脈絡拆解: AWS re:Invent 2022 Adam Selipsky Keynote

Published: 2022-11-29

Lastmod: 2024-04-16

起因

自從 2020 年初獲選成為 AWS Community Hero 之後，就遇上了疫情，三年沒飛。今年很榮幸受邀能夠實體登入位於 Las Vegas 的 AWS 年度開發者大會 re:Invent 2022，能夠再次體驗各種現場活動，相當開心。

前幾年開始比較認真切進去看各家開發者大會、技術大會、研討會的主題演講 (keynote) 等內容，特別是 AWS re:Invent 相關內容，開始比較會抓得到大架構、大綱，也能感受到為什麼某些題目與題目之間的順序要作調整、哪些議題在相比較之下會先被拉掉等等。

2022 年的這場主題演講，是 AWS CEO Adam Selipsky 的第二場 re:Invent 主題演講。接續他前任 CEO Andy Jassy 在 2020 提到「Still early days for cloud」，Adam 第一年的時候，承接以「It’s still early days」開場，而 Adam 的第二年，開始走出自己的風格。

跟以往一樣，全文開始我會先試著抓出演講架構，然後放一些觀察與推論，接著各段落放一些流水筆記，方便未來搜尋使用。文末整理有延伸閱讀，可以豐富大家對於演講內容的背景情境或資訊。也希望有機會帶大家一起推理為什麼這個時間點會推出這個產品或功能、全球各種產業的趨勢方向往哪裡走、台灣產業關注的議題方向不一定相同，但多看看、多比較、多參考，也許能避開一些誤判的雷區，在資源有限的情況下，如果走錯方向，市場不一定會給機會讓我們再砍掉重練。推理不一定是正確，就當作是種練習與分享。

新服務或新屬性，在本文中使用 [NEW 🚀] 方式標記，方便各位按下 Command/Ctrl+F 內頁搜尋。

本文刻意將大部分產品連結都先拿掉，讓大家能夠專注地閱讀（這年頭我們都少了些專注，是吧？）。若有需要產品連結，可以參閱我平常整理的 AWS 產品清單一覽表。

也歡迎大家給我一些回饋或指正。那我們就開始吧！

內容大綱

摘要 tl;dr

打好基礎釜底抽薪 —> 拆開非同步變簡單 —> 將所有東西連起來 —> 落地執行解決問題。

也分別可以對應到另外三場主題演講：

打好基礎釜底抽薪 👉 脈絡拆解: AWS re:Invent 2022 Monday Night Live Keynote With Peter Desantis
拆開非同步變簡單 👉 脈絡拆解: AWS re:Invent 2022 Werner Vogels Keynote
將所有東西連起來 👉 脈絡拆解: AWS re:Invent 2022 Swami Sivasubramanian Keynote
落地執行解決問題 👉 脈絡拆解: AWS re:Invent 2022 Adam Selipsky Keynote

架構

今年 Adam Selipsky Keynote 兩個小時的主題演講架構段落，順序剛好跟去年主題演講的順序整個相反過來，他沒有安排主題，但是藉由介紹四種環境，對應到四個段落。

這四個段落分別為：

Vast 廣闊 👉 Space 太空 👉 Data 資料
Unfathomable 深不可測 👉 Deep Sea 深海 👉 Security 安全
Extreme 極端 👉 Antarctica 南極 👉 Compute 運算
Possibilities 各種可能性 👉 Imagination 想像 👉 Application 產業應用

開場

各種客戶的各種應用情境
- BMW Group —> 工業
- RIOT Games —> 遊戲
- PrivatBank ¹ —> the largest Ukrainian bank —> 銀行
- Nasdaq —> 金融交易
- Startups —> 不只大型企業，依據 PitchBook ² 的資料，一千多家獨角獸公司裡頭有 83% 都在 AWS 運行。
現在處於領導地位，所以為下一代著想
- 100% renewable energy by 2025 —> ESG
- Water positive by 2030 —> ESG
Reduce cost —> 勉強算是延續去年的 IT still early 主題。不斷為客戶減少支出，也是整個亞馬遜飛輪的步驟之一。
- Carrier: 40% reduction in costs of ERP
- GILEAD: more than 60M saving
- AGCO: 78% reduction in costs
Flexibility to scale up or down
- Airbnb details road map to lower cloud costs —> savings of USD 63.5M 在疫情期間用量降低，所以雲端成本也馬上跟著減少 27%。
Innovate faster
Be prepared for anything

