Evaluating the Impact of Algorithmic Components on Task Graph Scheduling

To Appear at JSSPP 2025 - The 28th Workshop on Job Scheduling Strategies for Parallel Processing

Abstract

Scheduling distributed applications modeled as directed, acyclic task graphs to run on heterogeneous compute networks is a fundamental (NP-Hard) problem in distributed computing for which many heuristic algorithms have been proposed over the past decades. Many of these algorithms fall under the list-scheduling paradigm, whereby the algorithm first computes priorities for the tasks and then schedules them greedily to the compute node that minimizes some cost function. Thus, many algorithms differ from each other only in a few key components (e.g., the way they prioritize tasks, their cost functions, where the algorithms consider inserting tasks into a partially complete schedule, etc.). We propose a generalized list-scheduling algorithm that allows mixing and matching different task prioritization and greedy node selection schemes to produce 72 unique algorithms. We benchmark these algorithms on 20 datasets to study the individual and combined effects of different algorithmic components on performance and runtime. We show that while every algorithmic component we consider is well-established in the literature, many of their combinations which have never before been considered are pareto-optimal with respect to average performance and runtime. We also report interesting interactions between algorithmic components and dataset features, suggesting that performance is highly application-dependent.

Authors

Jared Coleman
Ravi Vivek Agrawal
Ebrahim Hirani
Bhaskar Krishnamachari

Evaluating Scheduling Algorithms for Adaptive Orchestration in Federated Tactical Edge Cloud Environments

To Appear at ICMCIS 2025 - The International Conference on Military Communication and Information Systems

Abstract

The IST-193 Research Task Group (RTG) has focused on adapting the popular Kubernetes cloud native platform to operate in military tactical network environments consisting of transient, mobile nodes and intermittent and limited network connectivity. A key focus area for the group has been designing the Federated Adaptive Orchestrator, which is responsible for deciding where services should be run among the tactical cloud environments that are available. This paper focuses on the evaluation of a variety of scheduling algorithms in terms of their performance during the orchestration process and compares them to the performance of a brute-force algorithm that tries to find the most optimal solution. This paper was originally presented at the NATO Science and Technology Organization Symposium (ICMCIS) organized by the Information Systems Technology (IST) Scientific and Technical Committee, IST-209-RSY - the ICMCIS, held in Oeiras, Portugal, 13-14 May 2025.

Authors

Alessandro Amato
Harrie Bastiaansen
Jared Coleman
Willem Datema
Mattia Fogli
Johan van der Geest
Bhaskar Krishnamachari
Thomas Kudla
Pablo Sanchez
Niranjan Suri

Multimodal Search on a Line

To Appear at SIROCCO 2025 - 32nd International Colloquium on Structural Information and Communication Complexity

Abstract

Inspired by the diverse set of technologies used in underground object detection and imaging, we introduce a novel multimodal linear search problem whereby a single searcher starts at the origin and must find a target that can only be detected when the searcher moves through its location using the correct of $p$ possible search modes. The target's location, its distance $d$ from the origin, and the correct search mode are all initially unknown to the searcher. We prove tight upper and lower bounds on the competitive ratio for this problem. Specifically, we show that when $p$ is odd, the optimal competitive ratio is given by $2p+3+\sqrt{8(p+1)}$, whereas when $p$ is even, the optimal competitive ratio is given by $c$: the unique solution to $(c-1{)}^4-4p(c+1{)}^2(c-p-1)=0$ in the interval $[2p+1+\sqrt{8p},\mathrm{\infty })$. This solution $c$ has the explicit bounds $2p+3+\sqrt{8(p-1)}\le c\le 2p+3+\sqrt{8p}$. The optimal algorithms we propose require the searcher to move infinitesimal distances and change directions infinitely many times within finite intervals. To better suit practical applications, we also propose an approximation algorithm with a competitive ratio of $c+\epsilon$ (where $c$ is the optimal competitive ratio and $\epsilon >0$ is an arbitrarily small constant). This algorithm involves the searcher moving finite distances and changing directions a finite number of times within any finite interval.

