Single service comparison

OBS is Huawei Cloud's general-purpose object store and offers an S3-compatible REST API, but the surface is not byte-identical to Amazon S3. Bucket policies, IAM principals, presigned-URL semantics, conditional writes, object-lock (WORM) modes, batch operations, and inventory reports differ in syntax or availability, and OBS uses Huawei AK/SK and IAM rather than AWS ARNs. Storage classes are comparable in intent (Standard, Infrequent Access, Archive) but minimum-duration, retrieval-tier, and restore mechanics diverge; for example, OBS Archive restore offers expedited and standard windows rather than S3 Glacier's separate expedited/standard/bulk tiers and Deep Archive class.

Use Dedicated OBS (DOS) when you need an isolated, single-tenant object-storage deployment with dedicated resources, custom compliance, or data-sovereignty constraints that the shared multi-tenant OBS cannot satisfy; it is NOT a drop-in performance tier or a second copy of S3, and most workloads should land on standard OBS. DOS has narrower regional availability, distinct procurement and quotas, and a different operational model, so treat it as a workload-variant choice driven by compliance or tenancy, not by storage class or price.

Integrations and operational responsibility differ materially. S3 couples natively with AWS KMS, CloudFront, Lambda, EventBridge, Macie, and Storage Lens; on Huawei, the equivalent chain is OBS with KMS (or DEW), CDN, FunctionGraph, CTS/CTS audit, and CBR for backup-style workloads, and the operator must rewire event notifications, replication, logging, and observability to Huawei services. Cross-region replication, versioning, default encryption, and lifecycle transitions exist on both sides but with different limits, region coverage, and configuration surfaces that architects must revalidate rather than assume parity.

Start with an assessment that maps each S3 bucket to a target: standard OBS for the vast majority of hot, warm, and archived datasets, and Dedicated OBS only where a compliance, tenancy, or sovereignty requirement mandates a dedicated deployment. Inventory every storage class, lifecycle rule, versioning/WORM policy, encryption setting, request volume, egress pattern, and dependent AWS service (Lambda, CloudFront, Macie, EventBridge) so you can record gaps and re-platform integrations on Huawei equivalents before cutover.

Move data with Huawei's Object Storage Migration Service (OMS) for bucket-to-OBS migration from AWS S3 and other sources, validating that OMS supports your source's authentication, listing scale, and incremental resync for the buckets in scope. For very large or time-constrained datasets, combine OMS with Data Express Service (DES) disk/appliance shipments, and re-point application code from the AWS SDK/S3 endpoint to the OBS endpoint, adjusting path-style vs virtual-hosted-style, region, signing, and presigned-URL logic.

Reconfigure IAM and policy as a re-platform, not a lift: convert S3 bucket policies and IAM role trusts to Huawei IAM policies and OBS ACLs/ACP, remap KMS/SSE-KMS keys to Huawei KMS, re-create lifecycle and replication rules in OBS syntax, and rebuild event-driven flows (notifications to FunctionGraph, CDN origin to OBS, audit through CTS) since these do not migrate automatically. Validate object counts, checksums, ACLs, encryption state, and CORS on a representative sample and re-run integrity checks after the final sync.

Plan for cost- and quota-model changes and a controlled cutover. AWS bills S3 by storage class GB-month, requests, retrieval tiers, and egress, while Huawei OBS bills capacity, requests, retrieval/replication, and outbound transfer with DOS on different procurement, so recompute TCO against peak GET/PUT rates, retention period, and cross-region or interconnect traffic. Confirm OBS bucket quotas, region availability, and DOS procurement lead time, then cut over DNS/endpoint references only after a validated final incremental sync and a documented rollback to the S3 source.