Share
Embed
Transcript
48 slides
-
Kong Solution Day 2026 Winter Kong Konnect 徹底解説 統合API管理・メータリングから AIを活用した運用(
KAi)までを一気に理解 村田 一平 Kong株式会社 Professional Services Staff Field Engineer -
免責事項 • 本日紹介する機能はTech Previewの機能も含まれます • 将来的に仕様が変わる可能性があり、実際に使用する際に 機能や挙動が紹介した内容と異なる場合がございます • 一部の機能は標準で有効化されていないため、試されたい機能が 見当たらない場合は弊社の営業やカスタマーサクセスまで
お問い合わせください • 機能によっては追加のライセンスが必要となる場合がございます -
Index 01 Kong Konnectとは 02 KonnectのConnectivity管理 03 KonnectのApplication管理 04 CLIによるアプローチ
- kongctl 05 Kong AIアシスタント - KAi 06 KAiデモ 07 まとめ -
自己紹介 村田 一平 Kong株式会社 Professional Services / Field Engineer 日立製作所
(開発・サポート・ PM・企画・何でも) VMware (Professional Services) Kong (Professional Services) -
Kong Konnect On-prem/cloud Kong Gateway Control Plane Self-Managed Data Planes
〜2021年 Kong Gateway Enterprise -
Kong Konnect On-prem/cloud Kong Gateway Control Plane Self-Managed Data Planes
〜2021年 Kong Gateway Enterprise 出展元:https://atmarkit.itmedia.co.jp/ait/articles/2010/09/news059.html -
Kong Konnect On-prem/cloud Kong Gateway Control Plane Self-Managed Data Planes
〜2023年 Kong Gateway Enterprise On-prem/cloud (self-managed) Konnect Managed by Kong Kong Gateway Control Plane Data Planes -
Kong Konnect On-prem/cloud Kong Gateway Control Plane Self-Managed Data Planes
2024年〜 On-prem/cloud (self-managed) Konnect Managed by Kong Konnect Kong Gateway Control Plane Data Planes Kong Gateway Control Plane Dedicated Cloud Gateway Data Planes -
Konnectの特長 • マルチクラウド環境 での設定管理を一元化 • 従来環境で必要だった Control PlaneやDatabaseの用意・運用が不要 • Data
Planeはレイテンシ観点でお客様環境に置くことも、運用効率化観点で SaaS側に持たせることも可能に • KonnectのSLAは99.9%の稼働率を保証 (≠ Data Planeの稼働率) • Konnectのダウンタイムは APIトラフィックを中断させない -
Kong Enterpriseとの「API Gateway」としての違い • 利用できるモードはHybridモードのみ • データベースは不要 DB Control Plane
Data Plane ユーザ サービス API お客様が管理する領域 Kongが管理する領域 DB Control Plane Data Plane ユーザ サービス API お客様が管理する領域 Kongが管理する領域 • Control Planeのメンテナンスは不要 • Data Planeの耐障害性は変わらず (アーキテクチャ依存) Kong Enterprise Kong Konnect Konnectだと... -
Konnectが提供する機能 Connectivity管理 API Gateway Event Gateway AI Gateway Service Mesh
Application管理 APIs (APIカタログ ) / Dev Portal Service Catalog Scorecard Analytics Identity Metering -
Connectivity管理 API Gateway Event Gateway AI Gateway Service Mesh •
軽量・高速 • クラウドネイティブ • マルチクラウド • 自動化・ APIOps • Kafkaプロトコル専用 プロキシ • Virtual Clusterによる 論理分割 • Kafkaポリシーの中央 管理 • Konnect UIによる 統合管理 • マルチLLMを単一I/F で利用 • トラフィック制御とコス ト最適化 • AIガバナンスと セキュリティ • AIエージェント支援 (MCP対応) • マルチPFとマルチ ゾーンの統合管理 • ゼロトラストセキュリ ティ(mTLS)の自動化 • マルチテナント ガバナンス • シンプル設計・ シンプル運用 APIの中央管理 ・ガバナンス Kafkaの中央管理 ・ガバナンス LLMの中央管理 ・ガバナンス Service Meshの中央管 理 ・ガバナンス 一言で言うと -
Connectivity管理 API Gateway Event Gateway AI Gateway Service Mesh •
軽量・高速 • クラウドネイティブ • マルチクラウド • 自動化・ APIOps • Kafkaプロトコル専用 プロキシ • Virtual Clusterによる 論理分割 • Kafkaポリシーの中央 管理 • Konnect UIによる 統合管理 • マルチLLMを単一I/F で利用 • トラフィック制御とコス ト最適化 • AIガバナンスと セキュリティ • AIエージェント支援 (MCP対応) • マルチPFとマルチ ゾーンの統合管理 • ゼロトラストセキュリ ティ(mTLS)の自動化 • マルチテナント ガバナンス • シンプル設計・ シンプル運用 APIの中央管理 ・ガバナンス Kafkaの中央管理 ・ガバナンス LLMの中央管理 ・ガバナンス Service Meshの中央管 理 ・ガバナンス 一言で言うと -
AI Gateway • 実体はAPI Gateway • LLMの管理向けに高度にカスタマイズ を実施 • 以下のような機能を提供