Authors

Jared Coleman
Dmitry Ivanov
Evangelos Kranakis
Danny Krizanc
Oscar Morales-Ponce

PISA: An Adversarial Approach To Comparing Task Graph Scheduling Algorithms

To Appear at IPDPS 2025 - The 39th International Parallel and Distributed Processing Symposium

Abstract

Scheduling a task graph representing an application over a heterogeneous network of computers is a fundamental problem in distributed computing. It is known to be not only NP-hard but also not polynomial-time approximable within a constant factor. As a result, many heuristic algorithms have been proposed over the past few decades. Yet it remains largely unclear how these algorithms compare to each other in terms of the quality of schedules they produce. We identify gaps in the traditional benchmarking approach to comparing task scheduling algorithms and propose a simulated annealing-based adversarial analysis approach called PISA to address them. We also introduce SAGA, a new open-source library for comparing task scheduling algorithms. We use SAGA to benchmark 15 algorithms on 16 datasets and PISA to compare the algorithms in a pairwise manner. Algorithms that appear to perform similarly on benchmarking datasets are shown to perform very differently on adversarially chosen problem instances. Interestingly, the results indicate that this is true even when the adversarial search is constrained to selecting among well-structured, application-specific problem instances. We present the first known lower bounds on the performance of many of the algorithms considered in this paper compared to other popular scheduling algorithms. This work represents an important step towards a more general understanding of the performance boundaries between task scheduling algorithms on different families of problem instances.

Authors

Jared Coleman
Bhaskar Krishnamachari

LLM-Assisted Rule Based Machine Translation for Low/No-Resource Languages

AmericasNLP @ NAACL 2024 - 4th Workshop on NLP for Indigenous Languages of the Americas

Abstract

We propose a new paradigm for machine translation that is particularly useful for no-resource languages (those without any publicly available bilingual or monolingual corpora): LLM-RBMT (LLM-Assisted Rule Based Machine Translation). Using the LLM-RBMT paradigm, we design the first language education/revitalization-oriented machine translator for Owens Valley Paiute (OVP), a critically endangered Indigenous American language for which there is virtually no publicly available data. We present a detailed evaluation of the translator's components: a rule-based sentence builder, an OVP to English translator, and an English to OVP translator. We also discuss the potential of the paradigm, its limitations, and the many avenues for future research that it opens up.

Authors

Jared Coleman
Bhaskar Krishnamachari
Khalil Iskarous
Ruben Rosales

Linear Search for an Escaping Target with Unknown Speed

IWOCA 2024 - 35th International Workshop on Combinatorial Algorithms

Abstract

We consider linear search for an escaping target whose speed and initial position are unknown to the searcher. A searcher (an autonomous mobile agent) is initially placed at the origin of the real line and can move with maximum speed $1$ in either direction along the line. An oblivious mobile target that is moving away from the origin with an unknown constant speed $v<1$ is initially placed by an adversary on the infinite line at distance $d$ from the origin in an unknown direction. We consider two cases, depending on whether $d$ is known or unknown. The main contribution of this paper is to prove a new lower bound and give algorithms leading to new upper bounds for search in these settings. This results in an optimal (up to lower order terms in the exponent) competitive ratio in the case where $d$ is known and improved upper and lower bounds for the case where $d$ is unknown. Our results solve an open problem proposed in [Coleman et al., Proc. OPODIS 2022]

Authors

Jared Coleman
Dmitry Ivanov
Evangelos Kranakis
Danny Krizanc
Oscar Morales-Ponce

Online Allocation of Sensing and Computation in Large Graphs

IEEE CIC 2023 - International Conference on Collaboration and Internet Computing

Abstract

We consider vehicle tracking over a large territory equipped with sensors and computational units, and propose online resource allocation algorithms that decide which sensor nodes to activate, and on which computational units to perform the corresponding tracking tasks. We show through numerical evaluation that our approach can notably outperform state of art algorithms in terms of delay, communication cost, and the number of required sensor measurements.

Authors

Xinlin Li
Merve Karakas
Osama A Hanna
Mehrdad Kiamari
Jared Coleman
Christina Fragouli
Bhaskar Krishnamachari
Gunjan Verma

DARSAN: A Decentralized Review System Suitable for NFT Marketplaces

ICBC 2023 - International Conference on Blockchain

Abstract

We introduce DARSAN, a decentralized review system designed for Non-Fungible Token (NFT) marketplaces, to address the challenge of verifying the quality of highly resalable products with few verified buyers by incentivizing unbiased reviews. DARSAN works by iteratively selecting a group of reviewers (called "experts") who are likely to both accurately predict the objective popularity and assess some subjective quality of the assets uniquely associated with NFTs. The system consists of a two-phased review process: a "pre-listing" phase where only experts can review the product, and a "pre-sale" phase where any reviewer on the system can review the product. Upon completion of the sale, DARSAN distributes incentives to the participants and selects the next generation of experts based on the performance of both experts and non-expert reviewers. We evaluate DARSAN through simulation and show that, once bootstrapped with an initial set of appropriately chosen experts, DARSAN favors honest reviewers and improves the quality of the expert pool over time without any external intervention even in the presence of potentially malicious participants.

