I'm facing a major challenge with migrating a large Azure Files workload from one region to another. We're working with a single storage account containing dozens of shares, with sizes ranging from 1GB to 15TB, all currently on the Transaction Optimized tier. Due to compute resources being fully utilized in our region, we need to move our Azure Files to a different region where compute can run efficiently.
The target storage account is the new Azure Files Provisioned v2, which promises significant cost savings. However, the move isn't as straightforward as we hoped. We initially tried using AzCopy, which has been quite disappointing, resulting in painfully slow transfer speeds—often just single-digit MBps. Our file mix includes many small documents and images which complicates things further since they're mainly accessed via API rather than SMB.
After tinkering with various settings for performance boosts, including RAM and concurrency levels, we still don't see the speeds we need for a 120TB project. I'm now considering alternatives but haven't tried Robocopy yet due to firewall and virtual WAN constraints. Ideally, I'd want to run multiple jobs simultaneously to improve throughput. Anyone have suggestions or insights on this?
4 Answers
Could you share the syntax you’re using for AzCopy? Sometimes the command parameters can make a big difference in how efficiently it operates.
Check out this thread for some additional tips: https://www.reddit.com/r/AZURE/s/l4uDvN9kj0. It might have some helpful insights for your situation.
Have you thought about using a cross-region restore from a recovery service vault? It might simplify things a bit, especially since you're migrating to a paired region. If you do that, you could delta sync with Robocopy until you've fully cut over. But remember, a 15TB restore will take some time, so plan accordingly!
120TB is a lot! Have you reached out to Microsoft support to see if they can assist with the migration on their end? Alternatively, using the Azure Data Box could be a viable option. The speeds you're experiencing seem pretty slow; consider if there’s a way to bypass the Azure Files layer for better performance.
I'm actually moving to a non-paired region, and we don’t use RSV for backups since this data isn’t primary. If it were to go down, it wouldn't be catastrophic, just inconvenient.