◦ 単一のリクエストで複数の LLMに問い合わせ → LLMのロードバランサ を実現 ◦ プロンプトのテンプレート提供 や修飾子の強制 ◦ NGワードなどを制限 する抑止機能 ◦ RAGへのインジェションおよび利用 ◦ セマンティックキャッシュ によるコスト削減・高速化 -
AI Gatewayことはじめ "message": { "role": "assistant", "content": "Kong AI Gateway
is a powerful API management solution designed to handle, secure, and scale APIs efficiently. Although specific implementations and features may vary, the Kong API Gateway generally functions through a set of core components and features that facilitate its operations. Here's an overview of how it works:\n\n### Core Components:\n\n1. **API Gateway**: Kong acts as a reverse proxy that routes incoming API requests to the appropriate back-end services. It listens for API calls and manages traffic to backend services, ensuring efficient resource utilization.\n\n2. **Plugins**: Kong supports a wide range of plugins that add functionalities such as authentication, logging, traffic control, caching, and monitoring. These plugins can be configured per API or globally across all APIs.\n\n3. **Admin API**: Kong provides an Admin API that allows developers to manage routes, services, consumers, and plugins programmatically. This API is used for setting up and configuring the gateway.\n\n4. **Database**: Kong can operate in a database-backed mode (using PostgreSQL or Cassandra) or in a DB-less mode where configurations are stored in flat files. The database stores metadata about services, routes, consumers, and plugins.\n\n### How It Works:\n\n1. **Routing Requests**: When a request is made to an API, Kong inspects the incoming request against the configured routes. It uses the request method, path, and headers to determine where to forward the request.\n\n2. **Executing Plugins**: Before the request is sent to the backend service, Kong executes any associated plugins. This could include authentication checks (like OAuth or API keys), traffic limiting, request transformations, or logging.\n\n3. **Forwarding the Request**: Once all plugins have been processed, Kong forwards the request to the appropriate backend service.\n\n4. **Handling Responses**: After the backend service processes the request and returns a response, Kong can again apply plugins to manipulate the response before sending it back to the client. This might involve transforming the response format or applying headers.\n\n5. **Logging and Monitoring**: Kong can log API usage and performance metrics, allowing for detailed monitoring and analysis. This assists in maintaining the health and performance of the system.\n\n6. **Security Features**: Kong provides various security features such as rate limiting, IP whitelisting, and JWT authentication to protect APIs from unauthorized access and abuse.\n\n### Evolution and Integrations:\n\nKong has also evolved to incorporate features that support microservices architectures, including service mesh capabilities. With the advent of Kong as a Service and integrations with tools like Kubernetes, it allows for dynamic service discovery, load balancing, and enhanced orchestration.\n\nIn summary, Kong AI Gateway works by efficiently routing API requests, applying a variety of configurable plugins for middleware functionalities, securing APIs, and providing insight into API usage, contributing to the overall management and scalability of microservices and API ecosystems.", "refusal": null, "annotations": [] }, -
