演講: 閱讀與觀察 2019 CNCF 調查結果

(圖說:分類用的儲存盒。圖片來源。)

今天很開心能參與 SDN x Cloud Native Meetup #27(由 Software Defined Network Develop Society TWCloud Native Taiwan User Group 兩個社群合辦)線上分享 2019 CNCF Survey 的內容。

投影片

摘要

這裡僅是簡單摘要,完整內容請參閱報告原文、或 SDN x Cloud Native Meetup #27 活動錄影 內容討論。

Key Takeaways

  • Survey Methodology: Scope v.s Industry
  • (Release cycles)+++ v.s (Check in code)-
  • (Use of containers in prod)+++
  • (Cultural changes)++
  • Serverless
    • Using/planning to use: 70%
    • Hosted : Installable = 80% : 20%
  • Kubernetes
    • Helm: ~70% (Packaging)
    • Not use tracing: 23%

時間軸

這份調查報告是 CNCF (Cloud Native Computing Foundation)

  • 於 2019 九月至十月之間,進行調查與回收問卷。
    • 總共回收 1,337 份問卷。
  • 2020 年三月四日,發布調查結果報告

調查取樣

  • The majority of respondents (71%) were from organizations with at least 100 employees, the largest portion of these coming from enterprises with more than 5,000 employees (30%).
    • 調查取樣偏向中大型企業。
  • Two-thirds of the respondents were in the software and technology industry, with the remainder coming from other professional service industries.
    • 前六大受訪產業:軟體、科技、金融服務、顧問諮詢、電信通訊、容器雲端相關。超過 60% 受訪者。
  • 調查結果與亞洲、傳統產業、地端現況有差距。亞洲區受訪者佔這個調查的 17%。

大分類

  • Application Development & Delivery
  • Containers
  • Cloud Native Tools
  • CNCF Technologies
  • Kubernetes

Application Development & Delivery

  • Release cycles continue to accelerate.
  • Those with daily release cycles increased from 15% in 2018 to 27%, and weekly release cycles have increased 20% to 28%.
  • Comparing the data from 2019 and 2018, respondents are checking in code less frequently.
  • Vendor lock-in strategy
    • 可能因企業規模、時程、應用場景而有所不同?
    • Enterprise
    • SME
    • Start-up

Containers

  • Most notably, the use of containers in production increased significantly.
  • This year, 84% of respondents are using containers in production, an impressive jump from 73% in 2018, and from 23% in our first survey in 2016.
    • Hint: Ref to SMR (4/5) Industry.
  • According to CNCF’s Cloud Native Landscape, there are more than 109 tools to manage containers, but 89% are using different forms of Kubernetes.
    • Hint: Ref to SMR (4/5) Industry.
  • Cultural challenges with the development team remain the top challenge in using/deploying containers (43%).
    • 導入與實作期間最大挑戰:文化改變。

Cloud Native Tools

  • While use in production is still rather low with 18% of those who responded indicating they use a service mesh project, 47% are evaluating the use of a service mesh.
  • Service Mesh:
    • Istio (69%) and Linkerd (64%) were the top two tools being evaluated, suggesting they will see significant increases in use in production in coming years.
  • Storage:
    • Top-3 are public cloud storage projects.
  • Serverless:
    • 60%~70% 受訪者,正在使用 (40%~50%) 或是打算使用 (20%) Serverless。
    • Of those who are using serverless, 80% use a hosted platform and 20% use installable software.
    • Of the hosted platforms in use, the top tool is AWS Lambda (53%). Google Cloud Functions (18%) and Azure Functions (14%) are a distant second and third.

(圖說:Serverless 導入情況。)

(圖說:已導入 Serverless 並選用代管 Serverless 平台的分佈。)

CNCF Technologies

  • 99% of respondents indicated they are using or evaluating at least one graduated or incubating CNCF technology in production.
  • Prometheus and CoreDNS also saw considerable jumps in use in production.
  • For incubating projects, Helm saw the largest increase since last year, growing 41% to reach 69% usage in production. etcd was the most widely used, with 83% use in production.

Kubernetes

  • Of those using Kubernetes, most respondents have 2-5 clusters in production (43%).
  • Helm remains the most popular tool for packaging Kubernetes applications.
  • the majority (70%) of respondents plan to autoscale their stateless applications, followed by 40% for task/queue processing applications, and 34% of stateful applications.
  • Of those who are not using Kubernetes autoscaling capabilities, 35% do not want to autoscale any works at this time, and 12% are using a third-party solution.
  • Namespaces are the most popular way to separate Kubernetes applications for all respondents, including those with multiple teams.
  • 23% report that they do not use tracing, compared with just over 3% for both monitoring and logging.
  • Minikube (39%) and Docker Kubernetes (32%) are the most popular Kubernetes environments during local container development.

後記

這次首次嘗試使用 Marp for VS Code 製作簡報投影片,這次投影片的 Markdown 原始檔案可以參考。

主要使用 Markdown 語法,加上一些 Marp 預設自定義的參數語法,限制頗多,特別是一開始我想嘗試只使用預設值狀態下,但相對帶來的優點也是有(對我來說),例如:

  • Markdown 文字輸入後,圖片就對齊完畢,不用動滑鼠調整位置。
  • 在投影片 Markdown 與文章 Markdown 之間複製貼上,簡單、粗暴、快速。
  • VS Code 可以快速預覽圖片,方便選擇。(不用動滑鼠)

之後再另篇記錄。

另外也歡迎大家到 AWS User Group Taiwan 逛逛聊聊 :)

備忘

  • Meetup #27: ~50 Attendees.
  • Youtube video: 87 views. (updated by 2020-0421)
comments powered by Disqus