As I called out in my last article, Pure Storage has an abundance of data and analytics collected from their global network of sensors. By using this data, Pure Storage is improving the way you can forecast your environmental workload and growth with the newest updates to Pure1 Workload Planner.
Pure1 has had a forecasting tool for a while, so what is so new about this? In days past, the forecasting tool would take data from current and historical workloads and project how long of a runway the array has when it comes to capacity. While this was a useful feature, it does not truly line up to the way the “real world” production data grows. Businesses expand, new more demanding software is deployed, data sets can expand erratically; none of which can really be projected linearly. Pure1 Workload Planner now has the capability to add variables to the forecasting model to give more realistic projections of future usage.
How is this Done?
Pure Storage has been collecting data from their pool of over 100,000 arrays world-wide. This information has compiled a large data lake for Pure1 Meta, Pure’s artificial intelligence and machine learning engine. Pure1 Meta has analyzed this data and created profiles for various workloads, which they call Workload DNA. Workload DNA profiles are built on an untold amount of production performance data and growth. Given this fact, these profiles are far more accurate than the old methods of analyzing growth over time and assigning a forecasted growth percentage based on historical growth.
Workload Planner in Action
When diving into the Planning page of Pure1, you will see the standard overview of the capacity forecasting as you had in the past. The updated feature set lies in the Simulate Workload button. This option takes you to an overview of the volumes on the selected array, presents a sortable list of the volumes, and provides 3 options for simulation:
Let’s dive into each option.
The Migrate feature allows admins to select a volume (or multiple volumes) and project the impact it would have on moving the workload to a different array in the cluster. The reduction in workload on the home array is also displayed in the modeling. Both impact on capacity and array workload are projected.
The Clone feature is exactly that: the ability to clone the volumes to multiple arrays and review the impact caused with the additional workload.
Scale is where Pure1 Workload Planner’s updates really shine. As stated earlier, not all workloads have static growth. Once the admin has selected the volumes, they can apply multipliers to said volumes. The multiplier can range from 0.5x, which cuts the workload in half, up to 5x and beyond. Scale applies these values to both workloads on the array and capacity projections.
Putting it All Together
What Pure1 Workload Planner does that makes these projections far more accurate is that the administrator can aggregate changes made on various volumes on the array. Starting with the baseline, volumes can be cloned, scaled, and migrated and the results will compound in the Simulation Summary. Often with data and workload growth, the existing hardware simply does not have the capacity to meet the changing workloads. Pure1 Planner can take the updated values generated by Workload Planner’s Simulate Workload feature and help size physical capacity changes with the Simulate Hardware option.
Simulate Hardware allows the admin to take the updated workload projections and apply the following options: Updating the array to a larger compatible controller, adding a disk shelf, or changing disk types. All of this information together provides a simple, best-in-class forecasting experience for admins to be able to accurately manage and operate their on-premises storage environment.
Over and over again, Pure Storage’s investments in their Pure1 platform are proving to return dividends for their customers. As the trend continues, it will be exciting to see where Pure takes the information provided by Meta AI/ML next.