[Paper Review] Read Mapping Near Non-Volatile Memory

S. Karen Khatamifard, Zamshed I. Chowdhury+4 more|arXiv (Cornell University)|Sep 7, 2017

Network Packet Processing and Optimization44 references4 citations

TL;DR

GeNVoM proposes a novel, energy-efficient, and high-throughput read mapping accelerator that leverages resistive non-volatile TCAM to enable scalable similarity matching for short DNA reads. By introducing a variation-aware sense amplifier tuning mechanism and hierarchical similarity matching, it achieves up to 113.5× higher throughput and 210.9× lower energy consumption than a GPU baseline, overcoming TCAM's limitations in handling noise and approximate matches.

ABSTRACT

DNA sequencing is the physical/biochemical process of identifying the location of the four bases (Adenine, Guanine, Cytosine, Thymine) in a DNA strand. As semiconductor technology revolutionized computing, modern DNA sequencing technology (termed Next Generation Sequencing, NGS)revolutionized genomic research. As a result, modern NGS platforms can sequence hundreds of millions of short DNA fragments in parallel. The sequenced DNA fragments, representing the output of NGS platforms, are termed reads. Besides genomic variations, NGS imperfections induce noise in reads. Mapping each read to (the most similar portion of) a reference genome of the same species, i.e., read mapping, is a common critical first step in a diverse set of emerging bioinformatics applications. Mapping represents a search-heavy memory-intensive similarity matching problem, therefore, can greatly benefit from near-memory processing. Intuition suggests using fast associative search enabled by Ternary Content Addressable Memory (TCAM) by construction. However, the excessive energy consumption and lack of support for similarity matching (under NGS and genomic variation induced noise) renders direct application of TCAM infeasible, irrespective of volatility, where only non-volatile TCAM can accommodate the large memory footprint in an area-efficient way. This paper introduces GeNVoM, a scalable, energy-efficient and high-throughput solution. Instead of optimizing an algorithm developed for general-purpose computers or GPUs, GeNVoM rethinks the algorithm and non-volatile TCAM-based accelerator design together from the ground up. Thereby GeNVoM can improve the throughput by up to 113.5 times (3.6); the energy consumption, by up to 210.9 times (1.36), when compared to a GPU (accelerator) baseline, which represents one of the highest-throughput implementations known.

Motivation & Objective

• To address the growing computational challenge of accelerating short read mapping in next-generation sequencing (NGS) due to exponential data growth.
• To overcome the limitations of traditional TCAM in handling approximate matches and noise (e.g., insertions, deletions, substitutions) common in NGS data.
• To design a scalable, energy-efficient accelerator that leverages near-memory processing and non-volatile TCAM for high-throughput similarity matching.
• To enable practical deployment of TCAM-based acceleration in bioinformatics by solving its key drawbacks: high energy consumption and lack of similarity matching support.

Proposed method

• Introduces a novel similarity matching mechanism for resistive non-volatile TCAM that enables approximate matching without restricting row size, using variation-aware sense amplifier voltage tuning.
• Employs a hierarchical, multi-phase similarity matching approach that progressively refines matches, improving accuracy and scalability.
• Designs a genomic data representation that enables efficient similarity matching while maintaining low storage complexity.
• Applies search space pruning techniques tailored to the similarity matching mechanism to reduce unnecessary TCAM accesses and energy consumption.
• Optimizes the accelerator to handle the dominant noise types in short reads—substitutions, insertions, and deletions—by integrating them into the matching logic.
• Adapts to any resistive CAM array, including recent 45nm spintronic and 90nm PCM-based designs, ensuring broad hardware compatibility.

Experimental results

Research questions

• RQ1 Can non-volatile TCAM be effectively used for high-throughput, energy-efficient read mapping despite its inherent limitations in handling approximate matches and high energy consumption?
• RQ2 How can a TCAM-based accelerator be designed to support similarity matching under NGS-induced noise (e.g., substitutions, indels) without sacrificing scalability or area efficiency?
• RQ3 What novel data representation and pruning techniques can be used to minimize TCAM accesses and energy usage in a similarity-matching context?
• RQ4 To what extent can a variation-aware sense amplifier tuning mechanism enable approximate matching in resistive TCAM without restricting row length?
• RQ5 How does the proposed GeNVoM architecture compare to state-of-the-art GPU accelerators in terms of throughput and energy efficiency for short read mapping?

Key findings

• GeNVoM achieves up to 113.5× higher throughput and 210.9× lower energy consumption compared to a highly optimized GPU-based accelerator for read mapping.
• The proposed variation-aware sense amplifier tuning enables effective approximate matching in resistive non-volatile TCAM without limiting row size, a key enabler for scalability.
• The hierarchical similarity matching mechanism improves mapping accuracy while maintaining high throughput by progressively refining matches across phases.
• Search space pruning reduces unnecessary TCAM accesses, directly contributing to energy savings and improved efficiency.
• The solution is compatible with existing resistive CAM technologies, including 45nm spintronic and 90nm PCM-based CAM arrays, enabling practical deployment.
• GeNVoM is specifically optimized for short reads (100–200 bases), which constitute over 90% of current NGS data, making it highly relevant to real-world bioinformatics workloads.