Authors

Sulyab Thottungal Valapu
Tamoghna Sarkar
Jared Coleman
Anusha Avyukt
Hugo Embrechts
Dimitri Torfs
Michele Minelli
Bhaskar Krishnamachari

Search and Rescue on the Line

SIROCCO 2023 - 30th International Colloquium on Structural Information and Communication Complexity

Abstract

We propose and study a problem inspired by a common task in disaster, military, and other emergency scenarios: search and rescue. Suppose an object (victim, message, target, etc.) is at some unknown location on a path. Given one or more mobile agents, also at initially arbitrary locations on the path, the goal is to find and deliver the object to a predefined destination in as little time as possible. We study the problem for the one- and two-agent cases and consider scenarios where the object and agents are arbitrarily (adversarially, even) placed along a path of either known (and finite) or unknown (and potentially infinite) length. We also consider scenarios where the destination is either at the endpoint or in the middle of the path. We provide both deterministic and randomized online algorithms for each of these scenarios and prove bounds on their (expected) competitive ratios.

Authors

Jared Coleman
Lorand Cheng
Bhaskar Krishnamachari

Graph Convolutional Network-based Scheduler for Distributing Computation in the Internet of Robotic Things

MILCOM 2023 WS-7 - Workshop On The Internet Of Things For Adversarial Environments

Abstract

Existing solutions for scheduling arbitrarily complex distributed applications on networks of computational nodes are insufficient for scenarios where the network topology is changing rapidly.New Internet of Things (IoT) domains like the Internet of Robotic Things (IoRT) and the Internet of Battlefield Things (IoBT) demand solutions that are robust and efficient in environments that experience constant and/or rapid change.In this paper, we demonstrate how recent advancements in machine learning (in particular, in graph convolutional neural networks) can be leveraged to solve the task scheduling problem with decent performance and in much less time than traditional algorithms.

Authors

Jared Coleman
Mehrdad Kiamari
Lillian Clark
Daniel D'Souza
Bhaskar Krishnamachari

Delivery to Safety with Two Cooperating Robots

SOFSEM 2023 - International Conference on Current Trends in Theory and Practice of Computer Science

Abstract

Two cooperating, autonomous mobile robots with arbitrary nonzero max speeds are placed at arbitrary initial positions in the plane. A remotely detonated bomb is discovered at some source location and must be moved to a safe distance away from its initial location as quickly as possible. In the Bomb Squad problem, the robots cooperate by communicating face-to-face in order to pick up the bomb from the source and carry it away to the boundary of a disk centered at the source in the shortest possible time. The goal is to specify trajectories which define the robots' paths from start to finish and their meeting points which enable face-to-face collaboration by exchanging information and passing the bomb from robot to robot.We design algorithms reflecting the robots' knowledge about orientation and each other's speed and location. In the offline case, we design an optimal algorithm. For the limited knowledge cases, we provide online algorithms which consider robots' level of agreement on orientation as per OneAxis and NoAxis models, and knowledge of the boundary as per Visible, Discoverable, and Invisible. In all cases, we provide upper and lower bounds for the competitive ratios of the online problems.

Authors

Jared Coleman
Evangelos Kranakis
Danny Krizanc
Oscar Morales-Ponce

The Snow Plow Problem: Perpetual Maintenance by Mobile Agents on the Line

ICDCN 2023 - International Conference on Distributed Computing and Networking

Abstract

We propose and study a new perpetual maintenance problem: The Snow Plow Problem. Inspired by snow removal in urban environments, we consider the perpetual maintenance by n mobile agents of a fixed-length interval over which some resource accumulates continuously over time (i.e. snow). In order to maintain the interval, agents must traverse it, collecting accumulated resources by plowing over continuous segments, and then return to some predefined destination point on the interval to dump collected resources. In a sense, our problem can be described as a combination of the well-studied patrolling and delivery problems. The maintenance cost for an agent is some combination of the distance traveled and the volume of resources collected between visits to the destination. We first study the case where the maintenance cost is simply the maximum amount of snow each agent must carry at any point in time and provide an optimal $O(n)$ algorithm for computing optimal trajectories for the mobile agents for scenarios where the destination is at an endpoint and snow falls uniformly across the interval. Then, we generalize the problem for any maintenance cost which can be expressed as the product of a positive, non-decreasing travel cost function $f(x)$ (the cost to travel to a distance $x$) and a positive resource cost function $g(x)$ such that ${\int }_y^{z}g(x)dx$ represents the volume of snowfall in one unit of time in the sub-interval $(y,z]\subseteq (0,1]$. We provide a $(1+ϵ)$-approximation algorithm that runs in $O(n\log 1/ϵ)$ time and an exact $O(n)$ algorithm for the case where $f(x)=ax$ and $g(x)=b$ for some positive constants $a$ and $b$. Finally, we generalize further to consider scenarios where the destination is at any point along the interval, providing another $(1+ϵ)$-approximation algorithm which runs in $O(n\log 1/ϵ)$ time.

