Workloads

The workload model aggregates all enabled scenarios into a single compute and storage demand summary. The aggregation logic lives in src/engine/workloads.ts.

Built-in workload sources

Surveyor can roll up demand from:

• Azure Virtual Desktop (AVD)
• AKS on Azure Local
• generic VM scenarios
• SOFS guest cluster
• MABS
• Arc-enabled service presets
• custom workloads

Each scenario contributes three numbers to the shared summary:

• total vCPUs
• total memory in GB
• total storage in TB

Important nuance

The workload summary is intentionally a roll-up layer. Some workload-specific math happens before aggregation:

• AVD computes session hosts, FSLogix storage, and bandwidth in src/engine/avd.ts
• SOFS computes profile demand, internal mirror compounding, and IOPS in src/engine/sofs.ts
• AKS computes control plane, worker, OS-disk, and PVC demand in src/engine/aks.ts

The aggregate numbers then feed:

• the Compute report
• the Final report
• exporters
• health checks

Scope boundaries

• AKS sizing in Surveyor covers base AKS infrastructure. Arc-enabled application platforms are modeled separately as service presets.
• RemoteApp is not yet a first-class workload type in the AVD data model. Current AVD estimates are the host-pool baseline.
• Custom workloads are the escape hatch for anything not represented by a dedicated planner.

Seeding from an existing estate (RVTools)

Custom workloads can be imported directly from an RVTools export so an existing VMware estate can drive an Azure Local sizing. Use Import RVTools on the Workload Planner and select either the full .xlsx export or the vInfo tab saved as .csv.

• VMs are grouped by their (vCPU, memory) signature — each distinct spec becomes one custom workload with vmCount set to the number of matching VMs. A fleet of identical hosts collapses into a single row.
• Memory is read from the Memory column (MiB → GB); storage is read from Provisioned (or In Use when provisioned is absent) and summed per group into the workload's logical storage (MiB → TB).
• Template VMs are skipped; powered-off VMs are included (they still occupy capacity) and reported in the import summary.