Cloud Services Cross-Reference: NoSQL & Specialty Databases

All four major cloud providers — AWS, Azure, OCI, and Google Cloud — offer purpose-built NoSQL and specialty database services spanning document stores, key-value caches, graph databases, time-series engines, wide-column stores, ledger/blockchain platforms, and search services. The landscape has shifted significantly in 2024-2025: AWS retired QLDB (July 2025) and put Timestream for LiveAnalytics into maintenance mode, Azure introduced Azure Managed Redis as a replacement for Azure Cache for Redis tiers, and Google Cloud reached general availability for Memorystore for Valkey while deprecating Memcached. OCI expanded its Oracle Graph SQL property graph support and Search with OpenSearch capabilities. This document cross-references services by category using the official product names current as of February 2026.

Document / JSON Databases

Document databases store semi-structured JSON data and support flexible schemas. This is one of the most competitive categories, with each provider offering distinct approaches.

Notes:

• Amazon DynamoDB is the AWS flagship NoSQL offering. It supports both key-value and document (JSON) models in a single serverless service with on-demand and provisioned capacity modes.
• Amazon DocumentDB is a fully managed MongoDB API-compatible service. It is architecturally distinct from MongoDB — it uses a purpose-built storage layer, not the MongoDB engine.
• Azure Cosmos DB for NoSQL is Microsoft's native document store and the recommended API for new applications. It uses a proprietary JSON model with automatic indexing, multi-region writes, and a 99.999% SLA.
• Azure Cosmos DB for MongoDB provides wire-protocol compatibility with MongoDB. Azure DocumentDB (vCore) is a newer offering powered by the open-source DocumentDB engine on PostgreSQL, targeting workloads requiring deep aggregation pipelines or MongoDB portability without refactoring.
• OCI NoSQL Database Cloud Service supports JSON, table, and key-value data models under a single service. It includes full SQL query access over JSON documents and ACID transaction support.
• Oracle Autonomous AI JSON Database is a variant of Oracle Autonomous Database specialized for JSON document workloads, with MongoDB API compatibility via Oracle SODA. It is distinct from the OCI NoSQL service: it runs on the full Oracle database engine and stores relational data alongside JSON.
• Cloud Firestore is Google's primary managed document database. It is the successor to Cloud Datastore (now deprecated), offering real-time sync, strong consistency, and serverless scaling.

Key-Value Stores

Pure key-value stores treat data as opaque values accessed by a primary key. DynamoDB also covers this model.

Notes:

• Amazon DynamoDB is the primary key-value and document store on AWS. All reads are by primary key (partition key + optional sort key), with optional secondary indexes.
• Azure Cosmos DB for Table provides an Azure Table Storage-compatible API with global distribution and single-digit millisecond latency. It is a drop-in replacement for legacy Azure Table Storage workloads that need higher performance and global reach.
• OCI NoSQL Database Cloud Service supports key-value as one of its three native data models (alongside JSON and table).
• Google Cloud does not offer a dedicated standalone key-value store. Firestore and Bigtable both serve key-based lookup patterns at different scales.

In-Memory / Cache Services

In-memory databases and caching layers deliver sub-millisecond latency by storing data in RAM.

Notes:

• Amazon ElastiCache is the primary caching service on AWS, supporting Valkey, Redis OSS, and Memcached engines. It is optimized for caching use cases (data is not guaranteed durable between snapshots). Serverless tier available.
• Amazon MemoryDB is a distinct service: a fully durable, Redis-compatible in-memory database. Every write is persisted to a Multi-AZ transaction log before acknowledgment. Intended for use as a primary database, not just a cache. Supports Valkey and Redis OSS.
• Azure Managed Redis (GA as of 2025) is the strategic successor to Azure Cache for Redis, built on the Redis Enterprise stack and available as a first-party Azure service without a Marketplace component. Basic, Standard, and Premium tiers of Azure Cache for Redis retire September 30, 2028; Enterprise tier retires March 30, 2027.
• OCI Cache is a fully managed Valkey and Redis 7.0-compatible service with support for both non-sharded (1–5 nodes) and sharded cluster topologies. Achieves sub-millisecond latency.
• Memorystore for Valkey (GCP) reached GA status with a 99.99% SLA. It delivers up to 2x QPS compared to Memorystore for Redis Cluster. Fully compatible with Redis OSS 7.2 APIs. Memorystore for Memcached is deprecated and will shut down January 31, 2029.

Graph Databases

Graph databases model entities as nodes and relationships as edges, enabling efficient traversal of highly connected data.