Application管理(1) APIs / Dev Portal Service Catalog Scorecard • APIの仕様・ドキュメントのハ
ブ・カタログ化 • 高度なカスタマイズ性 • Kong GWとの統合 • セルフサービスによる 開発者体験 • 他社サービスを含む一元化 されたサービスハブ • カスタムメタデータによる管 理画面のカスタマイズ • 各種エンティティの依存関係 の可視化 • ベストプラクティスに 基づいたサービスの評価 • 豊富な事前定義済み テンプレート • 柔軟なカスタム基準の作成 • ガバナンスと可視化の向上 開発者向け APIカタログ・ポータル 一言で言うと 各種サービスの ポータル DORAなどによる APIの評価と ベストプラクティス -
Application管理(2) Analytics Identity Metering • リアルタイムトラフィック分析 • AI/LLMの可視化 • ダッシュボードとレポート機
能 • リアルタイム使用量計測 • 柔軟な集計・課金モデル • PayPalなどの外部ツールと の連携 APIの分析 ・トラブルシューティング 一言で言うと APIのための 認証・認可基盤 API利用者向け従量課金 • M2M認証・認可の統合管理 • OIDC/OAuth Pluginを OOTBで利用可能 • Dev Portalとの統合 -
APIs(APIカタログ) / Dev Portal • 社内で利用できるAPIを全て可視化 • Open API Spec(OAS)やAPI説明ドキュメントを
併せて登録し、正確な仕様を全社で共有 • 開発者向けPortal画面は高カスタマイズ性 で 30以上のコンポーネントを提供 • 細かな権限設定 により、必要な情報を適切な人 間にのみ開示 APIを資産と捉えた場合、社内の APIが把握で きていないこと自体がセキュリティや ガバナンスのリスクであり、 敷いてはビジネス上のリスクである – Kong 村田, 2026 -
Scorecard • Service Catalog上のサービスの健全性を検証 • 適切かどうかを判断するスコアカードを用意 ◦ Kong Best Practices:
Kongの推奨プラクティスに沿っているか (エラー率、認証設定チェック) ◦ Security & Complicance: GitHubのPRのマージ状況や Dependabotの結果に問題がないか ◦ Service Documentation: OASやドキュメントが揃っていて、 OWASP観点で問題ないか ◦ Service Maturity: DORA指標などの業界指標観点で サービスが成熟できているか • 独自にスコアカードを作成可能 (例:PagerDutyと連携し、オンコールエンジニア が割り当てられているかなどを設定可) -
Metering • APIの利用状況を計測・課金 する仕組みを提供 • APIリクエストや AI GatewayのLLMトークン、任 意のイベントをメーターとして計測可能 (例:APIリクエスト数、LLMトークン数、GPU実行
時間、ユーザー数など) • 顧客毎に利用状況をひも付け て集計し、 サブスクリプションと 請求書を自動生成 • 豊富な課金設定 ◦ 定額、線形増加、一定量超過毎に ランクアップなど ◦ 月末〆、月途中などの支払日設定 ◦ 最低額・上限額設定、ディスカウント(利用 量、パーセント) ◦ 外部サービス連携 (Salesforce, PayPal, HubSpot, Stripe, etc) -
Konnectの自動化のアプローチ API開発者向け 提供機能 WebUI CLI decK K8s CRDs Terraform curl
(API) Kong Gatewayの 設定 ⭕ ⭕ ⭕ ⭕ ⭕ Kong Gatewayの 設定取得 ⭕ ⭕ ⭕ ❌ ⭕ APIs / Dev Portal の設定 ⭕ ❌ ❌ ⭕ ⭕ APIs / Dev Portal の設定取得 ⭕ ❌ ❌ ❌ ⭕ • APIでの実装はかなり複雑 • Terraformはstate管理を考えるとやや複雑になることも -
Konnectの自動化のアプローチ API開発者向け 提供機能 WebUI CLI decK K8s CRDs Terraform curl
(API) kongctl Kong Gatewayの 設定 ⭕ ⭕ ⭕ ⭕ ⭕ ❌ Kong Gatewayの 設定取得 ⭕ ⭕ ⭕ ❌ ⭕ ❌ APIs / Dev Portal の設定 ⭕ ❌ ❌ ⭕ ⭕ ⭕ APIs / Dev Portal の設定取得 ⭕ ❌ ❌ ❌ ⭕ ⭕ • APIでの実装はかなり複雑 • Terraformはstate管理を考えるとやや複雑になることも -
kongctlとは • Konnectの設定をYAMLで定義可能 ◦ CI/CDパイプラインに組み込んでAPI仕様書などを 自動でKonnectに登録可能 ◦ Dev Portalのデザインも CI/CDで管理
• 現状の設定をYAMLで取得可能 ◦ 設定のバックアップが簡単に ◦ バージョン管理も可能に • KAiとの対話も可能 ご質問はこちら -
KAi - Kong AI Assistant • Konnect UI上 / kongctl
から問い合わせ可能なAIアシスタント • 以下のようなことが対話的に相談可能(日本語OK) ◦ パフォーマンス問題の デバッグ支援 ◦ Konnectに関する質問全般 ◦ 利用状況の確認 (Service数、Consumer数等) -
まとめ Konnectとは... • Unified API & AI Platform ◦ ただのAPI
Gatewayではなく、Connectivity(Gateway/Mesh/AI)と Application(APIカタログ、サービスカタログ、Dev Portal)を一元管理 ◦ マルチクラウド / ハイブリッドクラウド環境上のAPI/AIを一元管理 • 運用負荷を低減 ◦ Control PlaneとData Planeを分離し、SLA99.9%かつ SaaS側の障害の影響を遮断 ◦ Control Plane / Data Plane / データベースのメンテナンス不要 ◦ 運用自動化・高度化のための各種手法(API, CLI, Terrafrom)を提供 • AI時代のガバナンス ◦ AI Gateway によるマルチLLMの統制とコスト最適化を実現 ◦ Konnectの機能自体にもAIを活用(KAi)