Private 5G 也重點服務了以工業 4.0 爲主的龐大傳感器和端側設備集羣，前文提到的工業元宇宙、車聯網自然也在同一序列。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"全球最大的非上市公司——美國科氏公司工業集團（Koch）已經和亞馬遜雲科技就 Amazon Private 5G 達成了合作，而科氏企業集團的核心是石油與化工，也是亞馬遜雲科技比較有代表性的服務案例。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"Amazon Graviton3：底層算力又升級了"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當然，不管是 Amazon IoT TwinMaker 等 IoT 服務，還是 Amazon Private 5G，依靠的還是底層實例中的芯片性能。在今年的雲棲大會上，阿里巴巴旗下半導體公司平頭哥發佈自研雲芯片倚天 710 並宣佈其性能超過 Amazon Graviton2 20%。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"從2019年 Amazon Graviton，2020年 Amazon Graviton2 到今天的 Amazon Graviton3，亞馬遜雲科技從芯片開始不斷改進計算服務。與 Amazon Graviton2 相比，Amazon Graviton3 集成了550億個晶體管，單核"},{"type":"text","marks":[{"type":"strong"}],"text":"性能提升超過25%，浮點和加密性能將提高兩倍。"},{"type":"text","text":"在機器學習方面，Amazon Graviton3 包括對 bfloat 16數據的支持，將能夠提供高達3倍的性能。"},{"type":"text","marks":[{"type":"strong"}],"text":"在性能飛躍的同時，Amazon Graviton3 能耗對比上一代產品下降了60%。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/8c\/8c4cbfc9f69c0191b98cc36972359d6c.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"新一代Amazon EC2 C7g 實例由 Amazon Graviton3 處理器支持，也是全球首個支持DDR5內容的雲上計算實例。與由Amazon Graviton2 處理器支持的當前一代 Amazon C6g 實例相比，性能提高 25%。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/15\/15ec15dff34cd3774a6ef4c3e92530ca.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當然，Amazon Graviton3 屬於通用芯片，專用芯片也更新了。亞馬遜雲科技公佈，由亞馬遜第二款機器學習芯片 Amazon Trainium 支持的新 Amazon Trn1 實例，將爲在雲中爲自然語言處理 (NLP)、計算機視覺、搜索、推薦、排名等用例訓練深度學習模型提供最佳的性價比，與 P4d 實例相比，通過 Amazon Trn1 實例訓練深度學習模型的成本降低達到 40%。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Amazon Nitro System 芯片也發佈了新品。Amazon Nitro 可以說是個超級黑科技，準確來講它是一個套架構，可以將服務器提供給用戶的資源最大化，減少虛擬化損耗。所謂“虛擬化損耗”，就是以往爲了維護服務其正常運行，在網絡、存儲、管理等系統功能方面做出的必然開銷，這種開銷要佔到服務器總體性能的三成。Nitro 架構就是通過定製硬件，關注這三成的性能問題。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"而這次發佈的 Amazon Nitro System 芯片，主要是支持 Amazon EC2 Instance 底層管理平臺，可以替 CPU 分擔工作負載。通用芯片、推理專用芯片、 Amazon EC2 支持芯片，這次發佈的是整整齊齊。Amazon Nitro SSDs 的 Im4gn\/Is4gen\/ I4i 實例提供 30 TB 的 NVMe 存儲，與上一代 I3 實例相比，I\/O 延遲降低了 60%，延遲可變性降低了 75%。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"Data Serverless ：無服務器應用迅速推廣"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當然，除了在網和端層面的服務能力擴展，此次在 Serverless 層面的更新也值得注意。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"業內都知道 Amazon Lambda 開啓了 Serverless 的時代，但真正獲得業內的廣泛贊同和跟隨，還是在 2019 年。這次 亞馬遜雲科技 re:Invent ，亞馬遜雲科技一口氣發佈了四個核心產品的 Serverless 版本：Amazon Redshift Serverless、Amazon EMR Serverless、Amazon MSK Serverless 和Amazon Kinesis data streams on-demand。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Amazon Redshift 我們已經提過，是最早的雲原生數倉；Amazon EMR 則是亞馬遜雲科技提供的 Hadoop 託管服務；Amazon MSK 是 Kafka 託管服務；而 Amazon Kinesis data streams on-demand 則是流式數據處理平臺。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/e4\/e4b3f278f30a1aba957d267fd4bee4da.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"而這些服務的 Serverless 版本則是讓使用者單擊幾下即可運行使用這些框架構建的應用程序，而無需配置、優化或保護集羣。