Scale your critical applications cost-effectively with Azure Disk Storage

To address today’s challenges like ensuring business continuity and managing reduced cash flow, organizations are accelerating their migration to the cloud. With a wide range of workloads migrating to the cloud, it is important to meet the unique scale needs of these workloads while keeping costs under control. Today, we are sharing new capabilities on Azure Disk Storage that continue to help you scale your performance cost-effectively—allowing you to meet your workload needs on-demand and only pay for what you need.

Boost disk performance on-demand with new disk bursting experience on Premium SSDs

In April 2020, we announced the general availability of disk bursting on smaller premium SSD disk sizes (512 GiB or below). Today, we are extending disk bursting support (in preview) to larger premium SSD disks sizes (above 512 GiB) with a new on-demand experience. Unlike the credit-based system where you can only burst if you have accumulated credits, our new on-demand bursting capability allows you to burst up to six times (up to 30,000 IOPS and 1,000 MBs of throughput) of the provisioned amount whenever needed. This provides you with the flexibility to scale performance to meet demand and optimize costs.

Disk bursting is most suitable for mission-critical workloads where a letdown in performance cannot be tolerated even for unexpected spikes in demand. If you are hosting an online transaction processing (OLTP) system, you may experience a sudden spike in order requests due to an increase in demand. You want to ensure that your storage system can handle the increased load to the backend databases to provide a consistent experience to customers.

You can easily take advantage of this new on-demand bursting capability by enabling it on supported Premium SSD disks. There is a burst enablement fee, and any additional transactions over the provisioned target are charged separately. Learn more about the preview scope, enablement steps, and pricing model.

If you have already adopted our existing credit-based bursting, we recently introduced new disk bursting metrics. These metrics provide you with visibility on the accumulated burst credits and help you better forecast your ability to burst. You can easily view the baseline and burst performance targets and the unused bursting credits. Learn more about using new disk metrics to monitor disk bursting.

Achieve sustained higher performance with performance tiers for premium SSDs

For planned events like a seasonal sales promotion or running a training environment, you need to achieve sustained higher performance for a few hours or days and then return to the normal performance levels. With performance tiers on premium SSDs (general availability announced in November), you now have the flexibility to scale the disk performance without increasing the disk size by selecting a higher performance tier. You can also change tiers at any time or bring it back to your baseline performance tier, enabling you to achieve higher performance and cost savings.

A baseline performance tier is set based on your provisioned disk size. You can set a higher performance tier when your application requires this to meet higher demand without increasing the size. For example, if you provision a P10 disk (128 GiB), your baseline performance tier is set as P10 (500 IOPS and 100 MBs). You can update the tier to match the performance of P50 (7500 IOPS and 250 MBs) and return to P10 when higher performance is no longer needed. Learn more about how to change disk performance tiers.

Choosing the optimal option to scale performance dynamically for your workload

With performance tiers and on-demand disk bursting, you now have two options to scale your disk performance. We recommend that you consider the following criteria when evaluating the option that best fits your workloads:

  • Planned versus unplanned performance scaling: If you have a planned event needing a sustained increase in disk performance (for example, holiday season), performance tiers would be the more suitable option. You can set the disk tier to a higher performance tier to better accommodate the increased load and bring it back to the original tier at any time. However, if you are unable to plan ahead or accurately predict the performance pattern of your workloads, disk bursting would be a better choice as it provides you with a higher allowance beyond your provisioned target.
  • Duration: Depending on the duration of the performance demand, the cost efficiency of the two options could vary. For scenarios where high demand results from short-running jobs or jitters in IO scheduling, on-demand disk bursting will be more cost-efficient as you will only pay for the burst transactions. If your workload does not exceed the provisioned target, you only pay for the burst enablement fee, a small fraction of the disk cost. In contrast, if you expect your workload to burst for days or even longer, performance tiers will be the better option. For example, if you host a tax reporting solution, you can expect demand to increase significantly during the last month of the reporting season. Scaling your performance by selecting a higher performance tier would be more cost-effective for this situation.

We recommend you refer to the Azure pricing calculator to evaluate the best option for your workload’s needs.

Get started

To get started with the new performance scaling capabilities, see guidance on on-demand disk bursting and performance tiers. In addition, Azure Disk Storage offers other unique capabilities—like shared disks, dynamic scaling on Azure ultra disks, and bursting on smaller premium SSD disk sizes—that also enable you to scale and run business-critical applications cost-effectively.

Refer to the Azure Disk Storage documentation for detailed instructions on all disk operations. Please email us at AzureDisks@microsoft.com to share your feedback or leave a comment in the Azure Storage feedback forum.

Source: Azure Blog Feed

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.