FireCloud, a scalable cloud-based platform for collaborative genome analysis: Strategies for reducing and controlling costs
Chet Birger, Megan Hanna, Edward A. Salinas, Jason Neff, Gordon SaksenaiD, Dimitri G. LivitziD, Daniel RosebrockiD, Chip StewartiD, Ignaty LeshchineriD, Alexander Baumann, Douglas Voet, Kristian CibulskisiD, Eric BanksiD, Anthony PhilippakisiD, Gad A. GetziD
Abstract
FireCloud, one of three NCI Cloud Pilots, is a collaborative genome analysis platform built on a cloud computing infrastructure. FireCloud aims to solve the many challenges presented by the increasingly large data sets and computing requirements employed in cancer research. However, cost uncertainty associated with cloud computings pay-as-you-go model is proving to be a barrier to adoption of cloud computing. In this paper we present guidelines for optimizing workflows to minimize cost and reduce latency. Our guidelines include: (i) dynamic disk sizing to efficiently utilize virtual disks; (ii) tuned provisioning of virtual machines (VMs) using a performance monitoring tool; (iii) taking advantage of steep price discounts of preemptible VMs; and (iv) utilizing the optimal parallelization of a tasks workload.
02 /Specification & Classification
Click any tag to browse matching papers- Citations
- 45
- Assay
- Research Focus
- Approach
Abstract summary
Claude Opus 5.5 · xhigh effort · 18 Oct 2026 · from the abstract · CC BY-NC-ND 4.0
Four levers for cheaper cloud genomics on FireCloud: size virtual disks dynamically, tune VM provisioning from monitoring data, buy preemptible machines, and parallelise tasks at the right width.
Cancer research is generating increasingly large data sets and computing demands, and the pay-as-you-go pricing of cloud platforms makes spending hard to predict, which the authors describe as a barrier to adoption. The paper responds with workflow-level configuration guidelines for FireCloud, one of three NCI Cloud Pilots.
Key results
ClaimedFour workflow-level choices are put forward as the main handles on cloud cost and latency.
The guidelines are: sizing virtual disks dynamically so capacity is not wasted, tuning virtual machine provisioning with a performance monitoring tool, exploiting the steep discounts on preemptible VMs, and choosing how widely to parallelise a task's workload. No cost or runtime measurements are reported. · Abstract
ClaimedCost uncertainty from the pay-as-you-go model is presented as a barrier to cloud adoption.
Stated as a barrier to adoption created by the pay-as-you-go model; no survey, usage statistics or spending data accompany the claim. · Abstract
Methods, data and reproducibility
The paper describes FireCloud, a collaborative genome analysis platform running on cloud infrastructure, and sets out guidelines the authors recommend for configuring workflows on it: dynamic virtual-disk sizing, VM provisioning informed by a performance monitoring tool, use of discounted preemptible VMs, and appropriate parallelisation of a task's workload.
- Methods & tools
- FireCloud
Limitations (2)
- The guidelines come with no cost or runtime figures, so the saving attributable to each of the four strategies is not quantified anywhere in the abstract. · noted in this summary
- The advice is framed around FireCloud and the pricing of its underlying cloud provider; whether it carries over to other platforms is not addressed. · noted in this summary
Who should read:Read this if you run genomic workflows on pay-as-you-go cloud infrastructure and want a short checklist of configuration choices that drive the bill and the turnaround time.
Reproducibility ledger
Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.
Sign in to file a reproducibility report.
Related papers
- Accumulating computational resource usage of genomic data analysis workflow to optimize cloud computing instance selection
Tazro Ohta, Tomoya Tanjo, Osamu Ogasawara·GigaSciencePeer-reviewed·Apr 1, 2019
- A cloud-based workflow to quantify transcript-expression levels in public cancer compendia
PJ Tatlow, Stephen R Piccolo·bioRxiv·Jul 12, 2016
- Benchmarking undedicated cloud computing providers for analysis of genomic datasets.
Seyhan Yazar, George EC Gooden, … Alex Hewitt·bioRxiv·Aug 7, 2014
- Hot-starting software containers for bioinformatics analyses
Pai Zhang, Ling‐Hong Hung, … Ka Yee Yeung·GigaSciencePeer-reviewed·Jul 31, 2018
- Optimizing genomics pipeline execution with integer linear programming
Olesya Melnichenko, Venkat S. Malladi·bioRxiv·Feb 6, 2024
- Aether: Leveraging Linear Programming For Optimal Cloud Computing In Genomics
Jacob M. Luber, Braden T. Tierney, … Aleksandar David Kostic·BioinformaticsPeer-reviewed·Dec 7, 2017