Single service comparison

AWS Dedicated Hosts are scoped to an instance family (e.g., m5, c5) and allow you to launch any supported instance size on that host, with host affinity and instance affinity rules controlling where instances restart. Huawei DeH is typed (s6/c6/m6 general, compute-plus, memory) and the ECS flavors you can deploy are constrained by the DeH type, so placement flexibility is narrower. Both expose an Auto Placement concept, but AWS offers per-instance host targeting via LaunchDomain-like controls and AWS License Manager integration for socket/core tracking, while DeH relies on host-type to flavor mapping and console allocation without an equivalent license-tracking service.

Scaling and HA differ materially. On AWS, Dedicated Hosts participate in Auto Scaling groups, capacity reservations, and can be combined with Dedicated Host Savings Plans across an instance family; recovery relies on ASG re-provisioning and affinity re-binding. On Huawei, DeH is a fixed physical resource in one AZ with no native Auto Scaling at the host level—ECS instances on a DeH are managed individually, and AZ-level resilience requires a multi-AZ workload design. For a physical DeH fault, Huawei documents automatic ECS migration to an unprovisioned backup DeH. DeH does not offer a host-aggregate or host-resource-group API comparable to AWS host management, so multi-host fleets need orchestration via ECS/AS plus custom placement logic.

Integrations and operational responsibility differ. AWS Dedicated Hosts integrate with AWS Config, License Manager, CloudWatch, Systems Manager, and IAM for compliance reporting and license reporting; the customer is responsible for instance OS patching, license compliance attestation, and host release timing. Huawei DeH integrates with ECS, EVS, IAM, and Cloud Eye for monitoring, but has no bundled license-entitlement reporting tool, and O&M responsibility for host faults is shared with Huawei (replacement of faulty DeHs is a Huawei-side operation). DeH is billed per physical host; ECS compute resources created on that DeH do not add a separate compute charge, although associated images, EIPs, EVS disks, and backup resources are billed separately.

Start with assessment: inventory AWS Dedicated Hosts by instance family, socket/core counts, host affinity rules, and any License Manager license configurations. Map each to a Huawei DeH type (s6, c6, m6) using the DeH type-to-ECS-flavor matrix in the Huawei documentation, confirming the target flavor supports your required vCPU and memory and that license-bound software (e.g., Windows, SQL Server per-core) is still licensed correctly on Huawei's physical topology. Flag instances that fall outside the DeH flavor set as candidates for plain ECS or BMS instead of DeH, since DeH flavors are more restrictive.

For data and instance migration, there is no one-click AWS-to-Huawei host migration path. Build the target by allocating DeHs of the matched type in the target AZ, then use image-level replication: create a Huawei ECS image from your on-premises or AWS-exported VM (via SMS or IMS image import, after validating image format and guest drivers), and launch ECS instances onto the allocated DeH with Auto Placement off so instances land on the intended host. For application data, copy EBS snapshots to OBS via the cloud or an intermediate transfer appliance, then restore to EVS disks attached to the new ECS instances; re-establish placement, affinity, and IAM permissions manually.

Validate before cutover: run smoke tests against each re-hosted instance, verify socket/core reporting for license compliance through manual attestation (Huawei has no License Manager equivalent), confirm performance against the s6/c6/m6 specifications, and test both Huawei's documented physical-host recovery and the workload's separately designed AZ-level failover. Plan cutover as a DNS/EIP re-point with a rollback window; use paired capacity across AZs where the workload needs protection from an AZ outage, and remember that automatically created AS instances are not dispatched to DeHs.

Anticipate cost-model and gap changes. AWS Dedicated Hosts are billed per host with instances on the host not charged separately (under Dedicated Host billing), plus Savings Plan discounts; Huawei bills the DeH subscription while ECS compute on it is free; recalculate TCO for the host commitment plus separately billed images, EVS, EIPs, bandwidth, and backup resources. Document missing parity points up front: no License Manager, no host-level Auto Scaling, narrower flavor-per-host constraints, and AWS region/instance-family feature breadth, so stakeholders accept a composed ECS+DeH+BMS pattern rather than expecting like-for-like replacement.