[Paper Review] Executing Bag of Distributed Tasks on the Cloud: Investigating the Trade-offs Between Performance and Cost

Long Thai, Blesson Varghese+1 more|arXiv (Cornell University)|Oct 30, 2014

Cloud Computing and Resource Management12 references4 citations

TL;DR

This paper proposes a heuristic algorithm to optimize the trade-off between performance and cost when executing Bag of Distributed Tasks (BoDT) on the cloud by decentralizing task execution near data sources. The method models data transfer and processing times, uses user-defined performance-cost preferences (via a beta parameter), and generates execution plans that achieve optimal results in 91% of cases, demonstrating significant cost-performance gains over centralized execution.

ABSTRACT

Bag of Distributed Tasks (BoDT) can benefit from decentralised execution on the Cloud. However, there is a trade-off between the performance that can be achieved by employing a large number of Cloud VMs for the tasks and the monetary constraints that are often placed by a user. The research reported in this paper is motivated towards investigating this trade-off so that an optimal plan for deploying BoDT applications on the cloud can be generated. A heuristic algorithm, which considers the user's preference of performance and cost is proposed and implemented. The feasibility of the algorithm is demonstrated by generating execution plans for a sample application. The key result is that the algorithm generates optimal execution plans for the application over 91\% of the time.

Motivation & Objective

• To investigate the trade-off between performance and monetary cost in executing Bag of Distributed Tasks (BoDT) on the cloud.
• To develop a heuristic algorithm that generates optimal execution plans based on user preferences for performance and cost.
• To compare centralized vs. decentralized execution strategies in terms of cost and performance.
• To model the impact of geographical data distribution and network transfer times on BoDT execution.
• To validate the feasibility and effectiveness of the proposed approach using a real-world application scenario.

Proposed method

• Models task execution using data transfer time $ trans_{t,c} = s_t \times trans_{l_t,c} $ and processing time $ comp_{t,c} = s_t \times comp $, where $ s_t $ is data size, $ trans_{l_t,c} $ is per-unit transfer time, and $ comp $ is processing speed.
• Defines total execution time per task as $ exec_{t,c} = trans_{t,c} + comp_{t,c} $, and computes total time across all tasks for each deployment configuration.
• Introduces a user preference parameter $ \beta \in [0,1] $ to balance performance (minimize time) and cost (minimize VM usage), with $ \beta = 0 $ favoring cost and $ \beta = 1 $ favoring performance.
• Applies a heuristic search to explore different VM deployment configurations across cloud regions, selecting the plan that best matches the user’s $ \beta $-driven preference.
• Uses a sample BoDT application (news aggregation) to evaluate the algorithm across varying $ \beta $ values and compare predicted vs. actual optimal plans.
• Compares centralized (single VM) and decentralized (multiple VMs near data sources) execution models to quantify performance and cost trade-offs.

Experimental results

Research questions

• RQ1 What is the optimal trade-off between execution performance and monetary cost when deploying BoDT applications on the cloud?
• RQ2 How does decentralizing task execution near data sources affect the performance and cost of BoDT workloads?
• RQ3 To what extent can a heuristic algorithm generate execution plans that align with user-defined performance-cost preferences?
• RQ4 How does the choice of deployment configuration (centralized vs. decentralized) impact the total execution time and cost?
• RQ5 What is the sensitivity of the optimal plan to changes in user preference (i.e., varying $ \beta $)?

Key findings

• The heuristic algorithm successfully generated the optimal execution plan for 10 out of 11 test cases, achieving a success rate of 91%.
• The algorithm effectively balances performance and cost by adjusting deployment configurations based on the user’s preference parameter $ \beta $, with performance and cost shifting predictably across different $ \beta $ values.
• Decentralized execution significantly reduces total execution time compared to centralized execution by minimizing data transfer latency through proximity to data sources.
• The model demonstrates that performance gains from decentralization are substantial, but come at a higher cost, highlighting the need for a preference-driven optimization approach.
• The framework clearly illustrates the performance-cost trade-off, showing that optimal configurations shift as user preferences change, with $ \beta $ serving as a key control parameter.
• The experimental results confirm that the proposed mathematical model and heuristic are feasible and effective for real-world BoDT workloads such as news aggregation systems.