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"雲端大數據架構，因爲亞馬遜雲科技這些 Serverless 版本服務的更新，其門檻正在飛速降低。以往像智能湖倉這種架構的搭建，讓架構師、工程師很頭疼，而現如今，工程師的工作正在變成單純的調參 —— Serverless 對產業生態的改變幾乎是永久性的。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"寫在最後"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"從 2021 亞馬遜雲科技 re:Invent 及今年各雲計算大會的發佈情況看，雲計算領域的產品更迭，一是注重底層基礎算力的升級，這屬於硬核實力的比拼；二則是注重服務外沿的擴展，如何理解雲、網、端三個角色，並提供儘量通用的公有云服務，成爲關鍵。元宇宙則是當下新興起的概念，爲整體的技術發展方向提供了新的想法和方向，值得我們特別思考。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}}]}
2021 re:Invent ，我們到底該關注哪些發佈？
{"type":"doc","content":[{"type":"heading","attrs":{"align":null,"level":1}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"2021 亞馬遜雲科技 re:Invent 剛剛落下帷幕，許多開發者都在討論今年亞馬遜雲科技 re:Invent 的重磅發佈。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"這並不奇怪，每年亞馬遜雲科技 re:Invent 都會有數十項新產品、新功能發佈，其中一部分，可能代表了日後雲計算及整個 IT 界基礎設施的發展方向，比如曾經的 Amazon RedShift、Amazon Lambda，前者引導了業內雲原生數倉的發展，後者則把無服務器帶進了業內開發者的視線。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"但今年的亞馬遜雲科技 re:Invent 有些不一樣，除了在產品性能上的常規迭代，更重要的是體現了雲計算的服務在概念上的延伸。最爲典型的，便是 Amazon IoT TwinMaker 與 Amazon IoT FleetWise。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"元宇宙與物聯網：世界真的在雲化"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"據官方介紹，Amazon IoT TwinMaker 是一款可以讓開發人員更加輕鬆、快捷地創建現實世界的數字孿生，如樓宇、工廠、工業設備和生產線。用戶可以通過 Amazon IoT TwinMaker，使用數字孿生來構建反映現實世界的應用程序，提高運營效率並減少停機時間。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"數字孿生是物理系統的虛擬映射，可根據其所代表的現實世界對象的結構、狀態和行爲定期更新。Amazon IoT TwinMaker 讓開發人員可以輕鬆彙集來自多個來源（如設備傳感器、攝像機和業務應用程序）的數據，並將這些數據結合起來創建一個知識圖譜，對現實世界環境進行建模。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"數字孿生技術最早是用於航空航天飛行器的健康維護與保障，屬於冷門技術。但隨着“元宇宙”概念的興起，數字孿生技術越來越爲人所熟知，因爲數字孿生的本質特徵是在信息世界對物理世界進行等價映射，因此成爲元宇宙概念的重要支撐技術之一，尤其是工業元宇宙，應用更是廣泛。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"亞馬遜雲科技 re:Invent 這次發佈 Amazon IoT TwinMaker 可以說，既與元宇宙相關，也與工業互聯網有關。從前，雲計算的服務範圍，聚焦在互聯網產業，面向所謂的傳統產業，主要提供轉型服務。但現在，雲計算的服務外延正在飛速擴展，通過元宇宙這座橋，已將範疇擴大至整個物理世界的虛擬化映射。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"與此緊密相關的是 Amazon IoT FleetWise。Amazon IoT FleetWise 使汽車製造商可以輕鬆地收集和管理汽車中任何格式的數據（無論品牌、車型或配置），並將數據格式標準化，方便在雲上輕鬆進行數據分析。當數據進入雲端後，汽車製造商就可以將數據應用於車輛的遠程診斷程序，分析車隊的健康狀況，幫助汽車製造商預防潛在的召回或安全問題，或通過數據分析和機器學習來改進自動駕駛和高級輔助駕駛等技術。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果說 Amazon IoT TwinMaker 提供的是從現實世界到虛擬世界的映射服務，那麼 Amazon IoT FleetWise 就是聚焦車聯網領域，解決的是車聯網長期以來的發展問題。車聯網概念最早出現於 20 世紀 60 年代，但 60 年來，一直有點“瘸腿發展”的意思——大部分人對車聯網的理解是，在車內提供網絡服務，而不是將車輛數據上傳分析。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"直到 2012 年特斯拉 Model S 出世，才把車聯網作爲必選功能，納入汽車生產製造的流水線裏。現在 Amazon IoT FleetWise 發佈，把車聯網相關服務全面引上雲端。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Amazon IoT FleetWise 的一個重要特徵是，可以在雲中構建車輛的虛擬表示，並應用通用數據格式來構建和標記車輛屬性、傳感器和信號。Amazon IoT FleetWise 使用車輛信號規範 (VSS) 對車輛建模進行標準化，以便“燃料壓力”等信號始終表示爲燃料壓力，並以每平方英寸磅力 (PSI) 和千帕 (kPa) 爲單位進行測量。