Azure NetApp Files Cool Access GA: Dynamic Throughput for Premium and Ultra Volumes

The Azure Updates entry, titled "Generally Available: Storage with cool access enhancement," was published September 29, 2026 with the "Launched" label. Microsoft defines that label as a fully released, production-ready product available to all Azure customers. One point to get straight first: this is an Azure NetApp Files feature, not a new Azure Blob Storage access tier. Blob tiering rules and pricing do not apply here.

What changed​

Microsoft's "What's new in Azure NetApp Files" page now lists the feature as GA. It describes it as an enhancement to storage with cool access for the Premium and Ultra service levels that more precisely aligns throughput with data tiering.

The key change: when cool access is enabled, maximum throughput is dynamically calculated based on the amount of data tiered to cool access storage, rather than applying a fixed reduction. Microsoft also says hot data retains its configured performance, and throughput is adjusted only when data is tiered to the cool tier.

Microsoft's stated benefit is more predictable QoS behavior while optimizing performance and cost as data access patterns evolve, without requiring manual tuning or reconfiguration.

Timeline:

In short: cool access has been GA since 2024. What went GA now is a smarter way of setting throughput limits on Premium and Ultra volumes that use it.

How the throughput math works​

Microsoft's cool access documentation gives more detail than the announcement. It says when cool access is enabled on the Premium and Ultra service levels, throughput is adjusted at regular intervals based on the size of the data in the cool tier. It also sets a threshold: if a volume is within the Premium or Ultra service levels with auto QoS enabled and has more than 100 GiB of cool access data, the throughput QoS limit is different for the data on hot and cool tier respectively.

The calculation is done per volume, not per pool. Microsoft notes that throughput is calculated per volume, not per capacity pool. Volumes within the same pool that have different amounts of data in the cool tier have different maximum throughput. This difference is expected.

That last point will matter to anyone reading monitoring dashboards. Two volumes in the same Ultra pool can show different throughput ceilings, and that is how the feature is designed to work, not a fault. Microsoft publishes per-volume formulas in its "Throughput in the Premium and Ultra service levels" guidance. Check those before you open a support case.

How it compares with Flexible​

The Flexible service level already works differently. Microsoft says Flexible service level capacity pools with cool access maintain the user-configured throughput limits. Unlike Premium or Ultra pools, performance isn't reduced when cool access is enabled.

The enhancement doesn't make Premium and Ultra behave like Flexible. Throughput on those tiers can still drop as data cools. The drop is now proportional to how much data has cooled rather than a flat penalty. The following is our analysis, not Microsoft's wording: if your volume keeps most of its data hot, the new model should do much less damage to performance than the fixed cut did.

What it does not change​

The throughput change leaves the basic tradeoffs of cool access in place:

• Latency on cold reads is still there. Tiered blocks live in an Azure storage account. Microsoft's 2024 performance testing gave a sense of the gap: in one single-VM test, the maximum throughput for a single VM was approximately 400 MiB/s. Hot tier performance was around 2.75x better, capping out at approximately 1,180 MiB/s. That was one test setup, not a guarantee.
• The billing model is the same. Microsoft says when cool access is enabled on a volume, you benefit from a reduced price. You don't receive additional discounts specifically for the reduced bandwidth. Instead, you pay the cool access price, which inherently includes the reduced throughput.
• Metadata stays hot. NetApp has noted that metadata is never cooled and always remains in the hot tier.

Neither the announcement nor the documentation gives numbers for savings or throughput gains from the enhancement itself. Treat any such figures you see elsewhere with caution.

Practical checklist for administrators​

Microsoft says the enhancement needs no manual tuning or reconfiguration. Existing Premium and Ultra cool access volumes should get the new behavior without any action from you. If you're rolling out cool access for the first time, these documented details apply:

1. Configure the capacity pool first. Before you create or enable a cool-access volume, configure a capacity pool with cool access.
2. Know that pool conversion is one-way. You can convert an existing capacity pool into a cool-access capacity pool to create cool access volumes. After a capacity pool is enabled for cool access, you can't convert it back to a non-cool-access capacity pool.
3. Set the coolness period deliberately. The coolness period can be set between 2 and 183 days. The default value is 31 days. A short period tiers data sooner and saves more. It also raises the chance of pulling data back from the cool tier.
4. Expect a delay on small or new volumes. Microsoft says when you enable cool access on a newly created or nearly empty volume, a short delay can occur before tiering activity begins while the volume accumulates enough eligible data for the tiering engine to act on.
5. Check large-volume requirements. With cool access enabled, you can create large volumes at sizes between 2,400 GiB and 7.2 PiB. For the 7.2 PiB tier, you must be registered to use large volumes up to 7.2 PiB. With large volumes up to 7.2 PiB, more than 80% of the data must reside on the cool tier.
6. Recheck throughput baselines. Volumes that were tuned around the old fixed reduction may now have different ceilings. Compare per-volume numbers before and after.

Scope and open questions​

The Azure Updates entry does not list regions or sovereign-cloud boundaries. During the preview, a Microsoft Learn page for Azure Government listed the feature as unavailable there. Nothing we found confirms that the GA release covers Azure Government, so admins in that cloud should check with Microsoft before planning around it.

Bottom line​

This is a small but useful release. Under the old fixed reduction, Premium and Ultra customers paid for top-tier performance and then took a flat cut once they enabled tiering. Throughput now tracks how much data has actually cooled, so hot data keeps its configured performance. The latency cost of reading cold data and the cool access pricing model are unchanged, and region and cloud coverage should be confirmed before rollout. For file shares with a small active set and a large archive, that makes cool access on Premium and Ultra easier to justify.