AI workloads will span on-premises, edge and cloud, making hybrid-by-design strategy essential: HPE Storage

Can you walk us through HPE’s data storage solutions portfolio? 

Our storage business unit is part of the larger hybrid cloud business unit, which is one of three primary business units at HPE. For storage, there are three pillars within our organisation — our unstructured data platform, structured data solutions and data protection capabilities. We have built out our product lines and road maps and we innovate along those three paths. But in the structure and non-structured space, we have products supported by a disaggregated, scaled-out multi-protocol architecture that gives outsized benefits to our customers. They are a more monolithic and shared-nothing architecture, so we invest heavily not only in the architecture but in the software and solution capabilities to deliver value to our customers.  

We help customers extract value from their data through the portfolio and advanced tech architectures that enable them to deploy at scale and cost-effectively. The challenge for the team is to continue to evolve and enhance our innovation in those areas. 

Where does AI come into the picture? 

AI is helping create this data explosion. We are building platforms that scale to meet compute-intensive workloads like AI, and that’s where our architecture plays a role. We are also building intelligence into the Storage platform. For example, we introduced the HPE Alletra Storage MP X10000, an intelligent data platform, in November last year. It supports object and unstructured data for purposes and provides inline data enrichment, a fundamental capability to unlock the value of data as it’s being stored on the device. It is built on top of a key-value store. That platform is intended to continue evolving and keeping pace with rapid advancements and requirements. 

We have two primary core storage platforms — HPE Alletra Storage MP B10000 and HPE Alletra Storage MP X10000. On the former, we announced Cyber Resilience capability integrated into the platform, which does inline anomaly detection. We also announced a unified file capability, which expands the platform. On the MP X10000, we announced support for the NVIDIA AI data platform, ensuring we will continue to partner with NVIDIA to develop AI capabilities over time. We also announced the first set of data intelligence capabilities in the platform, and in particular, our ability to do instant Retrieval Augmented Generation (RAG). We have the ability, as we consume the data, to create vector embeddings, and make that data readily accessible so it doesn’t operate as a batch process that runs analysis over the data. It calculates the vector embeddings and stores them, making them accessible in real-time. That is an AI capability we are bringing to the market. 

Given the massive volumes of data being processed by AI, do you see an increasing demand for data storage solutions as opposed to technologies of the past?

It is certainly driving unique and important requirements in storage. We are seeing customers increasingly having to come up with a data strategy to support their AI projects, which in many cases require data previously stored in a public cloud to be repatriated into their private clouds or on-premise environments. To process the vast amounts of data, you need storage close to the compute; this is a major trend driving customer behaviour. Initially, there was a big wave of data and workloads going to the cloud. Now, customers realise which workloads behave better or are better-placed on-premises versus the cloud. AI is certainly a workload that will live on-premises, on the edge and in the cloud as well, which is why it is so important to have a hybrid-by-design strategy.

Do you primarily address the requirement of data centres?

We sell to and tailor solutions for small- and medium-sized business customers and enterprises. Customers require a heterogeneous set of capabilities. Whether remote offices, data centres within their environment, or extensions into the public cloud, our solutions allow us to have mobility across all those environments, a key element to an efficient operation. Customers are looking for many ways to save money so they can invest in other value creation. We intend to provide a platform that can extract value from their data and operate at the lowest cost of ownership, which gives them the ability to focus on other value-creation activities.  

What challenges do enterprises deal with when it comes to data storage?

There is an explosion in data. Zettabytes of data are being created, further fuelled by an explosion in AI. Customers are recognising they can gain operational efficiency out of their data, or create value that allows them to interact with their customers more effectively. So, they are storing and saving more data. One must position the data to leverage and extract value from it, which is where many customers see a challenge.

Are enterprises leaning more towards public or private cloud adoption, or do they continue to pursue a hybrid cloud strategy?

We see everything. However, the hybrid cloud has become the predominant IT mode of operation. Around 10 years ago, there was a rapid movement of workloads into the public cloud. Over time, customers found an equilibrium where some could be managed cost-effectively, and they met the security and performance requirements. The cost requirements and other workloads belong on-premises, and we have seen a repatriation of workloads back. As a result, hybrid is here to stay. We will have customers with a mix of workloads on-premises. They might manage them in their private cloud, and or in a public cloud. In our HPE GreenLake cloud platform, we give customers the ability to have a cloud experience and a cloud management model on-premises. 

How is the future of data storage solutions evolving? 

Data storage solutions will continue to evolve to support these rapidly emerging AI workloads and leverage AI within the platform to do more enhanced and advanced provisioning of data, drive greater efficiencies and do more predictive analytics.

If you want to be competitive in your specific market, you have to extract value from data. This will continue to fuel innovation in the foreseeable future.