車輛建模後，上傳標準 CAN 數據庫 (DBC) 或 AUTOSAR XML (ARXML) 文件，以便 Amazon IoT FleetWise 可以讀取通過車輛控制器局域網總線 (CAN 總線) 發送的獨特的專有數據信號。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"看懂了吧，其實 Amazon IoT TwinMaker 和 Amazon IoT FleetWise 的底層理念如出一轍，都是在雲端構建虛擬映射，不過一個是針對工業領域，一個是針對汽車行業。可以說，這個世界正在虛擬化，同時也正在雲化。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"Amazon SageMaker Canvas：無代碼創建 ML 模型"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"如果說 Amazon IoT TwinMaker 和 Amazon IoT FleetWise 體現了雲服務在橫向上的概念延伸，那麼 Amazon SageMaker Canvas 則是在縱向上的概念延伸。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"大家都知道 Amazon SageMaker ，作爲一個已發佈四年的全託管機器學習服務。Amazon SageMaker 爲開發者提供了一套完備的“中央廚房”，使用 Amazon SageMaker 開發者只需準備好“食材”（數據）就可以直接開始做菜（訓練模型），大大提升了開發人員和數據科學家構建、訓練和部署機器學習模型的效率，開啓了全新的智能時代。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"但 AI 領域長期受到人才短缺問題的限制，而 AI 的應用領域正在增多，機器學習服務的門檻需要進一步降低。亞馬遜雲科技此次發佈 Amazon SageMaker Canvas 的目的即在於此 —— 用無代碼理念構建機器學習模型，做模型預測，保證在脫離數據工程團隊的情況下，依然可以提供服務。它利用與 Amazon SageMaker 相同的技術自動清理和組合您的數據，在幕後創建數百個模型，選擇性能最佳的模型，並生成新的單個或批量預測。支持二元分類、多類分類、數值迴歸、時間序列預測等多種問題類型。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/b0\/b0d72c06ee0096338f12780095197b68.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"此前業內還有許多關於低代碼、無代碼的爭議，但現在看來，這不是概念之爭，而是產業內有確實的需要。Amazon SageMaker Canvas 的發佈即是驗證了這一情況。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"站在整個 AI 的宏觀層面而言，無論是基於 AI 提供的預測服務還是分析服務，也脫離了單純對更高級別人工智能的追求，而兼顧 AI 能力對產業的賦能。這是雲計算對服務理念的進一步擴大和貫徹。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"Amazon Private 5G：用專有 5G 鏈接 IoT 設備"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"而在這種情況下， Amazon Private 5G 的發佈，自然也被許多人所關注，因爲它是支撐服務外擴的重要和必要嘗試。可以說，Amazon Private 5G 是本次 re:Invent 最重要的發佈之一。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在移動端，我們早就用上了 5G 通信服務，但企業需要的是專有 5G 服務網絡。Amazon Private 5G 可自動設置和部署網絡，並按需擴展容量以支持更多設備和網絡流量。亞馬遜雲科技 EC2 副總裁 David Brown 說：“藉助亞馬遜雲科技私有 5G，我們將混合基礎設施擴展到客戶的 5G 網絡，以簡化、快速且廉價地建立私有 5G 網絡。客戶可以從小規模開始，按需擴展，按需付費，並從亞馬遜雲科技控制檯監控和管理他們的網絡。”"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"而 Amazon Private 5G 也重點服務了以工業 4.0 爲主的龐大傳感器和端側設備集羣，前文提到的工業元宇宙、車聯網自然也在同一序列。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"全球最大的非上市公司——美國科氏公司工業集團（Koch）已經和亞馬遜雲科技就 Amazon Private 5G 達成了合作，而科氏企業集團的核心是石油與化工，也是亞馬遜雲科技比較有代表性的服務案例。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"Amazon Graviton3：底層算力又升級了"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當然，不管是 Amazon IoT TwinMaker 等 IoT 服務，還是 Amazon Private 5G，依靠的還是底層實例中的芯片性能。在今年的雲棲大會上，阿里巴巴旗下半導體公司平頭哥發佈自研雲芯片倚天 710 並宣佈其性能超過 Amazon Graviton2 20%。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"從2019年 Amazon Graviton，2020年 Amazon Graviton2 到今天的 Amazon Graviton3，亞馬遜雲科技從芯片開始不斷改進計算服務。與 Amazon Graviton2 相比，Amazon Graviton3 集成了550億個晶體管，單核"},{"type":"text","marks":[{"type":"strong"}],"text":"性能提升超過25%，浮點和加密性能將提高兩倍。"},{"type":"text","text":"在機器學習方面，Amazon Graviton3 包括對 bfloat 16數據的支持，將能夠提供高達3倍的性能。"