Volpex@Home: Difference between revisions

| developer            = [https://en.wikipedia.org/wiki/University_of_Houston University of Houston] Department of Computer Science; Parallel Software Technologies Laboratory (PSTL)

| sponsor              = [https://en.wikipedia.org/wiki/National_Science_Foundation National Science Foundation] (awards CNS-0834750 and MCB-0919974)

| programming language = C, [https://en.wikipedia.org/wiki/Python_(programming_language) Python], C++

'''Volpex@Home''' (originally '''Volpex@UH''') was a [https://en.wikipedia.org/wiki/Volunteer_computing volunteer computing] project run by the Department of Computer Science at the [https://en.wikipedia.org/wiki/University_of_Houston University of Houston] that used the [[BOINC]] middleware to execute communicating [https://en.wikipedia.org/wiki/Parallel_computing parallel] programs on ordinary, Internet-connected personal computers. The name Volpex stands for '''Parallel Execution on Volatile Nodes''', and the project's central goal was to make the idle time of everyday "volatile" desktop machines usable as a virtual [https://en.wikipedia.org/wiki/Computer_cluster cluster] for [https://en.wikipedia.org/wiki/Message_passing message passing] parallel applications that the standard BOINC model could not run.<ref name="nsf-volpex">{{cite web |title=Award Abstract #0834750: CSR-PSCE, SM: Collaborative Research: VOLPEX: A Framework for Parallel Execution on Volatile Nodes |publisher=National Science Foundation |url=https://www.nsf.gov/awardsearch/showAward?AWD_ID=0834750 |access-date=2026-08-30}}</ref><ref name="homepage">{{cite web |title=Volpex@Home (archived project homepage) |publisher=University of Houston |url=https://web.archive.org/web/20160120043345/http://volpexathome.cs.uh.edu/VolPEx/ |access-date=2026-08-30}}</ref>

The project also hosted a biological application called '''inCell@Home''', which ran [https://en.wikipedia.org/wiki/Parallel_tempering replica-exchange] [https://en.wikipedia.org/wiki/Molecular_dynamics molecular dynamics] simulations of protein behavior inside a crowded, cell-like environment in support of drug-design research.<ref name="homepage" /> The framework research was funded by two [https://en.wikipedia.org/wiki/National_Science_Foundation National Science Foundation] grants, and the project reached about 1,700 volunteers before it was completed and retired in 2017.<ref name="nsf-volpex" /><ref name="nsf-cheung">{{cite web |title=Award Abstract #0919974: Effects of macromolecular crowding on protein structure, folding, and interactions |publisher=National Science Foundation |url=https://www.nsf.gov/awardsearch/showAward?AWD_ID=0919974 |access-date=2026-08-30}}</ref><ref name="frwiki">{{cite web |title=Liste des projets BOINC (Volpex@UH statistics) |publisher=French Wikipedia |url=https://fr.wikipedia.org/wiki/Liste_des_projets_BOINC |access-date=2026-08-30}}</ref>

 

Most desktop computers sit idle for much of the day, representing a large pool of unused processing, storage, and communication capacity that grows with the spread of [https://en.wikipedia.org/wiki/Multi-core_processor multi-core] machines. The difficulty is that these machines are ''volatile'': their owners can switch them off, disconnect them, or reclaim them at any time and without warning.<ref name="about">{{cite web |title=About VOLPEX |publisher=University of Houston (archived) |url=https://web.archive.org/web/20150524061949/http://volpexathome.cs.uh.edu/VolPEx/volpexinfo.php |access-date=2026-08-30}}</ref> Platforms such as BOINC and [https://en.wikipedia.org/wiki/HTCondor Condor] handle this volatility well for independent, non-communicating tasks (the so-called [https://en.wikipedia.org/wiki/Embarrassingly_parallel "bag-of-tasks" or master-slave] model), but they do not natively support tightly coupled, message-passing parallel programs in which tasks must exchange data as they run.<ref name="nsf-volpex" /><ref name="faq">{{cite web |title=Volpex@Home Frequently Asked Questions |publisher=University of Houston (archived) |url=https://web.archive.org/web/20150524023014/http://volpexathome.cs.uh.edu/VolPEx/faq.php |access-date=2026-08-30}}</ref>

Volpex was designed to remove exactly that limitation, so that a broader class of scientific applications could run on volunteer hardware and many more research projects could benefit from free, donated compute time.<ref name="faq" /> Rather than relying only on [https://en.wikipedia.org/wiki/Application_checkpointing checkpointing], which the researchers argued is inadequate in environments with very high failure rates, Volpex used managed redundancy: it ran two or more replicas of every parallel process and regenerated failed replicas on demand from healthy ones, so that the application progressed at the rate of the fastest surviving replicas and continued seamlessly as long as at least one replica of each process remained alive.<ref name="nsf-volpex" /><ref name="about" />