1. Vast 廣闊 👉 Space 太空 👉 Data 資料

望遠鏡
- 1609 年，伽利略拿到凸透鏡改良望遠鏡的故事、望遠鏡 10x、
- 花了好幾個世紀人類才進到火箭世代、
- 然後我們將望遠鏡送上太空，直接偵測各種可見光與不可見光、2014 哈伯望遠鏡、2019 詹姆斯·韋伯太空望遠鏡
- 1,000,000,000,000,000,000,000,000 stars
- 為了探索這廣闊的星空，人類需要各式各樣的工具，包含發射各種火箭、衛星、打造太空站等等 —> 需要各種系統、各種團隊一起合作，才能拼湊出來解讀各種觀測數據
Data explosion
- 如何管理這大量且不斷產生的資料，是個問題。
- The vast realm… of data
  - expedia group - 600B AI prediction per year
  - SAMSUNG - 1.1 B users make 80,000 requests per second
  - Pinterest - 1 exabyte of data
Exploring the vast data realm
- Right tools (Databases + Analytics + ML/AI)
  - Databases
    - Amazon Aurora —> 先講相容，再講效能比原本的還快數倍
  - Analytics
    - AWS Analytics Services —> 故意留了一個在右下角沒有「Serverless」，其他全部都有
    - Amazon Redshift
    - [NEW 🚀] OpenSearch Service Serverless (Preview) —> 果然
  - ML/AI
    - AI services
    - Amazon SageMaker
  - Guest: Biljana Kaitovic, EVP, Digital & IT and Group CIO, ENGIE
    - energy provider in France
    - World Energy Actor | ENGIE
    - 1882 年發明交流電的特斯拉
    - 9,000 in 1882 —> 180,000 terawatt-hours TODAY —> 270,000 in 2050 (50% increase)
    - Accessible, affordable, secure, green
    - data and digital play a key role in energy transition —> 難怪今年跑出一堆 energy 相關議程?!
    - 1 - optimizing renewable energies
    - 2 - managing decentralized energy - 各家屋頂上的太陽能發電 - 提到 blockchain
    - 3- reducing energy usage - IoT
    - more than 1 petabyte data
    - commonDataHub
    - 60% savings on runtime in the cloud
- Integration
  - combine the data from different data source
  - ETL —> 惡夢
  - “… thankless, unsustainable black hole”
  - A better way
    - (4 points here)
  - A Zero ETL future…
  - [NEW 🚀] Amazon Aurora zero-ETL integration with Amazon Redshift (Available in Preview Today)
    - Near real-time analytics and ML on transactional data
    - Consolidate data from multiple Aurora databases
    - Updates automatically and continuously
    - Serverless - no infra to manage
  - Apache Spark
  - [NEW 🚀] Amazon Redshift integration for Apache Spark (GA)
- Governance
  - Control vs Access
    - Too much control —> Silos
    - Too much access
  - [NEW 🚀] Amazon DataZone (Coming Soon)
    - Bringing data and people together
      - data producers
        Set up business data catalog
        Populate with metadata, labels…(缺)
      - data consumers
        search and browse catalog, request access
        Define data projects
    - Analytics integrations
      - Redshift, QuickSight, Athena, Partner API
- Insights
  - Amazon QuickSight
    - [NEW 🚀] Operational paginated reports
  - Amazon QuickSight Q - 直接問 Q 問題
    - [NEW 🚀] ML-powered forecasting with Q (GA Today)
    - [NEW 🚀] "Why" questions with Q (GA Today)
- The vast realm of data summary 收尾 —> 準備進入下一個 realm?

2. Unfathomable 深不可測 👉 Deep Sea 深海 👉 Security 安全

壓力艙, 聲納
Confidence to explore —> 講回去 Digital transformation
- Secure Global Infrastructure
- Secure and agile
Confidence to build and explore
- Identity and access mgmt
- Detection and response
  - Amazon GuardDuty
    - Amazon GuardDuty EKS Protection <— Amazon GuardDuty now protects Amazon Elastic Kubernetes Service clusters
    - [NEW 🚀] Container runtime threat detection for GuardDuty
- Network and application protection
- Data protection
Thousands of security solutions - 在 Marketplace 上面有一堆安全解決方案
- OCSF - Open Cybersecurity Schema Framework
- [NEW 🚀] Amazon Security Lake (Available in Preview Today)
  - —> 好奇是不是 partner solutions 放上 marketplace 後，用量不夠多？乾脆重新打包，讓大家更方便取用，雙贏。
  - 分成四個象限來拆解：AWS vs Partner + Source vs Analytics
    - Memo: Adam 都有隨著投影片的版面，適當走位耶！:p

