HPE GreenLake has utilized new CPUs and denser SSDs to bring more AI and data lake workloads to customers looking for an alternative to the public cloud.
HPE is adding new all-flash nodes to its GreenLake file storage, which will increase the offering's performance and density of its as-a-service offering. It is aimed at supporting AI and data lake use cases.
HPE GreenLake for File Storage now has two new flash node options that double the performance over the previous version through an upgrade to the latest server CPUs, thereby lowering power use per performance. The updated shelves also support denser SSDs of up to 60 TB per drive that fit into a single rack unit (RU). The new all-NVMe just a bunch of flash (JBOF) storage unit can house up to 1.356 PB in a single RU.
Enterprises and vendors are continuing to rapidly invest in AI workloads, according to Scott Sinclair, an analyst at TechTarget's Enterprise Strategy Group.
"A large percentage of organizations have initiatives for this year to buy more storage for their AI environment," he said, referencing ESG survey data that he said indicated more than half of respondents plan to increase generative AI storage spend in 2024.
HPE is looking to bolster storage offerings for data-rich use cases in places where customers already are, such as file systems, Sinclair said.
Storage for the data masses
HPE released its GreenLake for File Storage last April in partnership with Vast Data, which uses the HPE Alletra MP unified array.
While AI adoption was making headlines last year, it was still nascent, according to Gokul Sathiacama, vice president of unstructured data at HPE Storage. As the requirements for AI became clear, HPE has added performance and density to storage hardware to meet those needs.
Most AI workloads are more easily accessible in the public cloud, according to Dave Pearson, an analyst at IDC. But that AI infrastructure is targeted at the biggest companies. Now, HPE and other storage vendors are providing options for AI workloads that targeted at mainstream enterprises.
"Announcements like this make it more possible for enterprises to operate their own infrastructure for AI," he said.
As more tools for AI become available, the shift from public cloud to hybrid environments will accelerate, Pearson said.
HPE GreenLake for File Storage is based on Vast Data's technology, Sinclair said.
"Vast's differentiation is that they're targeted at low latency workloads for AI environments," he said.
Predictive AI, for example, becomes more accurate the more data that is fed into it, which requires an AI environment to scale as the data grows, Sinclair said. That ability to scale aligns with Vast Data's software capabilities, and HPE is using those capabilities on its own hardware to provide an as-a-service model for customers, he said.
"What we're really seeing here is the potential of partnerships and innovation in different areas," Sinclar said.
Next generation storage
Dell expanded its own file offerings a few weeks ago and HPE today, storage vendors are taking advantage of the benefits of the latest technology, according to Steve McDowell, founder and analyst at NAND Research. He pointed to Intel's Xeon Scalable processors, fifth generation processors released in December 2023, that support the interconnect standard PCIe 5, which can double the bus speed for connected devices.
Announcements like this make it more possible for enterprises to operate their own infrastructure for AI.
Dave PearsonAnalyst, IDC
"I think we are months away from all the vendors having the same CPUs, but right now, it is a horse race to see who has the fastest controllers," he said.
Vendors are competing to get the latest technology to market quickly, which provides storage options when customers are looking to buy, McDowell said.
Release cycles for components such as new CPUs or new types of SSDs tend to land close together for the storage major vendors, Sinclair said.
GreenLake for ease of upgrade
HPE GreenLake competes with similar infrastructure and data services from vendors such as Dell and NetApp, but it might have an advantage, Pearson said.
"[In GreenLake, the] enterprise-wide cloud management and existing data services are both mature and easily integrated into this offering," Pearson said.
GreenLake customers also don't have to wait for hardware upgrades to storage technology. Given the as-a-service model, HPE would provide the installation services for customers, which would avoid potential system downtime, according to the vendor.
This is a core feature of the HPE Alletra MP platform, McDowell said. Customers can swap in the controller without swapping out the storage so that storage and compute can be scaled together or independently.
"That's the beauty of the disaggregation that they do," he said.
Sinclair also said that AI environments can be multi-layered and complex, and storage as a service can be a useful investment.
"It's eliminating a number of different complexities in terms of steps or multiple steps in that it is optimizing and tuning that environment, because HPE is doing that work for you," Sinclair said.
Adam Armstrong is a TechTarget Editorial news writer covering file and block storage hardware, and private clouds. He previously worked at StorageReview.com.
