6件の比較グループ
ストレージObject storage
Store unstructured objects behind HTTP APIs for applications, analytics, media, and archives.
AWS
Amazon S3
- 主な用途
- Broad object-storage platform and data-lake foundation
- 運用モデル
- Bucket policy, IAM, lifecycle, replication, and a large integration surface
- 特性
- Strong read-after-write consistency with many storage classes
- 注意点
- Data transfer, request tiers, and optional features make cost modeling detailed
技術ガイドを開く Google Cloud
Cloud Storage
- 主な用途
- Object storage integrated with analytics and machine-learning services
- 運用モデル
- Uniform access, lifecycle, Autoclass, and regional or multi-region placement
- 特性
- Strong global consistency with automatic class management options
- 注意点
- Location type and data-processing region should match the surrounding pipeline
技術ガイドを開く Azure
Blob Storage
- 主な用途
- Object and data-lake storage for Azure applications and Microsoft data tooling
- 運用モデル
- Storage accounts, containers, access tiers, lifecycle, and private endpoints
- 特性
- Block, append, and page blob types with hierarchical namespace option
- 注意点
- Storage-account boundaries and feature compatibility require early planning
Cloudflare
R2
- 主な用途
- S3-compatible objects served through Workers and Cloudflare's network
- 運用モデル
- Buckets, bindings or S3 API, lifecycle, event notifications, and cache
- 特性
- Strong consistency and no internet egress fee
- 注意点
- The ecosystem and storage-class range are narrower than hyperscaler object stores
技術ガイドを開く AWS
AWS Lambda
- 主な用途
- Short event-driven functions across the AWS service ecosystem
- 運用モデル
- Functions, versions, concurrency, event sources, and IAM roles
- 特性
- Invocation-based scaling with bounded execution duration
- 注意点
- Cold starts, packaging limits, and per-service integration settings shape design
技術ガイドを開く Google Cloud
Cloud Run
- 主な用途
- Portable HTTP services and jobs packaged as containers
- 運用モデル
- Revisions, service identity, concurrency, min instances, and traffic splitting
- 特性
- Request-driven container scaling, including scale to zero
- 注意点
- Long-running state must live outside instances and startup cost depends on the image
技術ガイドを開く Azure
Azure Functions
- 主な用途
- Event-driven functions integrated with Azure triggers and bindings
- 運用モデル
- Function apps, hosting plans, deployment slots, identities, and bindings
- 特性
- Multiple hosting plans trade scale behavior against isolation and predictability
- 注意点
- Plan choice materially changes networking, cold starts, and execution limits
Cloudflare
Workers
- 主な用途
- Low-latency APIs and web applications across a global edge network
- 運用モデル
- Deployments, routes, bindings, observability, and compatibility flags
- 特性
- Isolate-based execution with rapid global startup
- 注意点
- CPU and runtime constraints favor web-native, bounded workloads
技術ガイドを開く AWS
Amazon EKS
- 主な用途
- Managed Kubernetes integrated with AWS networking and identity
- 運用モデル
- Clusters, node groups or Auto Mode, add-ons, upgrades, and IAM
- 特性
- Upstream Kubernetes control plane with flexible worker capacity
- 注意点
- Kubernetes and VPC complexity remain part of the operating model
技術ガイドを開く Google Cloud
GKE
- 主な用途
- Managed Kubernetes with Autopilot and fleet policy options
- 運用モデル
- Clusters, release channels, Autopilot or nodes, fleets, and workload identity
- 特性
- Deep Kubernetes automation and Google-originated operational patterns
- 注意点
- Standard mode still requires deliberate node, networking, and upgrade management
技術ガイドを開く Azure
AKS
- 主な用途
- Managed Kubernetes for Azure and Microsoft-centric environments
- 運用モデル
- Clusters, node pools, Entra identity, networking, upgrades, and policy
- 特性
- Kubernetes integrated with Azure networking and governance
- 注意点
- Networking mode and identity choices are hard to change later
Cloudflare
Containers
- 主な用途
- On-demand full-process workloads coordinated by Workers
- 運用モデル
- Container images, instances, lifecycle API, Worker control plane, and storage
- 特性
- Container capacity attached to Cloudflare's edge application platform
- 注意点