Authors

Jared Coleman
Oscar Morales-Ponce

Line Search for an Oblivious Moving Target

OPODIS 2022 - International Conference on Principles of Distributed Systems

Abstract

Consider search on an infinite line involving an autonomous robot starting at the origin of the line and an oblivious moving target at initial distance $d\ge 1$ from it. The robot can change direction and move anywhere on the line with constant maximum speed 1 while the target is also moving on the line with constant speed $v>0$ but is unable to change its speed or direction. The goal is for the robot to catch up to the target in as little time as possible.The classic case where $v=0$ and the target's initial distance d is moving away from the robot with a known speed $v<1$. In this paper we design and analyze search algorithms for the remaining possible knowledge situations, namely, when $d$ and $v$ are known, when $v$ is known but $d$ is unknown, when $d$ is known but $v$ is unknown, and when both $v$ and $d$ are unknown. Furthermore, for each of these knowledge models we consider separately the case where the target is moving away from the origin and the case where it is moving toward the origin. We design algorithms and analyze competitive ratios for all eight cases above. The resulting competitive ratios are shown to be optimal when the target is moving towards the origin as well as when $v$ is known and the target is moving away from the origin.

Authors

Jared Coleman
Evangelos Kranakis
Danny Krizanc
Oscar Morales-Ponce

Multi-Objective Network Synthesis for Dispersed Computing in Tactical Environments

SPIE Defense + Commercial Sensing

Abstract

Tactical operations like search and rescue or surveillance necessitate the rapid synthesis of physically dispersed assets and mobile compute nodes into a network capable of efficient and reliable information gathering, dissemination, and processing. We formalize this network synthesis problem as selecting one among a set of potentially deployable networks which optimally supports the distributed execution of complex applications. We propose a general framework for studying this type of problem and provide solutions for some well-motivated variants. We discuss how the framework can be extended to support other objectives, parameters, and constraints as well as more scalable solution approaches.

Authors

Jared Coleman
Eugenio Grippo
Bhaskar Krishnamachari
Gunjan Verma

Network Synthesis for Tactical Environments: Scenario, Challenges, and Opportunities

SPIE Defense + Commercial Sensing

Abstract

We develop a network synthesis scenario, which is built around a concrete perimeter surveillance application, yet we believe captures a number of the challenges and requirements that are common to other tactical communication and computational network applications. The proposed scenario addresses the problem of binary population identification within a perimeter: our goal is to synthesize a sensing and computing network that classifies people moving within a given perimeter to one of two categories (e.g., friend or foe). We discuss several open challenges that we organize across the following clusters: sensor placement, communication network provisioning and optimization, computational task placement, dynamic resynthesis and resilience under adverserial settings. We also briefly discuss approaches that attempt to address such challenges.

Authors

Tzanis Anevlavis
Jonathan Bunton
Jared Coleman
Mine Dogan
Eugenio Grippo
Abel Souza
Christina Fragouli
Bhaskar Krishnamachari
Matthew Maness
Karl Olson
Prashant Shenoy
Paulo Tabuada
Gunjan Verma

Robotic Sorting on the Grid

ICDCN 2022 - 23rd International Conference on Distributed Computing and Networking

Abstract

Inspired by robotic applications, we study the problem of sorting a set of items over a physical domain with mobile agents. Given m mobile robots and a grid where each cell contains a single element, the objective is to design algorithms that allow robots to cooperatively sort the elements over the grid in the minimum time. We assume a synchronous model where robots do not communicate, can carry up to c elements, and can move between adjacent cells in one unit of time (grab and release time is negligible). First, we show that any algorithm requires at least $\mathrm{\Omega }\left(\frac{n^2}{mc}\right)$ units of time to sort an n-element line (an $n\times 1$) and present an algorithm that sorts the elements in $O\left(\frac{n^2}{mc}\right)$ time. Then, we show that any $n\times 1$-grid requires at least $\mathrm{\Omega }\left(\frac{n^3}{mc}\right)$ time and present an algorithm that completes in $O\left(\frac{n^3\log n}{mc}\right)$ time. Our algorithms have an equivalent competitive ratio to Shear Sort [Isaacd et al., Proc ICPP 1986] with only $m=n$ agents (compared to the $n^2$ processors required by Shear Sort). Finally, we present experimental results that show the capacity has very little impact on the overall runtime and that a simplification of the algorithm leads to better results.

