Graph Database
Azure Cosmos DB (Gremlin)
Azure Cosmos DB Gremlin API exposes Azure Cosmos DB's multi-model, globally distributed database as a property graph endpoint accessed with Apache TinkerPop Gremlin. It offers turnkey multi-region, multi-master replication, automatic indexing of all properties, five tunable consistency levels, and a serverless or provisioned-throughput RU billing model that scales elastically across logical partitions.
Azure
Service information
Shortname: Cosmos DB Gremlin
Huawei equivalent shortnames: GES
Keywords: graph, database, gremlin, relationships
Differences vs Huawei
Service boundaries and operating model differ sharply. Cosmos DB is a multi-model database whose Gremlin API is one of several interchangeable APIs over the same underlying store, with built-in global distribution, multi-master writes, and an SLA-backed latency floor. Huawei GES is a purpose-built graph engine focused on relationship querying and analytics it is not a multi-model store, and it exposes graph management, metadata, backup, and algorithm APIs rather than Cosmos-style multi-API multi-region replication. Treat them as analogous only for graph workloads, not as drop-in replacements for Cosmos DB's broader capabilities.
API and data model compatibility have version and semantic gaps. GES is compatible with Apache TinkerPop Gremlin 3.3.0, while Cosmos DB Gremlin targets the TinkerPop 3.4.x stream with extensions; Gremlin traversal steps, schema enforcement, and property/edge cardinality behavior differ. Cosmos DB uses partition keys with unlimited logical partitions and automatic indexing of every path; GES requires an explicit metadata (schema) file, predefined vertex and edge labels, and import via OBS or local files. Re-validate each traversal step, schema construct, and id/partition-key design rather than assuming parity.
Scaling, high availability, and operational responsibility diverge. Cosmos DB delivers multi-region active-active writes, configurable consistency across bounded staleness to strong, and serverless or autoscale RU provisioning managed by Microsoft. GES runs as a managed graph instance with a distributed processing engine for high-concurrency multi-hop queries, but its HA, replication, and regional topology follow the GES instance model and region set, not Cosmos-style turnkey multi-master. Backup/restore, observability, and capacity scaling are operator responsibilities on GES via its console and APIs, with cross-region behavior differing materially from Cosmos DB's automatic global replication.
Migration to Huawei
Assess the workload before choosing a target. Inventory Gremlin traversal variants, partition-key strategy, consistency requirements, region/HA topology, throughput (RU/s), index usage, and TTL. GES is the closest Huawei equivalent for property-graph query and analytics, but if the Cosmos DB use relies heavily on multi-master multi-region writes, sub-10-ms latency SLAs, or serverless autoscale, plan compensating architecture rather than assuming feature parity. Confirm GES Gremlin 3.3.0 compatibility against your traversal set early.
There is no documented one-click or DRS pipeline from Azure Cosmos DB Gremlin to GES. Export vertices and edges from Cosmos DB to structured files in OBS, define a GES metadata (schema) file mapping each vertex/edge label and property, and import using GES incremental import or Huawei Cloud CDM for orchestration. Recreate logic that relies on Cosmos-specific constructs (partition keys, automatic indexing, TTL) manually inside the GES schema and import job. Validate schema, constraints, and id mapping before any data cutover.
For application migration, retarget client Gremlin connections to the GES Gremlin endpoint and re-test every traversal against the 3.3.0-compatible engine, accounting for syntactic and semantic differences in steps, cardinality, and error handling. Replace Cosmos SDK-specific calls (e.g., change-feed, bulk executor, conflict resolution) with GES APIs or downstream consumers since GES does not provide direct equivalents. Update retry, timeout, backpressure, and connection-pool settings to GES's behavior, which differs from Cosmos's gateway or direct-mode clients.
Post-migration, recompute TCO and validate operational gaps. GES is billed per graph instance and storage rather than per-request RU; combined with OBS, CDM/DataOps, and observability services this changes cost shape versus Cosmos's serverless or autoscale RU plus storage and cross-region transfer. Validate backup/restore, monitoring via Cloud Eye and LTS, IAM/scope permissions at label and property level, and any data-governance needs with DGC before production cutover. Keep a dual-write or reconciliation phase because there is no native CDC bridge between Cosmos DB and GES.
Official Huawei Cloud documentation
Huawei Cloud
Huawei equivalent service
Shortname: GES
General function: Graph Database
Graph database and analytics engine.
Keywords: graph, database, relationship