Notes:

• Amazon Neptune is AWS's primary managed graph database. It supports three query models: Property Graph via Apache TinkerPop Gremlin and openCypher, and RDF via SPARQL. Neptune Analytics is a companion in-memory graph analytics engine that can process tens of billions of relationships and is integrated with Amazon Bedrock for GraphRAG workloads.
• Azure Cosmos DB for Apache Gremlin is Microsoft's managed graph offering, delivered as one of Cosmos DB's multi-model APIs. It uses the TinkerPop Gremlin query language. It is not a standalone service — it shares Cosmos DB's infrastructure and pricing model.
• Oracle Graph is integrated directly into Oracle Database 23ai and Oracle Autonomous Database on OCI. As of 23ai, SQL property graphs are a native database object queryable via standard SQL and the ISO/GQL graph query language. No separate service purchase is required; graph is included with the database. Supports import from Neo4j via a PL/SQL package.
• Spanner Graph is Google Cloud's graph capability, embedded natively within Cloud Spanner. It allows querying graph relationships alongside relational data using SQL and GQL. Spanner's multi-model architecture (relational + graph + key-value + vector + search) makes it unique in the industry.
• Neither OCI nor Google Cloud offer a dedicated standalone managed graph database service independent of their relational databases. AWS Neptune and Azure Cosmos DB Gremlin are the only first-party dedicated graph services.

Time-Series Databases

Time-series databases optimize for high-volume ingestion and range-based queries over timestamped data, typical of IoT, metrics, and telemetry workloads.

Notes:

• Amazon Timestream for InfluxDB (launched March 2024) is a fully managed wrapper around InfluxDB OSS 2.x, targeting near-real-time workloads with single-digit millisecond query response. Suitable for IoT and metrics with cardinality under 10 million unique dimension combinations. Priced by compute instance-hour.
• Amazon Timestream for LiveAnalytics is closed to new customers as of June 20, 2025; it remains operational for existing customers with no announced shutdown date. Existing customers should plan migration to Timestream for InfluxDB. It was a serverless, SQL-based service capable of ingesting gigabytes per minute and querying petabytes.
• Azure Data Explorer (ADX) is Microsoft's recommended time-series analytics platform following the retirement of Azure Time Series Insights. ADX uses the Kusto Query Language (KQL) and supports native time-series functions including anomaly detection, downsampling, and trend analysis. It ingests streaming data at high throughput and scales horizontally.
• Cloud Bigtable is Google's wide-column store (see Wide-Column section) and is the recommended platform for time-series workloads on GCP, specifically via row-key schema patterns that encode timestamps. It does not have a dedicated time-series API but has been used at 40 million transactions per second for such workloads. BigQuery additionally supports time-series SQL functions for analytical queries.
• OCI does not offer a dedicated managed time-series service. Time-series data can be stored in OCI NoSQL Database or Oracle Autonomous Database with time-series patterns.

Wide-Column Stores

Wide-column (or column-family) stores use rows with dynamic columns, optimized for sparse, high-throughput workloads at scale. Apache Cassandra is the dominant open-source engine in this category.

Notes:

• Amazon Keyspaces is a serverless, fully managed Cassandra-compatible wide-column database. It is API-compatible with CQL (Cassandra Query Language) and provides up to 99.999% availability with multi-region replication. Supports on-demand and provisioned capacity modes.
• Azure Cosmos DB for Apache Cassandra provides Cassandra CQL wire protocol compatibility over the Cosmos DB backend. Like other Cosmos DB APIs, it benefits from turnkey global distribution, automatic indexing, and Cosmos DB's SLA guarantees. It is not a standalone service.
• Cloud Bigtable is Google's proprietary wide-column NoSQL database service for petabyte-scale workloads. It does not expose a Cassandra-compatible API but is the foundational store behind several Google products (Search, Maps, Gmail). Bigtable uses a row-key + column-family model and supports HBase API.
• OCI does not offer a managed Cassandra-compatible or wide-column store service. Cassandra workloads on OCI run on self-managed VM deployments or via OCI Marketplace images.

Ledger / Blockchain Databases

Ledger databases maintain an append-only, cryptographically verifiable history of changes. Blockchain services extend this to permissioned multi-party networks.

