Choose the right compute SKU and configure redundant resources to balance cost and fault tolerance on CloudBox.
How CloudBox protects hardware workloads
CloudBox hardware protects your workloads at three layers:
Network devices: Dual-redundancy with dual upstream connections eliminates single points of failure and keeps the network highly available.
Storage devices: Data remains secure and reliable even if two storage servers fail simultaneously.
Computing devices: Failover and hot migration keep workloads running when redundant resources are available.
Configure computing resources
CloudBox delivers computing capacity through compute SKUs. Some SKUs include multiple Elastic Compute Service (ECS) instance families. When selecting a compute SKU, balance upfront cost against the redundant capacity needed for fault tolerance.
Select a compute SKU to match your workload
-
Match the vCPU-to-memory ratio of your chosen SKU to the ratio required by your workload.
For example, if you need 100 ECS instances each with 4 vCPUs and 8 GiB of memory, your total requirement is 400 vCPUs and 800 GiB—a 1:2 ratio. Choose a compute SKU whose instance family tops out at ecs.c6.26xlarge, which has the same 1:2 ratio. Choosing a SKU based on ecs.g6.26xlarge (1:4 ratio) wastes memory without providing additional compute benefit.
-
When your workload mix requires different vCPU-to-memory ratios, choose a multi-family SKU rather than over-provisioning a single family.
For example, if you need 100 instances at 4 vCPUs/8 GiB and 100 instances at 4 vCPUs/16 GiB, your total requirement is 800 vCPUs and 2,400 GiB. A SKU based solely on ecs.g6.26xlarge (104 vCPUs, 384 GiB each) wastes both memory and vCPUs. Instead, select a SKU that covers both ratios—one instance family topping out at ecs.c6.26xlarge (1:2) and another at ecs.g6.26xlarge (1:4).
For the full list of compute SKUs supported by CloudBox, see the "Compute SKU specifications" section of the Computing resources topic. If no available SKU meets your requirements, contact Alibaba Cloud technical support to customize one.
Reserve redundant resources for fault tolerance
CloudBox failover and hot migration can absorb hardware faults, but only when enough redundant resources are available at the time of failure. Two methods are available:
-
Add redundant resources when selecting your compute SKU.
These resources are visible and immediately usable—you can create ECS instances or other cloud service instances from them at any time. This gives you flexibility during normal operations, but does not guarantee that the right capacity is available exactly when a hardware failure occurs.
-
Purchase a Service Level Agreement (SLA) value-added service package.
The SLA package automatically reserves redundant resources based on your current configuration. These resources are invisible during normal operations and activate immediately on hardware failure, ensuring capacity is always available to maintain business continuity.