Click to learn more about author Zivan Ori.

This is Part Three of a three-part series: Part One is here. Part Two is here.

After comparing NVMe storage against local SSD, it is clear to see the NVMe storage is a better solution for AI/ML infrastructure across industries, regardless of their needs. Highlighted in the first part of the series, examples of how NVMe storage works alongside AI/ML are extremely visible and extend beyond robots and plots played out on screen. Common use case examples of automation include the following:
 

■ Financial Analytics – Automation and AI are used to fuel decision-making processes for investments in the financial services and financial technology (FinTech) industries. By mixing historical data and financial modeling, one platform can provide the horsepower required for predicting future investment strategies for their financial customers.

■ Image Recognition in Manufacturing – Automation in manufacturing has a long history of being applied to production lines, increasing the output capacity of their production systems, scaling from hundreds of units to thousands or even millions of units per hour. The financial impact of a quality issue on the production line can be disastrous if not caught in a timely manner. Taking photos of manufactured parts by using image recognition in real-time is essential to determining whether a part meets the required quality standards while also noting systematic quality issues as they happen.

■ Car Services – Ride sharing apps have paved the way for a new paradigm in public transit, allowing users and drivers to connect as quickly and easily as needed. Ride sharing companies use AI/ML to position drivers where they are most needed based on both past and current ride sharing requests. This reduces the amount of time spent driving, while also increasing customer satisfaction through shorter wait times, both of which improve the revenue potential for the ride sharing companies.

Beyond AI/ML, one vendor also provides more generalized computing services for their customers. They provide storage capacity for cloud services, using OpenStack and Kubernetes in conjunction with NVMe storage for high performance storage. They also leverage NVMe storage for big data analytics, using spark applications to perform multiple types of data analytics tasks, such as SQL, data mining and more.

Summary: Benefits of NVMe storage for AI/ML

NVMe storage is an ideal solution for countless AI/ML workloads, especially machine learning for multiple applications. By applying NVMe storage to machine learning for multiple applications, you can:

■ Create and manage larger shared datasets for training – By separating out storage capacity from the compute nodes, datasets for machine learning training can scale up to 1PB. As a result, the dataset grows, bringing additional NVMe storage online, increasing performance and avoiding limitations set by legacy storage controller bottlenecks.

■ Overcome the capacity limitations of local SSDs in GPU nodes – The limited space for SSD media in GPU nodes restricts the ability to manage larger datasets. Conversely, NVMe storage enables NVMe volumes to be dynamically provisioned over high performance Ethernet or InfiniBand networks.

■ Accelerate epoch time of machine learning by as much as 10x – NVMe storage eliminates the latency bottlenecks of older storage protocols and unleashes the parallelism inherent to the NVMe protocol by leveraging high performance NVMe-oF. Every GPU node has direct, parallel access to the media at the lowest possible latency.

■ Improve the utilization of GPUs – Having GPUs rest idle due to slow access to data for processing is costly. By offloading storage access to the idle CPUs and delivering storage performance at the speed of local SSDs, NVMe storage ensures that the GPU-nodes are kept busy with fast access to data.