Authors

Jared Coleman
Oscar Morales-Ponce

Message Delivery in the Plane by Robots with Different Speeds

SSS 2021 - 23rd International Symposium on Stabilization, Safety, and Security of Distributed Systems

Abstract

We study a fundamental cooperative message-delivery problem on the plane. Assume n robots which can move in any direction, are placed arbitrarily on the plane. Robots each have their own maximum speed and can communicate with each other face-to-face (i.e., when they are at the same location at the same time). There are also two designated points on the plane, $S$ (the source) and $D$ (the destination). The robots are required to transmit the message from the source to the destination as quickly as possible by face-to-face message passing. We consider both the offline setting where all information (the locations and maximum speeds of the robots) are known in advance and the online setting where each robot knows only its own position and speed along with the positions of $S$ and $D$.In the offline case, we discover an important connection between the problem for two-robot systems and the well-known Apollonius circle which we employ to design an optimal algorithm. We also propose a $\sqrt{2}$ approximation algorithm for systems with any number of robots. In the online setting, we provide an algorithm with competitive ratio $\frac{1}{7}(5+4\sqrt{2})$ for two-robot systems and show that the same algorithm has a competitive ratio less than 2 for systems with any number of robots. We also show these results are tight for the given algorithm. Finally, we give two lower bounds (employing different arguments) on the competitive ratio of any online algorithm, one of 1.0391 and the other of 1.0405.

Authors

Jared Coleman
Evangelos Kranakis
Oscar Morales-Ponce
Danny Krizanc

The Pony Express Communication Problem

IWOCA 2021 - 32nd International Workshop on Combinatorial Algorithms

Abstract

We introduce a new problem which we call the Pony Express problem. n robots with differing speeds are situated over some domain. A message is placed at some commonly known point. Robots can acquire the message either by visiting its initial position, or by encountering another robot that has already acquired it. The robots must collaborate to deliver the message to a given destination. The objective is to deliver the message in minimum time. In this paper we study the Pony Express problem on the line where n robots are arbitrarily deployed along a finite segment. The robots have different speeds and can move in both directions. We are interested in both offline centralized and online distributed algorithms. In the online case, we assume the robots have limited knowledge of the initial configuration. In particular, the robots do not know the initial positions and speeds of the other robots nor even their own position and speed. They do, however, know the direction on the line in which to find the message and have the ability to compare speeds when they meet. First, we study the Pony Express problem where the message is initially placed at one endpoint of a segment and must be delivered to the other endpoint. We provide an $O(n\log n)$ running time offline algorithm as well as an optimal online algorithm. Then we study the Half-Broadcast problem where the message is at the center and must be delivered to either one of the endpoints of the segment $O(n^2\log n)$ time and we provide an online algorithm that attains a competitive ratio of $\frac{3}{2}$ which we show is the best possible. Finally, we study the Broadcast problem where the message is at the center and must be delivered to both endpoints of the segment $\frac{9}{5}$, which we show is tight.

Authors

Jared Coleman
Evangelos Kranakis
Oscar Morales-Ponce
Danny Krizanc

Minimizing The Maximum Distance Traveled To Form Patterns With Systems of Mobile Robots

CCCG 2020, 32nd Canadian Conference on Computational Geometry, August 5-7, 2020

Abstract

In the pattern formation problem, robots in a system must self-coordinate to form a given pattern, regardless of translation, rotation, uniform-scaling, and/or reflection. In other words, a valid final configuration of the system is a formation that is similar to the desired pattern. While there has been no shortage of research in the pattern formation problem under a variety of assumptions, models, and contexts, we consider the additional constraint that the maximum distance traveled among all robots in the system is minimum. Existing work in pattern formation and closely related problems are typically application-specific or not concerned with optimality (but rather feasibility). We show the necessary conditions any optimal solution must satisfy and present a solution for systems of three robots. Our work also led to an interesting result that has applications beyond pattern formation. Namely, a metric for comparing two triangles where a distance of 0 indicates the triangles are similar, and 1 indicates they are fully dissimilar.

Authors

Jared Coleman
Evangelos Kranakis
Oscar Morales-Ponce
Jorge Urrutia
Birgit Vogtenhuber