Neural Trees: Using Neural Networks as an Alternative to Binary Comparison in Classical Search Trees
Douglas Santry, NetApp
Binary comparison, the basis of the venerable B Tree, is perhaps the most successful operator for indexing data on secondary storage. We introduce a different technique, called Neural Trees, that is based on neural networks. Neural Trees increase the fan-out per byte of a search tree by up to 40% compared to B Trees. Increasing fan-out reduces memory demands and leads to increased cacheability while decreasing height and media accesses. A Neural Tree also permits search path layout policies that divorce a key’s value from its physical location in a data structure. This is an advantage over the total ordering required by binary comparison, which totally determines the physical location of keys in a tree. Previous attempts to apply machine learning to indices are based on learning the data directly, which renders insertion too expensive to be supported. The Neural Tree is a hybrid scheme using a tree of small neural networks to learn search paths instead of the data directly. Neural Trees can efficiently handle a general read/write workload. We evaluate Neural Trees with weeks of traces from production storage and SPC1 workloads to demonstrate their viability.
Zoned namespace SSDs: Challenges and Opportunities
Zoned NameSpaces (ZNS) are a mechanism proposed in the NVM Express Workgroup to provide features and functionality similar to that of Open Channel SSD, but fully integrated with the NVMe model using a zone concept similar to that in the ZAC/ZBD extensions for SMR disk. The goals of this research are to investigate applications for ZNS SSD, in particular (a) RAID-like functionality over ZNS SSD, (b) strategies for file system support for ZNS, and (c) interfaces and strategies for direct application usage of ZNS SSD.
Hardware-Assisted Secure Flash-Based Storage
Modern storage systems have been developed for decades with the security-critical foundation provided by operating system (OS). However, they are still vulnerable to malware attacks and software defects. Adversaries can obtain the OS kernel privilege or leverage software vulnerabilities to bypass, terminate or destroy current malware detection and defense systems. For instance, encryption ransomware accounts for more than half of all malware attacks today, but current software-based defense systems often fail to enable the victims to say no to ransom collectors. Therefore, it is natural to utilize hardware techniques which have been proven effective in defending against malware attacks.