Nearly two years ago Dell Technologies took the storage array industry by storm with the introduction of PowerStore. In fact, the PowerStore architecture caused industry analysts to take note and in one case write, “This is the most important development in data storage hardware for many years and PowerStore will be the benchmark for the rest of the storage array industry to compare, contrast and compete with.

One trait that makes PowerStore the “benchmark for the rest of the storage array industry” is its native capabilities with regard to autonomous operations and machine learning functionality. What exactly are those game changing autonomous operations? Predicated on a modernized, container based storage operating system, PowerStore has the ability to autonomously monitor and manage cluster activities like data placement, volume migrations, and cluster expansions, as well as predictive capacity forecasting and autonomous internal pathing optimization.

Programmable Infrastructure

What we mean by programmable infrastructure is PowerStore’s ability to provide external programmability. More succinctly put, PowerStore is fully REST API programmable and provides a set of resources (objects), operations, and attributes that provide interactive, scripted, and programmatic management control of the PowerStore cluster.

Examples of REST programmability:

  • Provision new cluster components, such as appliances and volumes.
  • Configure replication destinations and sessions, and rules.
  • Fail over and fail back an appliance.
  • Create snapshots for backup purposes.
  • Gather metrics to use for historical analysis.
  • Gather configuration information and logs to use for auditing and trending analysis.

Although REST is growing in popularity and is the most versatile interaction method, there are still people who want to use shell scripting. In this case, PowerStore also provides a fully open CLI interface that can be leveraged from Microsoft or UNIX/Linux hosts.

And if software development kits (SDKs) are your thing, PowerStore supports programmability from Python and Microsoft PowerShell.

There are also automation plugins and playbooks for VMware vRealize OrchestratorAnsible, and container integration with KubernetesTanzu, and OpenShift.

Internal Automation and Machine Learning

While PowerStore has the external programmability aspect covered, the internal automation aspects are pretty incredible as well. There are MANY features native to PowerStore that are completely driven by artificial intelligence and machine learning (AI/ML).

Cluster Expansion

Adding appliances to a PowerStore cluster is a very simple process. Using the PowerStore Manager GUI, an administrator can use a wizard driven process to add an appliance as part of an initial cluster installation or anytime in the future. This process automatically discovers unconfigured PowerStore appliances on the network and walks you through a quick config wizard that auto-configures and joins the new appliance to the cluster.

Within a matter of minutes users have scaled-out their PowerStore environment with up to four appliances. All of these appliances are managed from the same instance of PowerStore Manager and can be viewed together from the Dashboard tab or individually from the Hardware tab.

Automated Data Placement

When working with a scaled-out cluster an admin may want to quickly create volumes. Having more than one appliance can often mean tedious decision making and careful, manual balancing of resources. PowerStore has simplified this via an integrated analytical engine known as the resource balancer.

This is a screenshot of the volume creation wizard of PowerStore Manager. Note the process for volume creation which includes the typical requirements of the name the volumes, the quantity of volumes to be created, and the size of the volumes being created.

Here is where the AI/ML engine kicks in. There is a “Placement” option on the volume creation screen. In this example you can see from the drop down that there are two appliances available in this cluster. An administrator may want to pin the volumes to a specific appliance in the cluster (perhaps one is a higher performing model) and he/she would simply select that appliance. The “Auto” option can also be selected and allows the cluster intelligence to take over and assign the volumes autonomously between the two appliances. This can save an administrator massive amounts of time and expedite the volume creation and balancing process.

Imagine a scenario where you want to quickly create tens, dozens, or maybe even 100 volumes. Where is the best place to put them? How quickly can you get this done? Now imagine PowerStore completely automating that process and assigning the rapidly created volumes instantly to one appliance or another keeping things balanced. This is the power of PowerStore.

Assisted Migration

While the above mentioned automated data placement capability is more initial volume creation, PowerStore is constantly monitoring all appliances in a PowerStore cluster. Take capacity for example. Let’s say that an appliance starts running low on physical capacity. PowerStore can make real-time migration recommendations. These recommendations are generated based on factors such as drive wear, appliance capacity, and health but the process itself is autonomous.

In the above screenshot you can see that PowerStore appliance A1 is projected (via machine learning) to run out of space in 8 days. Clicking on this alert will provide more details.

You can see we now have a remediation option called “Assisted Migration”. Selecting Assisted Migration will allow the admin to select volumes limited on space and PowerStore will then migrate those volumes to another appliance that meets the space requirements and needs of the volume. The admin doesn’t have to look at node utilization, pool capacity or any other metrics. PowerStore logic handles this autonomously.

Dynamic Node Affinity

Dynamic Node Affinity – or DNA – as I call it, is one of my favorite machine learning /autonomous features of PowerStore. This feature was introduced in PowerStore OS 2.1 that was released in January, 2022.

PowerStore’s ease is well known and admins do not have to assign volumes to pools or storage nodes. This is an autonomous process at volume creation. PowerStore sets the backend node affinity of a volume automatically making administrative life simple. While this creation process is simple, we know that workloads can change over time. As the array environment takes on more load over time the original backend affinity mapping may not be the most optimal for performance. Enter DNA! Dynamic Node Affinity will autonomously change the backend affinity node guaranteeing that PowerStore is always running at optimized and peak performance.

DNA is comparing and assessing data in real-time meaning that it is a true inline machine learning process. Before any action is taken on the backend affinity pathing PowerStore runs through several checks and validations comparing IOPS, CPU utilization, and overall latency for block workloads. If it is determined that greater performance can be achieved then the node affinity it switched. This process is completely autonomous and 100% transparent to any host connected to PowerStore.


With PowerStore administrators need not worry about constant monitoring, manual balancing, and migrations. The intelligence of the PowerStore architecture is built for autonomous operations and simplicity. This means administrative ease of use and confidence in overall performance and balanced resource utilization.

Dell is bringing enterprise storage to market with external programmability and internal autonomous operations by integrated machine learning and artificial intelligence. PowerStore is truly next generation and there is a reason that it is the fastest growing storage product in Dell Technologies history. Expect more… PowerStore more!

A guest post by Jodey Hogeland

Leave a Reply