- It is not a managed Kubernetes service and has a newer operational ecosystem
技術ガイドを開く AWS
Amazon SQS
- 主な用途
- Work queues with standard or FIFO delivery
- 運用モデル
- Queues, visibility timeout, redrive policy, consumers, and IAM
- 特性
- Pull-based queue semantics with at-least-once standard delivery
- 注意点
- Fan-out usually adds SNS or EventBridge
技術ガイドを開く Google Cloud
Pub/Sub
- 主な用途
- Topic-based event distribution and streaming ingestion
- 運用モデル
- Topics, subscriptions, retention, filters, schemas, and identities
- 特性
- Each subscription tracks delivery independently for fan-out
- 注意点
- Queue and ordered-processing patterns need subscription and ordering-key design
技術ガイドを開く Azure
Service Bus
- 主な用途
- Enterprise queues and topics with ordering and transactions
- 運用モデル
- Namespaces, queues, topics, subscriptions, sessions, and rules
- 特性
- Broker features include sessions, dead lettering, duplicate detection, and transactions
- 注意点
- Throughput tiers and namespace capacity need explicit planning
Cloudflare
Queues
- 主な用途
- Managed delivery to consumer Workers near the rest of a Cloudflare application
- 運用モデル
- Queues, producers, consumers, batching, retries, and dead letters
- 特性
- Push or pull consumption with at-least-once delivery
- 注意点
- The broker feature set is intentionally smaller than enterprise messaging suites
技術ガイドを開く Access foundation models and build governed AI applications with evaluation, retrieval, and observability.
AWS
Amazon Bedrock
- 主な用途
- Managed access to multiple model providers inside AWS
- 運用モデル
- Model access, inference profiles, guardrails, knowledge bases, and agents
- 特性
- Serverless model APIs with AWS identity and network controls
- 注意点
- Model availability, region, and feature support vary by provider and model
Google Cloud
Vertex AI
- 主な用途
- Unified custom ML and generative AI development on Google Cloud
- 運用モデル
- Models, endpoints, pipelines, RAG, evaluation, monitoring, and IAM
- 特性
- Managed model lifecycle plus Gemini and partner model access
- 注意点
- The platform is broad, so teams need clear boundaries between training and application layers
技術ガイドを開く Azure
Azure AI Foundry
- 主な用途
- Model catalog and AI application tooling integrated with Azure
- 運用モデル
- Projects, model deployments, content safety, evaluations, and connections
- 特性
- Strong Microsoft and Azure OpenAI integration
- 注意点
- Naming, portal surfaces, quotas, and regional model availability change frequently
Cloudflare
Workers AI + AI Gateway
- 主な用途
- Run or route inference close to edge applications
- 運用モデル
- Model bindings, gateway routing, caching, rate limits, logs, and provider keys
- 特性
- Edge-native inference and a control plane for multiple external providers
- 注意点
- Training and deep model lifecycle tooling are outside the core platform
AWS
CloudFront
- 主な用途
- CDN tightly integrated with AWS origins and security services
- 運用モデル
- Distributions, behaviors, cache policies, origins, functions, and invalidations
- 特性
- Global cache with detailed per-path behavior configuration
- 注意点
- Policy combinations and data transfer pricing take time to model
Google Cloud
Cloud CDN
- 主な用途
- CDN attached to Google Cloud external load balancing
- 運用モデル
- Backend services, cache modes, signed URLs, invalidation, and Cloud Armor
- 特性
- Global anycast entry through the load-balancing stack
- 注意点
- The load balancer is a required architectural component rather than an optional front end
Azure
Azure Front Door
- 主な用途
- Global application delivery, acceleration, and web protection
- 運用モデル
- Profiles, endpoints, routes, origins, rules, WAF, and private link
- 特性
- Layer 7 global routing and CDN in one service
- 注意点
- Classic and Standard/Premium generations differ, so migration state matters
Cloudflare
Cloudflare CDN
- 主な用途
- Reverse-proxy CDN and security in front of internet origins
- 運用モデル
- Zones, DNS proxying, cache rules, WAF, Workers, and tiered cache
- 特性
- Anycast network with security and compute on the same request path
- 注意点
- Some enterprise routing and cache controls depend on plan level