Turbo-charging storage

Some have said that “data is the new oil of the digital economy.” Today’s businesses are increasingly applying techniques such as deep learning on their data to drive further insights and efficiencies and gain competitive advantage. Optimizing performance and latency for these newer analytic workloads requires full utilization of the underlying IT infrastructure. However, at scale, these compute environments become limited by I/O constraints, and traditional storage architectures become a key performance bottleneck.

Because traditional storage vendors haven’t had good answers, we are excited to invest and partner with WekaIO. WekaIO removes barriers between the data and compute layers, providing a high-performance, software-defined storage solution for the most demanding artificial intelligence (AI) and data-intensive workloads. When enterprises need to analyze massive amounts of data, at-scale, and reach a decision in milliseconds, WekaIO’s solution shines. It already helps autonomous vehicle companies’ self-driving cars react to changing road conditions in real-time. Other workloads that WekaIO accelerates include AI, machine learning, genomics, research, and analytics.

With our partnership, HPE is excited to address enterprises’ high-performance data needs. We already have joint solutions available with WekaIO software integrated on pre-built configurations of HPE product lines. Our CEO, Antonio Neri, envisions that enterprises will increasingly use machine learning and AI for greater business agility, and our partnership with WekaIO helps HPE power a data-driven world.

Storage Performance without Compromise

Deep Learning frameworks, such as Tensorflow and PyTorch, commonly require data to be pulled from disk to memory at high speeds. Traditional solutions suffice for small data sets, but for large data sets and data-intensive workloads, storage bottlenecks and I/O issues aren’t an option. Here, WekaIO’s Matrix parallel file system provides the industry’s fastest, most scalable storage access for AI and technical compute loads.

Historically, if you wanted storage that was simple to deploy and made it easy to share data, you used scale-out network-attached storage (NAS)—and compromised on speed. If speed was the top priority, you could use a custom-built (and extremely expensive) high-performance computing (HPC) system with block-based local storage arrays—and sacrifice scalability. If scale was the top priority, then you used a parallel file system, which was inevitably difficult to deploy and use.

By redesigning the storage architecture from the ground up, WekaIO is the industry’s first solution to bring everything together:

• Scale: The Matrix parallel file system provides massive scale for capacity, throughput and metadata. It supports hundreds of petabytes of NVMe and exabytes of object storage within a single name-space.
• Speed: WekaIO’s shared storage performance is 10x faster than an all-flash NAS and is even 3x faster than non-volatile memory express (NVMe) local disk storage. How? WekaIO’s file system parallelizes everything using extremely small block sizes. The company also developed unique IP around the way it distributes metadata evenly across infrastructure nodes. These innovations  solve the problem of equally addressing data services and metadata services that has plagued traditional file systems.
• Simplicity: WekaIO’s flexible deployments are a strong customer selling point. Its software can be deployed on-premises, as a converged architecture or dedicated appliance, or entirely in the cloud. Data management is also simplified through its fully automated data tiering capabilities between flash and disk storage.

Betting on the Future

Today, most companies use WekaIO to solve their high-performance storage needs where typical solutions just can’t get the job done. But it doesn’t take much imagination to envision where WekaIO could go in the future.

The self-driving cars that every auto manufacturer is now exploring. Industrial IoT solutions that need to recognize changes in complex systems and react immediately. DNA sequencing to develop more effective, personalized disease therapies. Even retail, where companies want to analyze video from thousands of cameras to understand customer behavior and deliver personalized service. Anywhere businesses need to quickly analyze vast numbers of images, huge amounts of sensor data, or apply deep learning, fast storage becomes a core requirement.

We think, in the next few years, more and more businesses will need the kind of performance at scale that only WekaIO can provide. Looking even farther ahead, we envision a world where AI and deep learning are applied to millions of applications—many of which we can’t even predict today. Fast storage will power a major chunk of the future of enterprise technology. By backing WekaIO, HPE aims to be one of the primary companies helping to deliver it.