The problem
Fabric BCDR: All-or-Nothing, Vendor-Controlled, Proprietary Restore
Microsoft Fabric's built-in Business Continuity and Disaster Recovery does exactly one thing: it protects your entire capacity, all at once. You can't say "back up just this Lakehouse" or "replicate only these two Delta tables to cold storage." There's no item-level granularity — it's the whole workspace or nothing, and you're paying for it either way.
The deeper problem is what happens when you actually need a restore. Your data goes back to Fabric — on Microsoft's timeline, through Microsoft's workflow — into a format you can't touch until the restore completes. You can't point DuckDB at the backup to verify a table. You can't run a Spark job against it overnight while capacity is paused. The data is theirs until they hand it back.
And for teams that want disaster avoidance — not just recovery — there's no way to maintain a live, queryable replica outside Fabric that a different compute engine can use if Fabric is unavailable.
With native BCDR you don't choose what gets backed up, where it lives, or how you read it back. That's not a backup strategy — it's a vendor dependency.
The solution
LakeReplica: Granular Backup. Your Storage. Your Compute.
LakeReplica is a purpose-built replication tool for Microsoft Fabric. It moves your Lakehouse and Warehouse data — at whatever granularity you choose — to any destination you control: Azure Blob Storage, ADLS Gen2, AWS S3, or another OneLake workspace. Data lands as open Delta Parquet files, immediately queryable by any compute engine without a Fabric licence or a running capacity.
Primary use case: proper BCDR — item-level backup jobs, automated schedules, full migration history, and data you can actually read back on your own terms. Secondary benefit: once the ETL is done and capacity is off, your analysts and data scientists can keep querying the silver and final layers with DuckDB, Spark, or Polars directly from storage — no CUs, no waiting, no spin-up cost.
Pick exactly which Lakehouses, Warehouses, or table paths to back up. No forced full-workspace policies.
Backup lands in open format — readable by DuckDB, Spark, Polars, Trino. No proprietary restore workflow required.
Target Azure Blob, ADLS Gen2, AWS S3, or another OneLake workspace. You own the destination and the data.
Schedule recurring jobs — daily, hourly, post-pipeline — so your off-Fabric copy is always current.
Run DuckDB or your own Spark cluster against the backup directly. No Fabric capacity, no CU bill for reads.
Data never passes through our servers. AzCopy device-code auth — no MSAL tokens, no service principal setup.
Side by side
Fabric BCDR vs LakeReplica
| Capability | Fabric Built-in BCDR | LakeReplica |
|---|---|---|
| Choose specific items to back up | ✗ All-or-nothing workspace | ✓ Granular item selection |
| Storage destination you control | ✗ Microsoft-managed only | ✓ ADLS, S3, Blob — your account |
| Open Delta Parquet output | ✗ Proprietary restore format | ✓ Query immediately, any tool |
| Read backup without restoring | ✗ Must go through Fabric restore | ✓ DuckDB / Spark direct on storage |
| Query when Fabric capacity is off | ✗ Requires capacity or restore | ✓ Own compute, zero CU cost |
| Fabric-to-Fabric workspace copy | ✗ No cross-workspace replication | ✓ Full OneLake-to-OneLake support |
| Scheduled automated jobs | ✓ Capacity-level only | ✓ Item-level, custom cadence |
Real-world use case
Backup First. Then Never Pay Capacity Just to Read It Back.
The primary job is solid BCDR. Your Lakehouse contains months of curated silver and gold Delta tables — LakeReplica backs up exactly the items you care about, on a schedule you control, to storage you own. If Fabric has an incident, you have clean Delta Parquet files sitting in your ADLS account, ready to restore or query independently — no Microsoft restore ticket required.
The secondary payoff is cost. Your nightly pipeline finishes, LakeReplica copies the final tables, then your Fabric capacity pauses. At 9 AM the analyst runs ad-hoc queries against the same backup files with DuckDB or their own Spark cluster:
# Verify a backup table — or run ad-hoc analysis — while capacity is off
import duckdb
con = duckdb.connect()
con.sql("""
INSTALL delta; LOAD delta;
SELECT region, product, SUM(revenue) AS total
FROM delta_scan('abfss://backup@youraccount.dfs.core.windows.net/silver/sales/')
GROUP BY 1, 2 ORDER BY 3 DESC
""")
# Same files — your own Spark cluster, no Fabric licence needed
df = spark.read.format("delta") \
.load("abfss://backup@youraccount.dfs.core.windows.net/silver/sales/")
df.groupBy("region").sum("revenue").show()
The backup becomes a live, queryable analytics tier — not a dormant archive locked in a proprietary vault. Share the storage path with external teams that don't have a Fabric licence. Use it as a cost-efficient secondary compute layer. Or just know it's there if Fabric goes down.
"With LakeReplica, your backup isn't just insurance — it's data you can actually use, on your own compute, any time capacity is off."
How it works
Three Steps, Zero Lock-in
Pick a Fabric workspace, then choose the specific Lakehouse, Warehouse, or folder path you want to migrate. No forced all-workspace selection.
Point to another OneLake workspace, Azure Blob Storage, ADLS Gen2, or an S3-compatible bucket. Your data, your infrastructure.
AzCopy handles the transfer server-side — no data through our servers. Schedule it to trigger at the end of your pipeline, then safely pause your Fabric capacity. Your team queries the copy with their own engine at any time.
🚀 Application Coming Soon
Be First to Get Access
Full Fabric Lakehouse migration, open Delta output, zero lock-in. Join the waitlist and we'll reach out the moment we launch.
Join the Waitlist →lakereplica.com · No spam. Launch announcement only.