Media Aware Storage Framework (MASF)
Solving the Challenges of Media Modernization
MASF enables cost-efficient improvements in performance, latency and endurance over existing storage solutions through a building-block and storage framework. It assists with accelerating applications using fast caching, fast storage, and intelligence to navigate a multi-cloud system, increasing scale per server and reducing transaction costs for latency-sensitive workloads in order to deliver faster time to insights.
How it Works
MASF is a framework for optimal data placement across a broad range of environments including media types, storage architectures, storage methods, and operating systems.
Hinting Generator
IO classification identifiers from the application are delivered to the data placement block to enable the system to determine optimal data placement.
Data Policy
Data policies define rules for processing IO to different media pools, based on the various IO classifications.
Data Placement
Placement combines hints with policies to select appropriate media pool, optimizing system utilization.
Data Analyzer
A feedback loop analyzing workloads and placement trends in real time to provide policy recommendations to optimize system utilization.
Data Layout
Identifies and executes optimized layouts within the pool assigned by data placement.
Media Pool
A collection of one or more devices of the same media type presented to the data placement engine as available storage resources.
Why MASF?
Workloads across rapidly evolving segments such as HCI, HPC, AI and Databases are becoming more intense, demanding more data to be analyzed and at higher speeds.
In response, core count growth continues its relentless pace and highspeed networks such as 100GbE are becoming commonplace. Likewise, new storage media such as Intel® Optane™ SSDs and Intel® QLC 3D NAND have emerged to help break through storage bottlenecks and store massive amounts of data efficiently. But existing storage dynamics that assume a homogeneous and simple media structure cannot realize the full value of these media types. To unlock the value of modern media, and remove storage bottlenecks in your system, a more intelligent and data-driven approach to data placement is required.
The MASF design approach helps storage developers drive down storage cost while meeting the SLA requirements of increasingly demanding workloads.
With MASF, storage developers can define rules for processing IO to different media pools and combine hints with policies to select the most optimal pool. Furthermore, a feedback loop analyzing workloads and data placement provides policy recommendations ensuring policies are not static and unresponsive to the changing nature of the workload.
Combined, the building blocks of this framework help developers address the traditional storage focal points of cost, capacity, throughput and tail latency.