},{"type":"text","marks":[{"type":"strong"}],"text":"在性能飛躍的同時，Amazon Graviton3 能耗對比上一代產品下降了60%。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/8c\/8c4cbfc9f69c0191b98cc36972359d6c.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"新一代Amazon EC2 C7g 實例由 Amazon Graviton3 處理器支持，也是全球首個支持DDR5內容的雲上計算實例。與由Amazon Graviton2 處理器支持的當前一代 Amazon C6g 實例相比，性能提高 25%。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/15\/15ec15dff34cd3774a6ef4c3e92530ca.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當然，Amazon Graviton3 屬於通用芯片，專用芯片也更新了。亞馬遜雲科技公佈，由亞馬遜第二款機器學習芯片 Amazon Trainium 支持的新 Amazon Trn1 實例，將爲在雲中爲自然語言處理 (NLP)、計算機視覺、搜索、推薦、排名等用例訓練深度學習模型提供最佳的性價比，與 P4d 實例相比，通過 Amazon Trn1 實例訓練深度學習模型的成本降低達到 40%。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Amazon Nitro System 芯片也發佈了新品。Amazon Nitro 可以說是個超級黑科技，準確來講它是一個套架構，可以將服務器提供給用戶的資源最大化，減少虛擬化損耗。所謂“虛擬化損耗”，就是以往爲了維護服務其正常運行，在網絡、存儲、管理等系統功能方面做出的必然開銷，這種開銷要佔到服務器總體性能的三成。Nitro 架構就是通過定製硬件，關注這三成的性能問題。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"而這次發佈的 Amazon Nitro System 芯片，主要是支持 Amazon EC2 Instance 底層管理平臺，可以替 CPU 分擔工作負載。通用芯片、推理專用芯片、 Amazon EC2 支持芯片，這次發佈的是整整齊齊。Amazon Nitro SSDs 的 Im4gn\/Is4gen\/ I4i 實例提供 30 TB 的 NVMe 存儲，與上一代 I3 實例相比，I\/O 延遲降低了 60%，延遲可變性降低了 75%。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"Data Serverless ：無服務器應用迅速推廣"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"當然，除了在網和端層面的服務能力擴展，此次在 Serverless 層面的更新也值得注意。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"業內都知道 Amazon Lambda 開啓了 Serverless 的時代，但真正獲得業內的廣泛贊同和跟隨，還是在 2019 年。這次 亞馬遜雲科技 re:Invent ，亞馬遜雲科技一口氣發佈了四個核心產品的 Serverless 版本：Amazon Redshift Serverless、Amazon EMR Serverless、Amazon MSK Serverless 和Amazon Kinesis data streams on-demand。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Amazon Redshift 我們已經提過，是最早的雲原生數倉；Amazon EMR 則是亞馬遜雲科技提供的 Hadoop 託管服務；Amazon MSK 是 Kafka 託管服務；而 Amazon Kinesis data streams on-demand 則是流式數據處理平臺。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https:\/\/static001.geekbang.org\/wechat\/images\/e4\/e4b3f278f30a1aba957d267fd4bee4da.png","alt":null,"title":null,"style":null,"href":null,"fromPaste":false,"pastePass":false}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"而這些服務的 Serverless 版本則是讓使用者單擊幾下即可運行使用這些框架構建的應用程序，而無需配置、優化或保護集羣。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"雲端大數據架構，因爲亞馬遜雲科技這些 Serverless 版本服務的更新，其門檻正在飛速降低。以往像智能湖倉這種架構的搭建，讓架構師、工程師很頭疼，而現如今，工程師的工作正在變成單純的調參 —— Serverless 對產業生態的改變幾乎是永久性的。"}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"寫在最後"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"從 2021 亞馬遜雲科技 re:Invent 及今年各雲計算大會的發佈情況看，雲計算領域的產品更迭，一是注重底層基礎算力的升級，這屬於硬核實力的比拼；二則是注重服務外沿的擴展，如何理解雲、網、端三個角色，並提供儘量通用的公有云服務，成爲關鍵。元宇宙則是當下新興起的概念，爲整體的技術發展方向提供了新的想法和方向，值得我們特別思考。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}}]}
發表評論
所有評論
還沒有人評論，想成為第一個評論的人麼? 請在上方評論欄輸入並且點擊發布.
相關文章
4場和Zabbix春日約會，“承包”四月天
在全國各地和Zabbix相遇互動，4月活動集錦，來約會吧！ 時間 活動名稱 城市 4/20（週六） KCD雲原生社區日 上海 4/20（週六） OceanBase開發者大會