Azure Elastic SAN: Pooled, Cloud-Native Block Storage That Truly Acts Like a SAN

Hi there People!

You probably have ever lived via a Friday night time SAN enlargement, racking new cabinets and praying the zoning held collectively, the thought of getting that very same shared block storage mannequin in Azure (with out proudly owning a single fibre channel cable) sounds nearly too good to be true. In his session on the Microsoft Azure Infra Summit 2026, Kiran Cherukuwada, Principal PM in Azure Storage, walked us via precisely how Azure Elastic SAN does that, and the place it suits subsequent to the opposite block storage choices on Azure.

📺 Watch the session:

Most of us have been taught a easy rule. One workload, one disk, dimension it for peak, transfer on. That rule has been type to managed disks, but it surely will get costly quick when you could have dozens or lots of of workloads that every one peak at completely different occasions. Elastic SAN flips the mannequin. You provision a pool of capability and efficiency as soon as, then carve volumes out of it for a lot of workloads.

Right here is why that issues for IT execs:

• You cease over-provisioning every workload to its personal peak; the SAN absorbs the bursts.
• You get a SAN-style useful resource hierarchy (SAN, quantity teams, volumes) that appears and behaves just like the on-prem mannequin you already know.
• iSCSI connectivity means a large compute footprint, together with Azure Digital Machines, Azure Kubernetes Service, Azure Container Situations, Azure VMware Answer, and Nutanix Cloud Clusters.
• You possibly can drive storage throughput over VM community bandwidth, which regularly helps you to maintain a smaller (and cheaper) VM SKU.

Briefly, if in case you have many IO-intensive workloads sharing one area, Elastic SAN is the lever that turns “purchase peak for each workload” into “purchase mixed peak for the group.”

Azure Elastic SAN is the {industry}’s first absolutely managed SAN storage service within the cloud. It brings the on-prem SAN consumption mannequin to Azure as a single managed pool of block storage, shared throughout many workloads, accessed over the industry-standard iSCSI protocol.

Contained in the service you get three sources, matching the on-prem psychological mannequin:

1. The Elastic SAN itself. Prime-level useful resource. That is the place you provision general capability and efficiency, and the place billing occurs.
2. Quantity teams. The place you set community guidelines (service or non-public endpoints) and safety insurance policies. Any coverage you apply right here is inherited by each quantity within the group, so a quantity group is successfully your workload boundary.
3. Volumes. The LUNs that you simply mount on compute. They present up as uncooked block gadgets on a VM, as iSCSI targets to a Kubernetes node, or as VMware knowledge shops on AVS.

A single SAN can scale to a petabyte of capability, 2 million IOPS, and 80 GB/s of throughput. It’s regionally redundant by default, with a zone-redundant choice, and shared quantity help is there for clustered options like SQL Server Failover Cluster Situations and Azure VMware Answer. Community isolation is delivered through service endpoints and personal endpoints, and knowledge is encrypted at relaxation. Incremental snapshots are supported for quick point-in-time restore, and snapshots could be exported to managed disk snapshots whenever you want a hardened copy for backup or DR functions.

The place does it land within the block storage portfolio? Kiran framed it merely. Premium SSD v2 is the most effective worth/efficiency for devoted per-workload efficiency. Extremely Disk is for the mission-critical, every-microsecond-matters workloads. Elastic SAN is the most effective worth/efficiency choice at scale, when you could have many workloads that may share a storage pool.

The economics dwell within the provisioning mannequin. You purchase two sorts of items:

• Base unit. Every base unit offers you 1 TiB of capability plus 5,000 IOPS and 200 MB/s. Roughly 8 cents per GiB per 30 days in East US.
• Capability-only unit. Every capacity-only unit offers you 1 TiB of capability however no further efficiency. About 25 % cheaper, round 6 cents per GiB per 30 days in East US.

The sample Kiran confirmed is “dimension for efficiency first, then high up capability.” A 250 TiB SAN delivering 1 million IOPS and 40 GB/s got here out to roughly 200 base items plus 50 capacity-only items, touchdown round 20 grand per 30 days for the entire pool.

