You’ve already noticed that the data center is in the midst of a revolution. It is no longer a world where you can simply unpack a server or storage array, rack it, and walk away. The rise of various clouds, edge computing, machine learning, and software-defined everything makes the world increasingly complicated. Your infrastructure is now intelligent, and leveraging that intelligence is a critical skill for every IT practitioner. Nowhere is that truer than in the management of the storage systems which serve the data that is the lifeblood of every enterprise.
What if your storage system could tell you when something is wrong before it happens? What if we could harness the collective intelligence of tens-of-thousands of storage arrays deployed across thousands of organizations to let our storage “learn” about the impacts of its environment? That’s what predictive analytics is all about.
“Predictive Analytics” is a term that arose over the past decade to describe the ability of a technology provider to deliver proactive support and guidance based on near real-time analytics. These capabilities are enabled by the ability to easily collect data from datacenter equipment, coupled with the rise and power of AI-enabled analytics. Predictive analytics allows storage administrators to leverage a constantly evolving dataset to almost predict the future. Short of predicting the future, it can undoubtedly spot problems before they occur.
This type of AI-enabled software also allows for proactive support engagements. When an issue is detected, it typically results in an automatically opened support ticket or call. It’s not at all unusual for an enterprise customer to be notified by a support engineer, for example, that there is an issue that will soon need attention.
AI-driven solutions take things even further, leveraging machine learning to juice the capabilities beyond mere monitoring and alerting and building intelligence directly into the infrastructure. Intelligent infrastructure is the beginning of the autonomous data center, and what Pure Storage calls “self-driving storage.” There are a couple of solutions that are beginning to enable self-driving storage. Let’s look at the ones who stand above the rest.
It is impossible to talk about predictive analytics for storage without leading off with a discussion about HPE InfoSight. Introduced by Nimble Storage in 2010, InfoSight is the forebear for what’s happening in predictive analytics today. It’s arguably the first such solution that was available for the storage industry. Described by HPE as the “crown jewels” when it acquired Nimble Storage in 2017, HPE has since taken the technology across its product lines. No longer just about storage, InfoSight attempts to offer insights across the infrastructure.
While InfoSight can today provide insights about storage, servers and other infrastructure components, it’s important to note that the views across the infrastructure remain siloed. Insights about your HPE servers, for example, don’t inform how your storage system is viewed. This will change as InfoSight proliferates across the data center. HPE’s data scientists will begin to unpack meaning in the trillions of data points to understand the complex relationships between storage and the rest of the infrastructure. It’s a powerful promise that could lead to a compelling future.
InfoSight is a cloud-based offering that generates AI-assisted insights from data collected from arrays across HPE’s extensive customer base. This allows for the sorts of actions that we’ve already discussed—from trouble identification to automatic support call generation. HPE’s InfoSight takes things even further, though, as it has become more intelligent about the infrastructure that connects to its Primera, 3PAR, and Nimble Storage offerings. InfoSight can identify noisy neighbors in a virtualized environment, for example, with its “vmvision” capabilities, as well as correlate performance information across host, network, and storage layers.
This level of environmental awareness helps to guide IT practitioners in tuning and managing the infrastructure. The guidance given by InfoSight is based on insights gleaned from combining ongoing performance characteristics of the storage array with the health of the infrastructure. Those metrics, correlated with learnings derived from the data of tens-of-thousands of arrays that contributed to Infosight’s knowledge base over the past decade, drive automated recommendations. That long-sounding process happens in the blink-of-an-eye. An InfoSight customer receives instant guidance. It’s no wonder HPE recognized the value in the approach. It just makes sense.
Pure Storage Pure1 Meta
What if you could take the basic premise of predictive analytics and extend it to the realm of “what if” modeling? The system already knows how your storage infrastructure is working. It’s not a leap to imagine modeling new capabilities or workloads where the historical data of your infrastructure intelligently feeds the recommendations. That’s how Pure Storage thinks about predictive analytics. It puts heavy emphasis on the “predictive” part of the process.
