MegIS: High-Performance, Energy-Efficient, and Low-Cost Metagenomic Analysis with In-Storage Processing (2406.19113v1)

Abstract: Metagenomics has led to significant advances in many fields. Metagenomic analysis commonly involves the key tasks of determining the species present in a sample and their relative abundances. These tasks require searching large metagenomic databases. Metagenomic analysis suffers from significant data movement overhead due to moving large amounts of low-reuse data from the storage system. In-storage processing can be a fundamental solution for reducing this overhead. However, designing an in-storage processing system for metagenomics is challenging because existing approaches to metagenomic analysis cannot be directly implemented in storage effectively due to the hardware limitations of modern SSDs. We propose MegIS, the first in-storage processing system designed to significantly reduce the data movement overhead of the end-to-end metagenomic analysis pipeline. MegIS is enabled by our lightweight design that effectively leverages and orchestrates processing inside and outside the storage system. We address in-storage processing challenges for metagenomics via specialized and efficient 1) task partitioning, 2) data/computation flow coordination, 3) storage technology-aware algorithmic optimizations, 4) data mapping, and 5) lightweight in-storage accelerators. MegIS's design is flexible, capable of supporting different types of metagenomic input datasets, and can be integrated into various metagenomic analysis pipelines. Our evaluation shows that MegIS outperforms the state-of-the-art performance- and accuracy-optimized software metagenomic tools by 2.7$\times$-37.2$\times$ and 6.9$\times$-100.2$\times$, respectively, while matching the accuracy of the accuracy-optimized tool. MegIS achieves 1.5$\times$-5.1$\times$ speedup compared to the state-of-the-art metagenomic hardware-accelerated (using processing-in-memory) tool, while achieving significantly higher accuracy.

• The paper introduces MegIS, a cooperative in-storage processing system that minimizes data movement to improve metagenomic analysis performance.
• It employs a hardware/software co-design to extract, sort, and compare k-mers, achieving a speedup of 2.7x–37.2x over traditional tools.
• MegIS significantly reduces energy and cost overheads, cutting energy consumption by 5.4x–15.2x compared to state-of-the-art methods.

MegIS: High-Performance, Energy-Efficient, and Low-Cost Metagenomic Analysis with In-Storage Processing

The paper "MegIS: High-Performance, Energy-Efficient, and Low-Cost Metagenomic Analysis with In-Storage Processing" presents a significant contribution to the domain of bioinformatics by introducing MegIS, an in-storage processing (ISP) system designed to optimize the metagenomic analysis workflow. This system aims to alleviate the substantial data movement overhead typically encountered in standard metagenomic analysis, thereby enhancing performance, reducing energy consumption, and improving cost efficiency.

Background and Motivation

Metagenomics involves analyzing genomic fragments from multiple species within a sample, making it distinctly more complex than traditional genomics, which deals with isolated species. The typical metagenomic workflow includes sequencing, basecalling, and metagenomic analysis. The latter is notoriously data-intensive, requiring the movement of large volumes of data from storage systems to the main memory and processing units. This extensive data movement constitutes a significant performance bottleneck.

The authors identify that traditional methods and even recent hardware-accelerated methods do not adequately address this bottleneck. For instance, state-of-the-art metagenomic tools like Kraken2 and Metalign, suffer from I/O overheads due to the large size of reference databases they must query. Even advanced systems leveraging processing-in-memory (PIM) fail to eliminate this issue, as data still needs to be moved from storage to memory.

MegIS: Concept and Design

The core innovation of MegIS is its design as a cooperative ISP system that orchestrates data processing both inside and outside the storage device. This synergistic approach involves a hardware/software co-design to leverage the strengths of the SSD's in-situ processing capabilities while minimizing the data movement to and from the host system.

Key mechanisms and steps of MegIS include:

1. Data Preparation and K-mer Extraction (Step 1):
   • MegIS extracts k-mers from the input read queries and sorts them lexicographically, then partitions them into buckets that are processed and transferred in batches to the SSD. This step minimizes the amount of data transfer required and is performed on the host due to its superior computational resources and larger DRAM.
2. Intersection Finding and Taxonomic Identification (Step 2):
   • This step, performed inside the SSD, involves finding the intersection between query k-mers and the reference database k-mers stored on the SSD. MegIS reads data directly from the flash chips and performs lightweight computation, such as comparing k-mers and retrieving taxIDs using a specialized in-storage data structure called K-mer Sketch Streaming (KSS).
3. Abundance Estimation (Step 3):
   • MegIS allows for integration with different abundance estimation techniques, either through lightweight statistics or more precise read mapping. MegIS creates a unified index of reference genomes directly in the SSD, streamlining the process of read mapping.

Experimental Results

The evaluation of MegIS demonstrates impressive results:

• Performance: Compared to state-of-the-art tools like Kraken2 and Metalign, MegIS achieves a speedup of 2.7x–37.2x and 6.9x–100.2x, respectively, on various SSD configurations.
• Energy Efficiency: MegIS significantly reduces energy consumption, exhibiting a 5.4x reduction compared to Kraken2 and a 15.2x reduction compared to Metalign.
• Cost Efficiency: By offloading intensive data movements to the SSD and avoiding the need for high-bandwidth interfaces or extensive DRAM capacity, MegIS also shows notable improvements in system cost-efficiency.

Implications and Future Directions

Practically, MegIS offers a scalable and cost-effective solution for enabling high-throughput metagenomic analyses, making it suitable for applications in precision medicine, environmental monitoring, and infectious disease surveillance where rapid and accurate genomic insights are critical. Theoretically, MegIS paves the way for integrating ISP technology into other data-intensive bioinformatics applications, potentially addressing similar data movement bottlenecks.

Future developments could explore enhancing MegIS's capability to handle even larger datasets and integrating it with emerging sequencing technologies that perform real-time analysis. Additionally, further research could aim to refine the hardware accelerators and explore more advanced algorithmic optimizations specific to different types of genomic data.

Conclusion

MegIS stands out as a highly efficient system tailored for metagenomic analyses, leveraging in-storage processing to address fundamental data movement challenges. By co-designing hardware and software components, MegIS not only enhances performance and energy efficiency but also improves the overall cost-efficiency of metagenomic analysis workflows. This work represents an important step forward in making high-throughput, accurate genomic analysis more accessible and sustainable.