Questions on Combining Azure Reserved Instances, Savings Plans, Databricks DBU, and Data Factory RI Consideration

Hello Swaroop Gowda D,

Let me break this down by each area you asked about - Reserved Instances vs Savings Plans, Databricks DBU pre-purchase, and Azure Data Factory reserved capacity.

1.Reserved Instances (RI) and Savings Plans

• Yes, you can use both together. Reservations are always applied first to matching resources (same region, size family, etc.).
• Savings Plans kick in next - they cover any eligible compute usage not already matched by an RI.
• If neither applies, the workload is billed at pay-as-you-go (PAYG) rates.
• Think of it as: RI = precise discounts, Savings Plan = flexible discounts.

Documentation: Decide between a savings plan and a reservation

2.Azure Databricks DBU Pre-Purchase

1. Databricks does not use RIs; instead, you can pre-purchase Databricks Commit Units (DBCU) that are applied against DBU usage.
2. Your usage range (4–17 DBU/hour):
   • Monthly DBU-hours = DBU/hour × ~730 hours
   • 4 DBU → ~2,920 DBU-hours/month
   • 17 DBU → ~12,410 DBU-hours/month
   • Yearly DBU-hours = DBU/hour × 8,760 hours
   • 4 DBU → ~35,040 DBU-hours/year
   • 17 DBU → ~148,920 DBU-hours/year
3. Cost formula: DBU-hours × Price per DBU. (Replace with your region’s actual DBU price — e.g., if $0.40/DBU, that’s ~$1.1K–$5K per month across your range.)
4. Savings: Pre-purchase gives up to ~37% off PAYG DBU rates, depending on 1- vs 3-year term. It’s best to pre-purchase only the baseline you know you’ll use 24×7, and let spikes run on PAYG.

Documentation: Azure Databricks Unit pre-purchase

3. Azure Data Factory Reserved Capacity

• Data Factory reserved capacity applies to mapping data flows compute (General Purpose, Memory Optimized, etc.), purchased by region, compute type, and core count.
• In your case: General Purpose with 4 cores (+4 driver cores) → you can buy a reservation that matches this setup.
• If your workloads are steady and predictable, reserved capacity significantly reduces costs vs PAYG. For bursty workloads, PAYG may be more flexible.
• You can use the Reservations blade in the Azure Portal to compare the estimated savings before purchase.

Documentation: Reserved capacity in Azure Data Factory

Summary:

• RIs and Savings Plans work together RIs first, Savings Plans second.
• Databricks → use DBCU pre-purchase if you have predictable baseline DBU usage (saves up to ~37%).
• ADF → reserved capacity is valuable if you run mapping data flows consistently at your current core count.
• For exact savings numbers, I’d recommend plugging your usage into the Azure Pricing Calculator and the Reservations blade in the portal. Hopefully this gives you a clearer picture. Let us know if you have further questions. Thank you!

hi there ) thanks for bringing these cost questions here, its a smart move to think about this stuff.

okay, lets break this down. first up, mixing reserved instances and savings plans. yes, u can absolutely use both together in the same subscription. azure's billing system is designed to handle this. it will apply your ri discounts first to the eligible vms, then the savings plan will kick in to cover other compute usage like container instances or azure functions. the key thing is, u won't double dip on discounts. the billing applies the best discount for each resource, so u are safe. its a great way to maximize savings across different types of workloads.

for your databricks question, calculating those dbu costs. since your workload varies between 4 and 17 dbu per hour, u need to find your average hourly usage. lets say u average around 10 dbu per hour. then u multiply that by 24 hours and 30 days to get a rough monthly estimate. 10 * 24 * 30 = 7200 dbu per month. u can then check the azure databricks pricing page for the prepurchase discount tiers. usually, prepurchasing dbus in a commit tier like 100k dbus per year gets u a significant discount versus pay as you go. its all about committing to a volume upfront.

this might help in other tools too, always worth modeling your average consumption before committing.

about data factory and reserved instances... for your general purpose setup with 4 cores plus 4 driver cores, u are essentially looking at 8 cores of integration runtime. purchasing an ri for data factory can be a bit tricky because the ri covers the vcpu hours of the ir, but not the actual pipeline execution costs. u need to make sure u have a stable, always on ir to make the ri pay off. if your ir runs 24/7, the ri can save u up to 30 percent over payg. but if it's intermittent, the savings might not be there.

aha, and here is a pro thought. use the azure pricing calculator to model both scenarios. plug in your core counts and estimated hours for data factory. then, compare the payg cost to the one year ri upfront cost. the calculator will show u the savings percentage.

worth looking into azure cost management's cost analysis tool. it can show u your past usage for databricks and data factory, which makes forecasting so much easier. u can see your actual dbu consumption and ir uptime right there.

hope this gives u a clearer path forward. optimizing costs is a journey, but these tools make it much easier. let me know if u have more questions.

Best regards,

Alex

and "yes" if you would follow me at Q&A - personaly thx.
P.S. If my answer help to you, please Accept my answer

https://ctrlaltdel.blog/