By: Ashley Markel on 4/7/2021 11:10:38 AM
NVMe, or Non-Volatile Memory Express, is a protocol released in 2011 for accessing high-speed storage systems media that came popular with the introduction of Solid-State Drives (SSDs). But what is NVMe and why is it important for machine learning and analytics applications?
As businesses and systems continue to consume and process increasingly more data to support manufacturing, it is important to rethink how this data is captured, preserved, accessed and transformed. And because of the speed, NVMe is revolutionizing how the user configures deep learning systems for data storage, data access, and overall architecture when combined with other more traditional storage methods.
This article will explain what NVMe is and share a deep technical dive into how the storage architecture works. Upcoming blogs will cover what features and benefits NVMe brings businesses, use cases where it’s being deployed today and how customers take advantage of NVMe SSDs. Additionally we will cover our platform Offerings and fully featured flash storage systems for everything from IoT Edge applications to Artificial Intelligence, AI, and machine learning systems.
NVMe builds off Solid State Drive (SSD) technology (similar to RAM) and introduces impressive write speeds of 3500MB/s moving the limit of actual data read/write speeds to the interface used.
NVMe was originally introduced as drop-in replacements to rotary drives, easing the transition for the consumer from slower rotation based hard drives and moving towards newer solid-state drives. The ease of adding SSD to existing SATA ports came at the expense of limiting data transfer to the SATA protocol, currently maxed out between 300-600 MB/s. Companies worked on standardizing the interfaces from 1990’s to the early 2000’s while also introducing speed improvements by leveraging different backplane technologies.
Referencing URTech, “It was not a surprise to anyone that SSD engineers kept working on improvements and today you have three primary options for SSD." It was the necessity to enhance the interface protocol speed to reach the theoretical max of 3500MB/s which naturally transitioned the SSD interface closer to the backplane of the PC; i.e. PCI Express bus. A simple way to look at it - utilizing SSD directly integrated into the backplane via PCI express bus is NVMe technology.
SSD options that exist today:
1. Rotary drives via traditional SATA port
2. SSD drives via traditional SATA port
3. (NVMe) SSD drives via PCI bus
Per Western Digital, The NVMe protocol capitalizes on parallel, low latency data paths to the underlying media, similar to high performance processor architectures. This offers significantly higher performance and lower latencies compared to legacy SAS and SATA protocols. This not only accelerates existing applications that require high performance, but it also enables new applications and capabilities for real-time workload processing in the data center and at the Edge.
AI and deep learning systems are generally GPU and RAM intensive processes requiring millions of transactions per second. When the user or application requires saving this data mid-process to a drive for analysis or manufacturing outputs the system can quickly become starved by the read/write system. This is where NVMe drives can help relinquish the bottleneck being placed on data storage. As the machine learning models used within industry moves from the cloud to the edge, and as we see an exponential rise in the amount of data being stored, it will be increasingly more important to eliminate the critical path of the read/write system. NVMe’s unique features help to avoid the bottlenecks from traditional scale-up database applications to emerging Edge computing architectures, and allow for scale-up to meet new data demands.
Stay tuned to as we provide data backed examples on the impact of NVMe on an advanced analytics machine learning application.