Synapse Migration
Plan Azure Synapse Spark pool to Microsoft Fabric capacity SKU sizing when migrating analytics workloads to a Fabric workspace.
Overview
Synapse-migration is an agent skill for the Operate phase that provides Synapse Spark pool to Microsoft Fabric capacity SKU sizing reference for migration planning.
Install
npx skills add https://github.com/microsoft/skills-for-fabric --skill synapse-migrationWhat is this skill?
- Synapse Spark pool node size/count → recommended Fabric SKU table (F8 through F256+)
- Planning note: workflow targets capacity already assigned to the workspace
- Fabric capacity shared across Spark, Warehouse, Power BI, and Pipelines
- Quick reference for Capacity Units, max Spark vCores, concurrent jobs, and burst
- Workload tiers from dev/test (<1 GB) through enterprise 20+ large nodes
- Synapse Spark pool → Fabric SKU mapping table with 6+ pool configuration rows
- Fabric SKU quick reference including Capacity Units and max Spark vCores
- Node tiers referenced: Small (4 vCore), Medium (8), Large (16), XL/XXL and 20+ node enterprise shapes
Adoption & trust: 24 installs on skills.sh; 427 GitHub stars; 3/3 security scanners passed (skills.sh audits).
What problem does it solve?
You are moving Synapse Spark workloads to Fabric but cannot translate existing pool vCores and node shapes into an appropriate shared Fabric capacity SKU.
Who is it for?
Independent data engineers or analytics solo founders planning Synapse → Fabric moves with shared capacity for Spark, warehouse, and BI.
Skip if: Greenfield apps with no Synapse history, or teams that only need notebook syntax without platform sizing.
When should I use this skill?
Planning or executing Azure Synapse to Microsoft Fabric migration and needing capacity SKU estimates from existing Spark pool configs.
What do I get? / Deliverables
You pick a defensible F SKU range from Synapse pool config and workload class before running the migration workflow on the assigned workspace capacity.
- Recommended Fabric F SKU range from Synapse pool vCore and workload profile
- Planning notes on shared capacity across Spark, Warehouse, Power BI, and Pipelines
Recommended Skills
Journey fit
Operate infra is the shelf for capacity planning and platform migration on live or near-production data estates, even though discovery starts earlier. Infra subphase covers shared Fabric capacity, Spark vCore mapping, and workspace-level SKU tradeoffs—not notebook feature coding alone.
How it compares
Capacity planning reference for Fabric migration—not a generic ETL generator or local SQLite skill.
Common Questions / FAQ
Who is synapse-migration for?
Builders maintaining Azure Synapse Spark pools who are adopting Microsoft Fabric and need agent help with SKU mapping and migration context.
When should I use synapse-migration?
During Operate infra when rightsizing Fabric capacity; during Validate scope when estimating cloud cost for a analytics migration; during Build backend when designing concurrent Spark job limits.
Is synapse-migration safe to install?
It is planning documentation; it does not execute cloud changes—review the Security Audits panel on this Prism page and your Azure governance before agent-driven deployments.
SKILL.md
READMESKILL.md - Synapse Migration
# Capacity Sizing Reference Use this table to estimate the Fabric capacity SKU needed based on current Synapse Spark pool configuration. This is a **planning reference only** — the migration workflow operates against whatever capacity is already assigned to the target workspace. > **Note**: Fabric capacity is shared across all workload types (Spark, Warehouse, Power BI, Pipelines) in the workspace. If other workloads run on the same capacity, size up accordingly. ## Synapse Spark Pool → Fabric Capacity Mapping | Synapse Pool Config | Total vCores | Typical Workload | Recommended Fabric SKU | Fabric CUs | Spark vCores Available | |---|---|---|---|---|---| | 3 Small nodes (4 vCore / 32 GB) | 12 | Dev/test, small datasets (<1 GB) | **F8** (dev) or **F16** | 8 / 16 | 8 / 16 | | 3–5 Medium nodes (8 vCore / 64 GB) | 24–40 | Standard analytics, medium datasets | **F32** | 32 | 32 | | 5–10 Medium nodes (8 vCore / 64 GB) | 40–80 | Production ETL, multiple concurrent jobs | **F64** | 64 | 64 | | 3–10 Large nodes (16 vCore / 128 GB) | 48–160 | Heavy ETL, large datasets (10+ GB) | **F64** or **F128** | 64 / 128 | 64 / 128 | | 10–20 Large nodes (16 vCore / 128 GB) | 160–320 | Enterprise workloads, many concurrent jobs | **F128** or **F256** | 128 / 256 | 128 / 256 | | XL/XXL nodes or 20+ nodes | 200+ | Large-scale data engineering | **F256+** | 256+ | 256+ | ## Fabric Capacity SKU Quick Reference | SKU | Capacity Units (CUs) | Max Spark vCores | Max Concurrent Spark Jobs | Burst | Typical Use | |---|---|---|---|---|---| | **F2** | 2 | 2 | 1 | Smoothed | Sandbox/POC only | | **F4** | 4 | 4 | 1 | Smoothed | Individual developer | | **F8** | 8 | 8 | 1 | Smoothed | Dev/test | | **F16** | 16 | 16 | 1–2 | Smoothed | Small team dev | | **F32** | 32 | 32 | 2–4 | Smoothed | Small production | | **F64** | 64 | 64 | 4–8 | Smoothed | Standard production | | **F128** | 128 | 128 | 8–16 | Smoothed | Enterprise production | | **F256** | 256 | 256 | 16–32 | Smoothed | Large enterprise | | **F512+** | 512+ | 512+ | 32+ | Smoothed | Enterprise-scale data engineering | > Spark vCores and concurrent job limits are approximate and depend on node sizes selected in Custom Pools and current burst utilization. Fabric uses **burst and smoothing** — short spikes can exceed the CU baseline, but sustained usage is throttled to the SKU limit. ## Sizing Decision Guide | Factor | How to Assess | Impact on SKU | |---|---|---| | **Peak concurrent notebooks/SJDs** | Count max parallel jobs during Synapse peak hours | More concurrency → larger SKU | | **Largest single-job resource need** | Check Synapse executor/driver memory configs | Large executors → need enough CUs to allocate them | | **Data volume per job** | Measure typical input dataset sizes | >10 GB per job → F64+; >100 GB → F128+ | | **Shared capacity with other workloads** | Will Warehouse / Power BI / Pipelines share this capacity? | Shared → size up 1–2 tiers | | **Burst vs. sustained** | Is Spark usage spiky (batch ETL) or continuous? | Spiky → can use smaller SKU with burst; sustained → size for peak | | **Dev vs. production** | Dev can use Starter Pool on F8; prod needs Custom Pool | Dev = F8–F16; Prod = F32+ | ## Cost Model Comparison | Aspect | Synapse Spark | Fabric Spark | |---|---|---| | **Billing unit** | Per-node, per-minute (when pool is active) | Per-capacity, per-hour (always-on or paused) | | **Idle cost** | Zero (auto-pause after timeout) | Capacity cost continues unless paused/deallocated | | **Scale model** | Node count autoscale (min–max per pool) | Capacity SKU (fixed CUs, burst smoothing) | | **Pause/resume** | Auto-pause per pool (minutes granularity) | Capacity pause/resume (via Portal or REST API) | | **Reservation pricing** | Azure Reserved Instances (1yr/3yr) | Fabric capacity reservations (1yr) | | **Trial** | N/A | Fabric Trial capacity (F64 equivalent, 60 days) | > **Cost tip**: For dev/test migrations, use a **Fabric Trial capacity** (free F64 for 60 days) or **F8 with p