BOINC@TACC

BOINC@TACC is a volunteer computing project operated by the Texas Advanced Computing Center (TACC) at the University of Texas at Austin that integrates BOINC-based volunteer computing with supercomputing and cloud computing. Rather than running scientific applications directly on TACC's supercomputers, the project routes qualified High-Throughput Computing (HTC) jobs from TACC systems such as Stampede2 to computing resources donated by volunteers, including desktops, laptops, tablets, and cloud-based virtual machines. Launched in 2017, BOINC@TACC was the first use of volunteer computing by a major high-performance computing center.

BOINC@TACC
BOINC@TACC logo
The University of Texas at Austin, home to the Texas Advanced Computing Center
Project
Status	Complete
Category	Multidisciplinary (aerospace engineering, computational biology, earthquake engineering, bioinformatics)
Compute	CPU & GPU
Requires	VirtualBox, Docker
Development
Developer	Texas Advanced Computing Center, University of Texas at Austin
Author	Ritu Arora
Sponsor	National Science Foundation (award #1664022)
Maintainer	BOINC@TACC project team, Texas Advanced Computing Center
Initial release	2017 (as TACC-2-BOINC); renamed BOINC@TACC on January 1, 2019  (7 years ago)
Completed	No
Discontinued	July 11, 2022  (4 years ago)
Repository	https://github.com/ritua2/BOINCatTACC
Software
Written in	PHP, Python, Shell, C++
Operating system	Windows, Linux, macOS (BOINC client); Docker-based for cloud volunteers
Metadata
Website	https://boinc.tacc.utexas.edu/
License	LGPL-3.0

Overview

BOINC@TACC allows researchers with TACC or XSEDE allocations to run small, loosely coupled high-throughput jobs on volunteered hardware instead of spending their supercomputer allocation time on them. The project's stated goal is to free up TACC's supercomputers for tightly coupled, compute- and memory-intensive parallel jobs by offloading independent, "embarrassingly parallel" work to a BOINC server that, in turn, distributes it to volunteer clients. The project's lead is Ritu Arora, a research scientist at TACC. BOINC@TACC has supported research spanning aerospace engineering, computational biology, and earthquake engineering, among other domains.

Volunteers can contribute either by running the standard BOINC client (which requires VirtualBox for the project's virtual-machine-based jobs) or, for institutional donors with cloud virtual machines, an alternative Docker-based mechanism developed specifically for the project. Task availability on BOINC@TACC has historically been intermittent, since work is generated only as TACC researchers submit qualifying jobs rather than on a continuous schedule; during the COVID-19 pandemic, for example, volunteers reported a batch of virtual-machine-based tasks related to COVID-19 research becoming briefly available.

History

After several live iterations, the project began operating in 2017 under the name TACC-2-BOINC. To get started, the TACC team deployed an instance of a BOINC server on a cloud computing platform and then developed the software needed to integrate BOINC with TACC's supercomputing and cloud computing platforms. On Jan 1, 2019, the project was renamed BOINC@TACC and moved to its current project URL. The project is funded primarily through NSF award #1664022, with additional support from XSEDE, TACC, and the Science Gateway Community Institute. Contributors to the BOINC@TACC codebase have included Carlos Redondo, an aerospace engineering student and developer on the project, along with Thomas Johnson and Anubhaw Nand, with additional input from BOINC founder David Anderson.

Software infrastructure

The BOINC@TACC codebase, hosted on GitHub under the LGPL-3.0 license, builds on the boinc-server-docker software originally developed by Marius Millea and extends it with several project-specific components.

Job routing and application deployment

A command-line back-end, stampede2-backend, runs on Stampede2 and other TACC SLURM-scheduled Linux systems to determine whether a submitted job is eligible to run on volunteered BOINC clients rather than on TACC resources. Researchers can also submit and track jobs through a web-based user interface built for the project. Applications are packaged as Docker images through a component called MIDAS, which automatically generates Docker containers from source code so they can run either inside VirtualBox on traditional BOINC clients or directly inside Docker on volunteered cloud virtual machines through a mechanism called VolCon. Pre-built Docker images maintained by the project team have included AutoDock Vina, BEDTools, BLAST, Bowtie, GROMACS, HTSeq, LAMMPS, NAMD, and OpenSees.

Reef cloud storage

Result files produced by BOINC@TACC jobs are stored using Reef, a lightweight, reusable cloud storage component built for the project (also described in the literature under the name Greyfish). Reef can be deployed either as a single storage node ("pocket Reef") or, for larger workloads, as a distributed system with a manager node and one or more remote storage nodes.

Credit system

Like other BOINC projects, BOINC@TACC computes each volunteer's Recent Average Credit (RAC) as an exponentially weighted moving average that gives more weight to recently granted credit and decays older contributions with a half-life of one week. Following BOINC's general credit-accounting scheme, whenever new credit $c$ is granted at a time interval $\mathrm{\Delta }t$ after the previous update, the RAC is recalculated as

${\text{RAC}}_{\text{new}}={\text{RAC}}_{\text{old}}\cdot e^{-\mathrm{\Delta }t\ln 2/T}+\frac{c}{T}(1-e^{-\mathrm{\Delta }t\ln 2/T})$

where $T$ is the credit half-life (seven days). As with several smaller BOINC projects, BOINC@TACC has not published its statistics through third-party aggregators such as BOINCstats or Free-DC, so detailed current figures for active users, hosts, and average performance are not publicly available.

Volunteering

Volunteers can join BOINC@TACC either by attaching the standard BOINC client to the project (requiring VirtualBox for virtual-machine-based tasks), or, for donors offering cloud-based virtual machines, by using the project's alternative VolCon software. Researchers eligible to submit jobs include TACC users, researchers from any of the 14 University of Texas System institutions, and, historically, researchers accessing TACC resources through XSEDE.

A BOINC Manager client showing tasks attached from multiple BOINC projects; volunteers use the same client software to contribute to BOINC@TACC

Publications

The following papers, describing the BOINC@TACC project and its software infrastructure, are listed on the official BOINC project publications page.

Papers describing the project

1. Redondo, Carlos and Ritu Arora. Greyfish: An Out-of-the-Box, Reusable, Portable Cloud Storage Service. PEARC '19: Practice and Experience in Advanced Research Computing (2019). DOI: 10.1145/3332186.3333055.
2. Arora, Ritu, Carlos Redondo and Gerald Joshua. Scalable Software Infrastructure for Integrating Supercomputing with Volunteer Computing and Cloud Computing. Software Challenges to Exascale Computing (2019).
3. Wilson, Lucas A and David P Anderson. VAST: Integrating Volunteer Computing into Existing Cyberinfrastructure. (2017). DOI: 10.13140/RG.2.2.28139.64805.

See also

• BOINC
• Texas Advanced Computing Center
• Volunteer computing
• World Community Grid