GPU Scheduler for De Novo Genome Assembly with Multiple MPI Processes

TL;DR

This paper introduces three GPU schedulers for the ELBA algorithm to improve performance in de novo genome assembly, enabling multiple MPI processes to efficiently utilize GPUs and achieve significant speed-ups.

Contribution

It proposes novel GPU schedulers that allow multiple MPI processes to share GPUs effectively, enhancing ELBA's scalability and performance in genome assembly tasks.

Findings

Achieved 7-8x speed-up with the best scheduler.

Significant performance improvements over baseline ELBA.

Trade-offs identified between parallelism and scheduler overhead.

Abstract

$\textit{De Novo}$ Genome assembly is one of the most important tasks in computational biology. ELBA is the state-of-the-art distributed-memory parallel algorithm for overlap detection and layout simplification steps of $\textit{De Novo}$ genome assembly but exists a performance bottleneck in pairwise alignment. In this work, we proposed 3 GPU schedulers for ELBA to accommodate multiple MPI processes and multiple GPUs. The GPU schedulers enable multiple MPI processes to perform computation on GPUs in a round-robin fashion. Both strong and weak scaling experiments show that 3 schedulers are able to significantly improve the performance of baseline while there is a trade-off between parallelism and GPU scheduler overhead. For the best performance implementation, the one-to-one scheduler achieves $\sim$7-8$\times$ speed-up using 25 MPI processes compared with the baseline vanilla ELBA GPU scheduler.