3. Extreme 極端 👉 Antarctica 南極 👉 Compute 運算

Memo: 這段的開場影片，在現場真的有配合打超亮閃燈模擬閃電！
南極洲
- Captain Robert F. Scott
- Roald Amundsen
- 小馬 ponies, 雪地狗
“Good enough” simply isn’t good enough
- F1 - Reduced simulation time by 70%
- Epic games - 100M+ players supported (Real-time gaming)
- Nielsen - Hundred of billions of events per day
600+ instance types （去年 475）
AWS Nitro
Graviton3 - 25% faster than Graviton2
- directTV use Graviton3 to support 5G network
- [NEW 🚀] C7gn Instance for EC2, Powered by Graviton3 (Available in Preview Today)
- Memo: EC2 instance type 名字裡面有 g
Machine Learning
- Training —> 訓練 model
  - Trn1 Instance for EC2 -50% lower training cost —> 去年發表的
    - qualtrics XM - 40% reduction in cost
- Inference —> 用 model 跑預測
  - [NEW 🚀] Inf2 Instances for EC2 (Available in Preview Today)
High Performance Computing (HPC)
- Dana-Farber Cancer Institute - 48x more target drug molecules identified
- Designed for HPC
- EC2 Instances for HPC workloads
  - Compute intensive - Hpc6a
  - Compute and network intensive - [NEW 🚀] Hpc7g (Coming soon)
  - Data and memory intensive - [NEW 🚀] Hpc6id (GA Today)
Guest: Cedrik Neike, CEO, Digital Industrials, SIEMENS
- Space Perspective - Revolutionizing space travel —> Space Perspective dot com
- 1- Siemens software on AWS
  - Siemens star-CMM+ Multiphsics simulation software
  - Amazon EC2 c5 instances
- 92% Fortune 500 公司使用 Siemen Software
- 24 of 25 cars manufacturers use our software
- Dream it —> Make it
- Industrial Revolution 4.0
  - Siemens + AWS
  - AWS has helped Siemen move industrial software to the cloud —> 昨天在路上和 Arm 的人聊到他們也有個部門和 AWS 合作，將兩邊軟體（韌體+雲端）串接起來
  - A modern factory produces 2,200 TB of data per month (OT side)
  - Unlocking the potential of industrial data
  - example: NORTHVOLT
    - Factory gateway —> Kinesis —> Lambda —> DynamoDB —> RDS + S3 —> Athena
  - We scale good ideas
    - MENDIX - 200,000 users?
- 收尾
  - 1- Siemens software on AWS
  - 2- Combinging AWS cloud services with industrial data
  - 3- Enabling builders with low-code application tools
Spatial Simulation 空間模擬
- 3D engines - Unreal Engine & Unity
- 車流控制模擬
- [NEW 🚀] AWS SimSpace Weaver (GA) - Run massive spatial simulations without managing infra
All your extremes —> 前一段收尾?!

4. Possibilities 各種可能性 👉 Imagination 想像 👉 Application 產業應用

Imagination not only lets us create new worlds, it lets us live in them. –> 以為要開始講 web3、元宇宙 XDD 但其實是真實客戶在各個產業領域的應用場景，每次 keynote 最喜歡聽這一段，離現場比較近距離。
Removing constraints
- Amazon Connect - Contact center in the cloud
  - 10M+ Connect interactions per day
  - priceline 有上千人晚上在家接線
  - [NEW 🚀] ML driven forecasting, capacity planning, and scheduling (GA)
  - [NEW 🚀] Contact Lens with agent performance management (Available in Preview Today)
  - [NEW 🚀] Agent workspace with guided step-by-step actions (Available in Preview Today)
  - (今年有罐頭掌聲耶?!)
- Supply chain complexity
  - [NEW 🚀] AWS Supply Chain (Available in Preview Today)
    - Unified UI
    - Select data source e.g. EDI, SAP, S3 –> 以後上架 Amazon.com （應該）就可以輕鬆串接 EDI 惹！ :p
    - Data lake
    - Inventory
    - ML prediction
      - New insights
      - In review
Combining disparate ideas
- Protected environments for analyzing data
  - [NEW 🚀] AWS Clean Rooms (Preview in a few weeks)
    - 清資料的好地方! –> 以後進去無塵室也可以改成呼叫 API 嗎？（歪樓 :p
    - Market behavior data, reporting data, user data —> AWS Clean Rooms
    - Relevant campaigns while protecting customer data
    - Clean room setup —> Data collaboration (先講大綱，再講細節)
- Identity Resolution
- Healthcare
  - Guest: Liz Homans, CEO, Lyell Immunopharma
    - 40/100 people will develop cancer in their lifetime
    - 2000 of you could one day be a person with cancer
    - 250 of us are already cancer survivors
    - Cell therapy is the next frontier —> 大到太空，小到細胞
    - 90% of all cancers are solid tumor cancers
    - 接著進入 T 細胞小教室 :p
    - Testing T-cell therapies today
    - LyFE
    - All of possible workshop
  - AWS for Health
  - Challenges of omics
    - cost
    - scale
    - specialist tools
    - privacy
  - Omics 小教室
  - [NEW 🚀] Amazon Omics (GA)
- Retail
  - Just walk out technology by amazon
  - “Amazon one” 揮揮手結掌，喔不，是結帳？！
Collaboration
收尾
Transforming X
- Transforming innovation
- Transforming penalties of failure
- Transforming decision making
- Transforming culture