Data continues to grow at a staggering rate, and companies continue to find more ways to use that data to make better business decisions. That’s great news, but it can present real challenges to the way data is stored, managed and accessed. Decision-makers today need fast access to large data stores, and failure is not an option. Data must be always available and always redundant, and able to grow quickly without outgrowing its storage.
In other words, what most organizations need today is a more software-defined, modular scale-out storage infrastructure—one that can handle big data analytics and automation, and one that can work seamlessly in both on-premises and cloud environments.
Hewlett Packard Enterprise’s new Primera storage system wants to be that storage platform. The solution uses modern technologies such as artificial intelligence (AI) and machine learning, NVMe and storage-class memory, and services-oriented software designs. The company says these features give the platform significant benefits in terms of scalability, availability, functionality and performance.
Essentially, Primera combines the best of several of HPE’s products and acquisitions—the simplicity of Nimble storage, the mission-critical heritage of 3PAR and the intelligence of InfoSight—into one large enterprise storage solution, said Eric Burgener, a research vice president with IDC.
The HPE Primera storage system uses a scale-out architecture, which supports massive parallelism. “As a storage protocol technology, NVMe supports at least three orders of magnitude higher parallelism than SCSI, and HPE Primera embraces this with a design that is highly optimized for solid state storage,” Burgener said in a recent white paper on the topic.
With HPE’s new "all-active architecture," all controllers and all cache are active at all times, he explained, resulting in low latency and high throughput. It also can accelerate applications significantly; recent testing by HPE found that it resulted in 122% faster Oracle performance and 93% time savings.
In addition, the platform has a modular, service-centric design for the storage OS. Because it is built this way, storage administrators can add or modify features without having to recompile the entire operating system. This enables more frequent, faster updates that are easier to implement and less risky to perform.
Because the platform integrates HPE InfoSight, it can take advantage of InfoSight’s AI and machine learning to evaluate system performance, energy consumption and capacity utilization, as well as data collected from virtual machines, software, applications, servers, cloud storage and networks. This results in a global view of a system or workload state at any point in time, Burgener wrote in the white paper.
Finally, a streamlined multi-node hardware design, architected for high availability, allows it to store up to 1PB capacity in 2U, or 2PB in 4U. Each controller node includes up to 12 host ports per node with 25GbE or 32Gb FC connectivity and redundant, hot-pluggable controllers, disk devices, and power and cooling modules.
While there are competing solutions on the market, Burgener said Primera has made a real effort to fully address application service availability, and has made it easy to use and upgrade. “They also don’t always support the massive parallelization necessary to optimize performance and capacity utilization of the solid state media that has come to dominate high-end systems. This can hinder workload consolidation and consistent performance,” he added.
Users can either purchase the HPE Primera storage system outright or deploy it as a managed service, paying only for what they use through the HPE GreenLake consumption-based IT solution.
