IT pros who manage storage infrastructures are facing ever-increasing complexities, escalating security threats and unprecedented amounts of data. If storage issues arise, they can result in significant downtime or compromised data. Modern application workloads demand storage solutions that are flexible, proactive and capable of responding quickly to all types of issues. Enter predictive analytics for storage.
To address the changing needs of IT teams, a number of vendors now offer storage products that incorporate predictive analytics and its related technologies, making it possible to proactively address issues at multiple levels in the storage stack.
Although offerings vary from one vendor to the next, they all employ predictive analytics to some degree, collecting a wide range of telemetric data to help ensure the accuracy of their analysis. IT teams evaluating storage systems should determine the ability of those products to support predictive analytics so the organization can be better prepared for today's dynamic and data-intensive workloads.
How predictive analytics works for storage
Predictive analytics uses statistical methodologies to uncover patterns in historical and current data, and it then forecasts specific outcomes. As part of this process, predictive analytics uses technologies such as data mining, analytical queries, predictive modeling, AI and machine learning.
Data storage products have been steadily incorporating these advanced technologies to reduce downtime, improve resource utilization and optimize application workloads. In many cases, vendors offer these tools in conjunction with their own storage platforms to provide a more complete storage infrastructure that can address data center needs.
Machine learning, a sub-discipline of AI, plays a vital role in predictive analytics for storage by automatically training the predictive algorithms using a constant influx of telemetric data from across the storage stack. A tool that incorporates predictive analytics continuously collects and analyzes the data, identifying patterns that can help forecast trends, optimize components, discover hardware failures, pinpoint bottlenecks and predict potential issues before they arise.
Predictive analytics in the market
Pure Storage utilizes predictive analytics for storage in its Pure1 service. Pure1, a cloud management service, includes Pure1 Meta, which uses predictive analytics and machine learning to analyze real-time data from over 10,000 cloud-connected Pure Storage arrays, with visibility into more than 100,000 workloads.
Pure1 Meta continuously monitors storage systems to proactively resolve issues before they cause serious problems. The service provides information about the health and function of the entire infrastructure stack, including virtual machines (VMs). The service also offers AI-based forecasting to determine capacity and performance requirements over time and to model hardware and workload consolidation options.
In addition, Hewlett Packard Enterprise has enhanced its InfoSight cloud management service to support predictive analytics and AI-based operations. InfoSight collects and analyzes billions of sensor data points from over 9,000 customer environments. These include not only Nimble and 3Par storage arrays, but it also includes compute, network and virtualization components throughout the infrastructure stack.
With this data, InfoSight can provide global insights into system status and health, helping to forecast and prevent problems, as well as optimize performance and resource usage. The service continuously analyzes and correlates millions of sensor measurements every second, constantly learning from the data in order to deliver more effective results.
Hitachi Vantara has also jumped into the market, offering an AI-based suite of operations software. The suite includes the Infrastructure Analytics Advisor, which uses predictive analytics, machine learning and other advanced technologies to detect anomalies, perform root cause analysis and predict behavior. It can also help plan infrastructures, improve performance and resource usage, and optimize systems across multivendor storage environments.
The Hitachi suite also includes Automation Director, an AI-based orchestration tool for managing and delivering IT resources. Automation Director automates core IT service delivery operations across various components, ranging from VMs to data protection mechanisms to SAN zones.
Another storage vendor to embrace predictive analytics is DDN Storage's Tintri suite, including its Tintri Analytics, a cloud-based service for modeling capacity and performance requirements. Powered by Apache Spark and Elasticsearch, Tintri Analytics offers machine learning algorithms to create storage and compute models that can extend 18 months into the future, using up to three years of data collected from hundreds of thousands of environments.
Making the most of predictive analytics
By incorporating predictive analytics into storage strategies, IT teams can take a proactive approach to identify and address issues before they negatively affect storage systems. The use of predictive analytics for storage can also lead to better performance and resource utilization, while lowering administrative and support overhead.
Even so, predictive analytics is a relatively young technology when it comes to storage. To be effective, it requires an ample amount of accurate data and algorithms that can properly assess that data. If either requirement isn't met, an organization can end up addressing the wrong issues or not catching issues before they become critical, resulting in wasted time and money.
When evaluating storage systems that incorporate predictive analytics, IT teams should perform due diligence to ensure they're getting products that can perform reliable analytics based on the most relevant data. Only then will they realize the full benefits that predictive analytics can bring to their storage infrastructures.