The magic ingredient is dynamic efficiency sharing. With conventional disks you provision every workload to its personal peak. With Elastic SAN, you provision the mixed peak. So a SQL Server needing 60,000 IOPS, an AVS cluster needing 40,000, and an Oracle workload needing 100,000 IOPS seem like 200,000 IOPS of devoted disk. But when they by no means peak concurrently, you’ll be able to land a 150,000 IOPS SAN and let every workload hit its peak on demand. That’s actual a refund.

The second lever is throughput over community bandwidth. As a result of Elastic SAN connects over iSCSI, storage I/O flows via the VM’s community pipe, not the VM’s disk throughput cap. Most VMs have much more community bandwidth than disk bandwidth, so you’ll be able to drive greater storage throughput from a smaller VM SKU. That smaller SKU is cheaper to run, and (that is the quiet win) it may possibly additionally minimize per-core database licensing prices. As one attendee requested within the dwell Q&A, “Why is it potential to transcend the VM disk throughput restrict with SAN?” The reply: iSCSI visitors makes use of VM community bandwidth like every other VM-to-VM visitors, so the disk throttle doesn’t apply.

One sincere tradeoff: that very same community bandwidth can be utilized by your app-tier-to-database visitors. So if you’re planning to push storage exhausting, dimension the VM with each flows in thoughts.

The place does this really repay?

• Combined enterprise workloads on Azure VMs. SQL Server, Oracle, customized OLTP, sharing one SAN. Kiran’s demo ran SQL TPCC, an AVS cluster benchmark, and an Oracle OLTP load concurrently off a single 30-base-unit SAN, and the metrics blade confirmed precisely how every quantity group consumed efficiency.
• Extending Azure VMware Answer storage. As an alternative of shopping for costly vSAN nodes simply to develop storage, you join AVS to an Elastic SAN datastore. Gen2 AVS non-public clouds skip the ExpressRoute gateway requirement and allow you to use a single non-public endpoint on the quantity group.
• Container Storage. Azure Container Storage v2 with Elastic SAN backing is usually obtainable. The quick connect and detach habits signifies that even when a node or cluster goes down, the information sits on the SAN and persists.
• Raise and shift from on-prem SAN. Kiran shared one migration instance: a workload with 100-plus vCPUs working off a mid-tier all-flash SAN array landed on Elastic SAN with roughly 64 % TCO financial savings and efficiency that exceeded the unique array.

Briefly, it is a “many workloads, one pool” story. You probably have one heavy workload, premium SSD v2 could also be a greater match.

Here’s a sensible order of operations:

1. Dimension the SAN. Add up the mixed peak IOPS and throughput for the workloads you propose to consolidate, then choose base items to cowl efficiency and capacity-only items to high up storage.
2. Lock down the community. Entry is closed by default. Select service endpoints or non-public endpoints per quantity group, and open them solely to the appropriate subnets.
3. Place compute in the identical zone. For greatest latency, deploy your VMs (or AVS cluster) in the identical area and availability zone because the SAN.
4. Tune the shopper. Use Gen 5 (D, E, or M collection) VMs with Accelerated Networking on, configure the iSCSI initiator, arrange native MPIO on Home windows or Linux, and use the Join scripts from the portal which default to 32 classes per quantity.
5. Watch the metrics. The SAN’s Metrics tab exhibits transactions, ingress, and egress on the SAN, quantity group, and particular person quantity stage. Drop the granularity to 1 minute if you end up troubleshooting.
6. Plan snapshots. Use Elastic SAN quantity snapshots for quick dev/check restores. Export to managed disk snapshots whenever you want hardened backup or cross-region DR.

If you’re coming from on-prem, the partnership with Cirrus Knowledge (free within the Azure Market) is the advisable path emigrate storage on the block stage.

Catch the total Microsoft Azure Infra Summit 2026 session playlist right here

Cheers!

Pierre Roman