Breakthrough Performance with Tree-based File Systems

Using the right file system is one of the most critical factors in achieving system performance, and also the most often overlooked. The amount of data stored and processed on embedded devices is increasing rapidly. As ever larger amounts of data bog down


Using the right file system is one of the most critical factors in achieving system performance, and also the most often overlooked. This is true for desktops, servers and even more so for embedded devices. The amount of data stored and processed on devices is increasing rapidly, initially driven by consumer devices like Smartphones, MP3/video players, DVR, etc. This trend is now carrying over to other segments of embedded industry including Industrial, Military, and Automotive. Customers expect highly responsive devices regardless of the amount of data stored. As ever larger amounts of data bog down system performance, device manufacturers compensate by adding faster processors and more memory, which in turn increases device cost and impacts battery life.

In this paper, we will look at a more efficient way of achieving the device responsiveness – improving data management performance by using a highly optimized file system. The paper will first evaluate file system architectures based on certain key constructs that have significant impact on its performance and then provide an overview of a new file system that employs these constructs.