The Volpex framework began as academic computer-science research. A collaborative [https://en.wikipedia.org/wiki/National_Science_Foundation National Science Foundation] grant, "VOLPEX: A Framework for Parallel Execution on Volatile Nodes" (award CNS-0834750), started on 1 September 2008 with Jaspal Subhlok as principal investigator and Rong Zheng and Edgar Gabriel as co-principal investigators at the University of Houston.<ref name="nsf-volpex" /> The associated MPI library, VolpexMPI, was first described in a 2009 paper by Troy LeBlanc, Rakhi Anand, Edgar Gabriel, and Jaspal Subhlok.<ref name="leblanc2009">{{cite journal |last1=LeBlanc |first1=Troy |last2=Anand |first2=Rakhi |last3=Gabriel |first3=Edgar |last4=Subhlok |first4=Jaspal |title=VolpexMPI: An MPI Library for Execution of Parallel Applications on Volatile Nodes |journal=Recent Advances in Parallel Virtual Machine and Message Passing Interface (EuroPVM/MPI 2009), Lecture Notes in Computer Science |volume=5759 |pages=124-133 |date=2009 |publisher=Springer |doi=10.1007/978-3-642-03770-2_19 |url=https://link.springer.com/chapter/10.1007/978-3-642-03770-2_19}}</ref>

'''VolpexMPI''' was a subset implementation of the [https://en.wikipedia.org/wiki/Message_Passing_Interface Message Passing Interface] (MPI) standard tailored to volatile nodes. Ordinary message-passing exchanges were converted into asynchronous Put/Get operations, and process replication provided robustness. The key fault-tolerance technique was fully distributed, sender-based message logging, which let a restarted or replacement replica reconstruct the messages it needed without a central checkpoint server.<ref name="leblanc2009" /> A later, portable variant called VolpexPyMPI was written in [https://en.wikipedia.org/wiki/Python_(programming_language) Python], ran on both Linux and Windows, and accepted user-level MPI programs written in C or [https://en.wikipedia.org/wiki/Fortran Fortran], trading some performance for ease of use and portability across heterogeneous hardware and operating systems.<ref name="leblanc2010ccgrid">{{cite conference |last1=LeBlanc |first1=Troy P. |last2=Subhlok |first2=Jaspal |last3=Gabriel |first3=Edgar |title=A High-Level Interpreted MPI Library for Parallel Computing in Volunteer Environments |book-title=2010 10th IEEE/ACM International Conference on Cluster, Cloud and Grid Computing (CCGrid) |date=2010 |pages=405-414 |doi=10.1109/CCGRID.2010.85 |url=https://ieeexplore.ieee.org/document/5493405}}</ref>

Because each process had several replicas, a receiving process had to decide which sender replica to contact first. The group developed communication target-selection algorithms for this; a hybrid strategy gave performance close to that of an all-fast-machine configuration, and evaluations used the [https://en.wikipedia.org/wiki/NAS_Parallel_Benchmarks NAS Parallel Benchmarks] across heterogeneous network and processor configurations.<ref name="anand2010">{{cite conference |last1=Anand |first1=Rakhi |last2=Gabriel |first2=Edgar |last3=Subhlok |first3=Jaspal |title=Communication Target Selection for Replicated MPI Processes |book-title=Recent Advances in the Message Passing Interface (EuroMPI 2010), Lecture Notes in Computer Science |volume=6305 |date=2010 |pages=181-190 |publisher=Springer |doi=10.1007/978-3-642-15646-5_21 |url=https://link.springer.com/chapter/10.1007/978-3-642-15646-5_21}}</ref>

[[File:BOINC project architecture.png|thumb|alt=A diagram of the BOINC client-server architecture showing a server feeding work to many volunteer clients over the Internet|The BOINC client-server architecture that Volpex extended with its Dataspace communication layer.<ref name="rohit2011" />]]

== Applications ==

This tied the computer-science framework directly to a biophysics question: experimental and computational work by the Cheung group showed that [https://en.wikipedia.org/wiki/Macromolecular_crowding macromolecular crowding] dramatically affects protein structure, folding, and activity, for example compacting the enzyme phosphoglycerate kinase and increasing its activity under crowded conditions that mimic the inside of a cell.<ref name="cheung-research">{{cite web |title=Cheung Group Research: Protein structure, function, and dynamics under cell-like conditions (archived) |publisher=University of Houston |url=https://web.archive.org/web/20120303170028/http://mynsm.uh.edu/groups/cheunggroup/wiki/f569a/Research.html |access-date=2026-08-30}}</ref> The crowding research combined coarse-grained and atomistic molecular simulation in a multi-scale approach and was supported by NSF award MCB-0919974.<ref name="nsf-cheung" />

 

Volpex remained a small research project throughout its life. In early January 2012 it counted 292 registered users (135 active), 824 hosts (404 active), and 64 teams across 36 countries, with about 1.1 million total credits, a recent average credit of roughly 3,449, and an average throughput of about 17.2 [https://en.wikipedia.org/wiki/FLOPS GFLOPS].<ref name="boincstats-early" /> By the time the project was retired, the French Wikipedia's BOINC project table recorded 1,746 users, 6,688 hosts, 252 teams, and 65,282,403 total credits as of 15 October 2017.<ref name="frwiki" />

* [https://en.wikipedia.org/wiki/Volunteer_computing Volunteer computing]

* [https://en.wikipedia.org/wiki/Distributed_computing Distributed computing]

* [https://en.wikipedia.org/wiki/Grid_computing Grid computing]

* [https://en.wikipedia.org/wiki/Desktop_grid Desktop grid]

* [https://en.wikipedia.org/wiki/Message_Passing_Interface Message Passing Interface]