Pando: Personal Volunteer Computing in Browsers

Abstract

The large penetration and continued growth in ownership of personal electronic devices represents a freely available and largely untapped source of computing power. To leverage those, we present Pando, a new volunteer computing tool based on a declarative concurrent programming model and implemented using JavaScript, WebRTC, and WebSockets. This tool enables a dynamically varying number of failure-prone personal devices contributed by volunteers to parallelize the application of a function on a stream of values, by using the devices' browsers. We show that Pando can provide throughput improvements compared to a single personal device, on a variety of compute-bound applications including animation rendering and image processing. We also show the flexibility of our approach by deploying Pando on personal devices connected over a local network, on Grid5000, a French-wide computing grid in a virtual private network, and seven PlanetLab nodes distributed in a wide area network over Europe. CCS Concepts• Computing methodologies → Distributed computing methodologies; • Software and its engineering → Development frameworks and environments;

λHive

“…The main ones are MRSJ, Acio.js and Pando.All these are based on the Map-Reduce paradigm, which simplifies the distribution of computation, but limits the possible application cases, that is, they do not allow solutions of general use, but focus on solving high computing problems. (Dean & Ghemawat, 2004) (Ryza & Wall, 2010) (Constela, 2019) (Lavoie, Hendren, Desprez, & Miguel, 2019) Hive differs strongly from these alternatives since:…”

Genet: A Quickly Scalable Fat-Tree Overlay for Personal Volunteer Computing using WebRTC

“…We implemented a scalable version of Pando [29], a tool that leverages personal devices' browsers for executing computations in parallel, based on our JavaScript implementation [26] of the Genet fat-tree overlay. When a new browser window, executing on the device, successfully connects, it first joins as a leaf in the fat-tree and computes results, therefore acting as a processor.…”

“…Typical applications of volunteer computing have different data transfer patterns because each participant receives a different sub-set of data. BOINC submits the same computation to a small number of participants (at least three) until a majority agrees [41], while the current version of Pando [29] does not use redundancy because the code is executed on trusted devices. In addition, in both cases, each participant will return different results to the root.…”

“…To show the probabilistic balancing scheme is useful, we measured the depth of nodes and found that at least 83% of nodes have the same depth as they would have in a deterministic scheme; this percentage grows to 92.5% as the tree grows larger and is independent of the failure level of nodes if they reconnect through the root after a disconnection. To show the design can quickly scale, we measured the time required for all participants to become connected within a fat-tree overlay fully implemented and tested in Pando [29], a tool for personal volunteer computing [28] that targets shorter-running tasks than is typical for well-known and larger-scale volunteer computing projects [5]. We succeeded in connecting a thousand browser windows in 22-55 seconds on a local network and could fully deploy the Collatz application on 320 cores, reaching maximum throughput in less than 30 seconds.…”

Personal volunteer computing

“…We show the applicability of personal volunteer computing today by measuring the combined performance we obtained on two samples of personal devices: the laptops and smartphones we have accumulated over the years, and the smartphones of our colleagues at work. We use Pando [53,54], a tool we built for personal volunteer computing applications, to show that both samples of personal devices, in their aggregate computing power, are competitive with a top-of-the-line laptop from two years ago.…”