HPE’s Intelligent Storage Update

HPE Primera

Hewlett Packard Enterprise’s announcements this morning gave a welcome surprise to the enterprise storage world. Its annual showcase event, HPE Discover, is only two weeks away. Discover is usually where the good news is delivered, and today’s news is indeed good.

A crowded press release and blog post announced a flurry of features and enhancements that span most of HPE’s storage line-up, touching on nearly everything that the company holds strategic. There are updates to both HPE’s midrange and high-end storage offerings, enhancements to data protection and disaster recovery and even a fascinating glimpse into the future of intelligent infrastructure.  

Let’s take a look at what it all means.

A self-tuning storage array?

Hewlett Packard Enterprise talks a lot about intelligent infrastructure. HPE’s acquisition of Nimble Storage nearly three years ago was as much about bringing Nimble’s stellar InfoSight AI-driven predictive analytics platform into HPE as it was about closing HPE’s midrange storage gap.

HPE demonstrated this when it immediately set to work extending the reach of InfoSight across its entire data center portfolio. InfoSight now touches everything, from storage to supply chain, and it sits prominently as a cornerstone of HPE’s cybersecurity and cyber-assurance efforts. That’s an admirable level of innovation and execution for such a short period.

InfoSight’s secret sauce is its real-time access to operational data gathered from products deployed both in the field and in HPE’s own QA labs. This telemetry is not just data from ongoing operations—it also extends back in time, across the history of nearly every enabled product line.

InfoSight takes this wealth of data and correlates it against detailed profiles of known bugs, performance issues and even tricky operational issues. These correlations, if they show something interesting for a given installation, are passed along to HPE’s support team, who then reach out directly to the impacted customer. This is “predictive analytics” (if grossly simplified). Customers love this level of predictive support.  

Now imagine that, instead of correlating all of that data off-line to predict potential trouble, the array itself could possess some of that intelligence. That is the capability that Hewlett Packard Enterprise is delivering to Primera.

Primera’s new AI capability works something like this: InfoSight’s deep knowledge base feeds a machine-learning algorithm that maps telemetry data to potential operational issues, training a storage-savvy neural network in the process. That neural network is then delivered to Primera storage arrays installed across HPE’s customer base.

Primera, with today’s newly announced capabilities, continuously feeds its ongoing operational data into an onboard inference engine (which HPE confirms is a software-only solution, and does not use additional silicon). The inference engine operates on the storage-savvy neural network, resulting in a model continuously trained and refined for that array’s environment.

Primera’s locally trained model predicts application performance and resource needs. It also has some capabilities to optimize resource utilization and tune the array. It’s too early to know how well this approach works. HPE, in its analyst briefing before the launch, was vague about what sorts of optimizations and recommendations the engine would be trusted to produce. There are many unknowns, but I’d argue that the unknowns don’t matter—at least not in the long-term.

This represents another bold step by HPE towards delivering the underpinnings of an intelligent and autonomous infrastructure. The smart guys with their clusters of GPUs will figure out the math as we move forward. The critical accomplishment of HPE, today, is its enabling of the infrastructure.

Done well, this could be a game-changer. I fully expect that we will see HPE begin to push all of its enterprise infrastructure products in this direction as we move forward. 

Nimble gets faster and more reliable

Server-Class Memory (“SCM”) has found a home as an ultra-fast cache across the storage industry, with nearly every flash-array vendor using the technology to speed up read-heavy workloads. HPE joins that club with its announcement of an SCM cache capability in its Nimble line.  

HPE tells us that the new SCM caching will deliver up to a 2X performance boost on read-intensive applications, delivering sub-250 microsecond average latency. This is a welcome update that helps HPE close its performance gap in the midrange storage space.

SCM support will be released both as an option on new arrays and also as an upgrade kit for existing arrays. This is excellent news for HPE’s channel partners, who can now walk into existing installations and effectively double the performance of an in-place Nimble array. This is an attractive option for customers already facing a year of economic uncertainty.

Also new to the Nimble lineup is built-in, three-site, asynchronous replication. This allows disaster recovery to now encompass either a third physical site or replication to an HPE Cloud Volume. The most intriguing part of this new capability is the idea that customers are now allowed to treat an HPE Cloud Volume as a first-class citizen. That’s a new mindset for the HPE cloud team, and I’m anxious to see where it leads us.


This isn’t all that HPE announced today. Primera is also getting near-instant asynchronous replication over distance, an all-NVMe varian, and CSI driver support for Kubernetes. HPE’s VMware vVols support now includes SRM integration for both Nimble and Primera, while Nimble also picks up cross-stack analytics for Microsoft Hyper-V. Details of all of these new updates are described in an informative blog post on the HPE community website.

HPE delivered a robust and welcome set of announcements today, keeping it competitive in midrange storage with Nimble, while bringing new capabilities to Primera at the high-end. This is also the first set of announcements from HPE that doesn’t mention 3PAR. Primera’s replacement of 3PAR seems to have happened quickly.HPE is working to keep Nimble current, but the architecture is overdue for a refresh. The mid-range of the storage market is changing.

Dell Technologies demonstrated innovation with its new ground-up Dell EMC PowerStore architecture. IBM Corporation delivers a simply stellar amount of raw performance and capabilities into the midrange with its IBM FlashSystem 7200. Pure Storage, who just updated its FlashArray hardware earlier this year, continues to disrupt how we think about storage architecture with its own set of software announcements, released on the same day. It’s time to bring Nimble forward another step.

Hewlett Packard Enterprise is at its best when delivering on its mission to provide enterprises a consistent experience that is AI-driven, built-for-the-cloud and delivered as-a-service. HPE CEO Antonio Neri is guiding this strategy forward with a very steady hand.