Notes:

• Amazon QLDB was AWS's purpose-built ledger database with a cryptographically hash-chained journal. It reached end of support on July 31, 2025 and is no longer available. AWS recommends Aurora PostgreSQL with audit logging for former QLDB workloads, though without cryptographic verifiability.
• Amazon Managed Blockchain remains active. It supports Ethereum, Polygon, Bitcoin, and Hyperledger Fabric. The ordering service for Hyperledger Fabric is backed by QLDB technology internally.
• Azure SQL Ledger is a feature embedded within Azure SQL Database and Azure SQL Managed Instance that provides blockchain-backed tamper-evidence for relational tables. It is not a standalone service.
• Azure Confidential Ledger is a standalone managed service providing a decentralized, tamper-proof ledger for data entries backed by blockchain. It is distinct from the Azure SQL Ledger feature.
• Oracle Blockchain Platform Cloud Service on OCI is based on Hyperledger Fabric and Hyperledger Besu (Ethereum-compatible). It supports NFTs, tokenization, and smart contract development. In October 2025, Oracle announced the Digital Assets Data Nexus platform, combining Oracle Blockchain with Oracle AI Database 26ai for financial institution use cases.
• Google Cloud does not offer a first-party managed blockchain or ledger database service. Hyperledger Fabric and Ethereum deployments on GCP use third-party Marketplace solutions.

Search Services

Search services provide full-text search, vector search, and analytics over unstructured and semi-structured content.

Notes:

• Amazon OpenSearch Service is AWS's fully managed distribution of OpenSearch (and previously Elasticsearch). It supports full-text search, log analytics, distributed tracing, and k-NN vector search for semantic and hybrid retrieval. Serverless option available.
• Azure AI Search (renamed from Azure Cognitive Search) is Microsoft's cloud search platform. In 2025 it gained an agentic retrieval engine for parallel, iterative, LLM-assisted search alongside its classic single-request engine. It provides hybrid search (full-text + vector), AI enrichment pipelines (OCR, NER, key phrase extraction), and deep integration with Azure OpenAI via its Foundry IQ feature family.
• OCI Search with OpenSearch is Oracle's managed OpenSearch 2.x service. It supports semantic search and hybrid search via the KNN plugin with Lucene, Faiss, and nmslib backends. Ingestion pipelines can automatically generate document embeddings at ingest time. The service is used for application search, log analysis, and security analytics in OCI reference architectures.
• Google Cloud does not offer a first-party general-purpose managed search service equivalent to OpenSearch or Azure AI Search. Full-text and vector search is handled at the database layer (Spanner, AlloyDB, BigQuery) or via Vertex AI Vector Search for embedding-based retrieval. Application search workloads typically use Elasticsearch or OpenSearch deployed on GKE, or Vertex AI Search (formerly Enterprise Search, part of the Vertex AI platform) for enterprise document search.

Summary: Service Coverage by Category

The following table provides a quick at-a-glance view of which providers offer first-party managed services in each specialty.

Unique to single provider:

• Azure Cosmos DB is the only service that exposes a single backend through five compatible APIs (NoSQL, MongoDB, Cassandra, Gremlin, Table), making it the most multi-model managed database available.
• Amazon DynamoDB is the only purpose-built, globally replicated, serverless key-value/document database that is not an API wrapper over another engine.
• Cloud Bigtable is the only first-party wide-column store that is not Cassandra-compatible, using its own HBase-compatible API at petabyte scale.
• Amazon MemoryDB is the only fully durable (Multi-AZ transaction log) in-memory database as a managed cloud service, positioned as a Redis-compatible primary database rather than a cache.
• Oracle Graph is the only graph capability delivered natively inside a relational SQL engine (Oracle Database 23ai), requiring no separate service and queryable via standard SQL.

References

• AWS: Choosing an AWS NoSQL Database
• AWS: Amazon DynamoDB
• AWS: Amazon DocumentDB
• AWS: Amazon ElastiCache
• AWS: Amazon MemoryDB FAQs
• AWS: Amazon Neptune
• AWS: Neptune Analytics vs Neptune Database
• AWS: Amazon Timestream for InfluxDB
• AWS: Amazon Keyspaces
• AWS: Amazon Managed Blockchain
• AWS: Amazon OpenSearch Service
• Azure: Cosmos DB Overview
• Azure: Choose an API in Azure Cosmos DB
• Azure: Cosmos DB for Apache Gremlin
• Azure: Azure Managed Redis - What's New
• Azure: Azure Cache for Redis Retirement
• Azure: Azure Data Explorer Overview
• Azure: Azure AI Search Introduction
• OCI: NoSQL Database Cloud Service Key Features
• OCI: OCI Cache Overview
• OCI: OCI Search with OpenSearch
• OCI: Oracle Graph Technology
• OCI: Oracle Blockchain Platform Cloud Service
• OCI: Oracle Autonomous AI JSON Database
• GCP: Google Cloud Database Products
• GCP: Memorystore for Valkey GA
• GCP: Memorystore for Memcached Deprecation
• GCP: Cloud Bigtable Time-Series Schema Design
• GCP: Spanner Graph
• Neo4j: Google Cloud Partnership