Pure Storage Cloud 6.10.0 on Azure: NVMe Backend, 308TiB Arrays, and More Regions

The latest 6.10.0 release of Pure Storage Cloud on Azure introduces several significant updates, including a backend protocol upgrade to NVMe, a substantial increase in maximum capacity for specific array models, and an expansion into new global regions. These changes are designed to modernize Pure Storage Cloud and provide greater deployment flexibility.

Backend Modernization with NVMe on Azure

A core update in version 6.10.0 is the transition from a SCSI-based backend to an NVMe-based backend for Pure Storage Cloud SKUs that use Premium V2 SSDs. NVMe is a high-speed storage protocol that offers more direct communication with storage devices. This change can lead to lower latency, higher IOPS, and more efficient CPU utilization.

The primary impact of this enhancement is an increase in the backend IOPS and bandwidth caps imposed by Azure on the controller virtual machines. The table below illustrates the performance uplift for the affected MP2R2 SKUs:

Table 1: NVMe backend compared to SCSI.

From the table above, it’s clear that both IOPS and bandwidth are seeing significant improvement, positively influencing certain workloads. An increase in backend IOPS is expected to bring benefits in a mixed read/write workload with small IO sizes. An increase in backend bandwidth can be beneficial for non-reducible mixed read/write workloads with high array utilization.

Consider the Pure Storage Cloud V20MP2R2 SKU as an illustrative example for bandwidth consumption. This SKU utilizes the Azure e32bds v5 VM, which shows the following performance characteristics:

• Frontend network bandwidth for reads: 1.8GBps (usable)
• Backend bandwidth for connections to backend disks: 2.7GBps (combined for both read and write operations)
• Provisioned bandwidth on backend SSDs: 2.5GBps
• Provisioned bandwidth on the two NVRAM disks: 900MBps

If a non-reducible mixed read/write workload demands 1.8GBps for reads and 900MBps for writes, the array experiences a bottleneck at the VM backend due to the following:

• User-requested reads consume 1.8GBps from the 2.7GBps backend bandwidth allocation.
• That leaves 900MBps available for NVRAM write operations, which would be sufficient. However, NVRAM data is eventually flushed to SSDs for persistent storage.
• This process involves first reading data from NVRAM and then writing the data to SSDs. Each of these two operations consumes 900MBps of bandwidth from the total backend allocation.
• The total required backend bandwidth becomes 1.8GBps (user-requested reads) + 900MBps (user-requested writes) + 900MBps (flush – reads from NVRAM) + 900MBps (flush – writes to SSD), totaling 4.5GBps.
• This exceeds the 2.7GBps of backend bandwidth available with SCSI.

Transitioning to the NVMe backend resolves this bottleneck by providing 4.8GBps of backend bandwidth. Furthermore, it’s important to note that this example shows a non-reducible workload. Workloads with even a conservative 2:1 data reduction rate become feasible under these conditions (but can still get bottlenecked elsewhere).

How to get this feature: The transition is fully automated. Upgrading an array to version 6.10.0 will replace the existing SCSI-based controller VM with an NVMe-enabled equivalent. There are no manual steps, no changes to the user interface, and no additional costs. As of this release, NVMe is the only supported backend protocol for Pure Storage Cloud on Azure going forward.

Increased Maximum Capacity for V20MP2R2 Arrays

This release brings a significant capacity expansion for the Pure Storage Cloud V20 array on Azure by utilizing Premium v2 SSDs (V20MP2R2). The maximum raw capacity has been increased from 168TiB to 308TiB.

It’s important to note that this enhanced capacity is exclusively available for newly deployed arrays running Purity version 6.10.0 or higher. Existing arrays will not inherit this increased capacity limit upon upgrade.

For a comprehensive understanding of the achievable capacity points across various configurations and array types, please refer to the detailed table below:

Table 2: Achievable capacity points for various configurations.

New Azure Regions Now Available

Deployment options for Pure Storage Cloud with Premium v2 SSDs have been expanded to include several new Azure regions:

Table 3: New Azure regions.

A subset of these new locations is designated as “no-zonal regions,” which operate with a single data center. According to Azure, Premium v2 SSD performance (throughput and latency) might be lower compared to regions with multiple availability zones. Testing suggests these regions are best suited for arrays not operating at consistently high utilization (above 70%).

Additional Platform Updates

Other highlights from this release include: 

• More restrictive managed disks policies: The NetworkAccessPolicy and PublicNetworkAccess policies for managed disks are now explicitly set to their most restrictive values. This change is applied automatically upon upgrade with no customer impact.
• Other Purity improvements: This release also incorporates additional security updates, fixes, and general improvements from the Purity 6.10.0 release. Please consult the main Purity release notes (Pure1 login required) for a comprehensive list.