Pure Storage is a fun company to watch. It is innovative, aggressive, and cocky. It also has a steely discipline in executing its compelling vision of how storage systems should evolve to serve the rapidly changing needs of IT.
Pure Storage’s Pure1 Meta fits the company’s reputation to a tee. It takes the concept of AI-enabled predictive analytics and drives it just a little bit further. Pure1 Meta delivers a set of capabilities that make managing data and the systems that serve that data a more autonomous process. This is Pure’s play to usher in the age of “self-driving storage.”
Pure1 Meta is smart software. It collects telemetry data from across the company’s installed base, to the order of a trillion data points per day. It takes those trillion data points and derives insights. Pure calls these insights “fingerprints.” Pure1 Meta applies AI and machine learning techniques to correlate these fingerprints with what it knows about the operation of those systems. These fingerprints, gleaned globally, are leveraged to provide very localized guidance to an IT administrator.
I’ll save some virtual ink and tell you that Pure1 Meta provides all of the capabilities that you expect from a predictive analytics solution. Pure1 Meta is a cloud-based solution that provides health monitoring and alerting, automatic support case generation, software patching and maintenance, and a fleet-wide view of your storage infrastructure. Pure even provides apps for your phone to manage it all. This is where the comparison to the competition stops.
Storage systems should never be considered in a vacuum. A storage infrastructure must be measured against its ability to serve data that meets the multi-dimensional needs of the processes consuming that data. This is true whether the workloads are a bunch of VDI sessions or a high-performance cluster of deep learning machines. A predictive analytics solution for storage needs to operate in this world, making recommendations that extend outside the boundaries of the array itself. It needs to provide full-stack analytics.
Full-stack analytics is one area where Pure Storage is aggressively evolving the capabilities of Pure1 Meta. Leveraging what it calls “workload DNA,” Pure maintains a continuously refined set of profiles that becomes an intrinsic part of its analytics engine. This data then enables Pure1 Meta full-stack analytics.
The full-stack analytics offered by Pure1 Meta extend beyond the array. It provides real-time and on-going guidance based on the dynamics of the environment in which each array operates. Meta looks at performance and capacity. It also offers the ability to model new workloads to understand the impacts on the infrastructure beforehand.
Pure1 Meta has a significant set of workload planning capabilities. It allows you to model workloads, simulate the impact of hardware upgrades, and even model workload migration. It’s the ability to combine AI-enabled predictive analytics with a full-stack view of the infrastructure, along with Pure’s unmatched modeling and prediction capabilities, that make Pure1 Meta the new benchmark for Predictive Intelligence solutions. It’s not quite self-driving, but it gets close.
The power of predictive analytics is far more than just about fine-tuning the support environment. Predictive analytics is about utilizing the power of AI-driven analytics and machine learning to understand your environment. This helps you deploy and manage your overall storage infrastructure, both today and into the future.
Every storage vendor has an intelligent monitoring solution. NetApp’s OnCommand Insight is a competent offering. IBM ’s Storage Insights is compelling in that it marries predictive analytics with unique data placement capabilities, crossing on-prem and cloud. Dell EMC’s CloudIQ is rapidly evolving to provide predictive analytics capabilities across Dell’s storage line. All of these will continue to evolve. They simply must.
Nimble Storage brought predictive analytics to the storage world, but HPE slowed down InfoSight’s storage-centric feature set to migrate InfoSight across its portfolio. This is the right strategic call for HPE. It will ultimately allow the company to deliver an infrastructure-spanning solution where each element understands the others. That’s a very compelling future for HPE’s customers. In the near-term, however, InfoSight for storage remains siloed as the company focuses its efforts more horizontally.
Pure Storage has long been a bellwether for the storage industry. That continues today with Pure1 Meta. It shows the promise of what machine-learning driven predictive analytics can be.
I fully expect that every serious storage provider will ultimately deliver what Pure1 Meta provides today. Their customers will demand it. I hope it doesn’t take too long.