Optimized Delay of File Accessing using Raptor Code System in Cloud Computing

Avinash Sharma, Mithilesh SagarInternational Journal of Trend in Scientific Research and Development 2020

Abstract

Cloud storage systems rely on fountain codes to ensure data availability across distributed nodes, but decoding overhead often leads to significant access delays. This paper introduces the Raptor Code System (RCCS), which integrates Raptor codes with Genetic Algorithm (GA) optimization and a multi-stage retrieval strategy. Experimental results using CloudSim demonstrate that RCCS reduces file access delays by up to 23% compared to traditional Luby Transform (LT) code systems, offering a more efficient solution for high-performance cloud storage.

Introduction

As cloud computing scales, the latency associated with retrieving fragmented data becomes a critical bottleneck. Traditional erasure coding methods, such as Luby Transform (LT) codes, often suffer from high computational overhead during the reconstruction phase. This "decoding delay" occurs when the system must wait for a sufficient number of fragments to arrive or struggles with inefficiently encoded data. The proposed Raptor Code System (RCCS) aims to minimize these delays by optimizing the mathematical structure of data fragments before storage and employing a sophisticated multi-stage request model.

Methodology

The RCCS framework introduces three primary innovations to reduce retrieval latency:

1. Raptor Codes vs. LT Codes

Unlike standard LT codes, Raptor codes incorporate a pre-coding step. This allows for successful file recovery even if the fountain code process is incomplete, significantly reducing the computational burden on the CPU during decoding.

2. Genetic Algorithm (GA) Optimization

A novel aspect of RCCS is the use of a Genetic Algorithm to optimize Raptor code fragments prior to cloud storage. By applying selection, crossover, and mutation operators, the system ensures that fragments are stored in their most efficient form for rapid reconstruction.

3. Multi-Stage Retrieval Modeling

To prevent network congestion and minimize tail latency, the authors developed a mathematical model for a two-stage request process. Instead of a single large-scale request, the system initially requests a batch of $n_1$ fragments. A second batch ($n_2$) is only requested if decoding fails, balancing reliability with network efficiency.

Comparison of one-stage and two-stage request processes in the RCCS framework.

Key Insights

A significant finding of this research is the linear approximation of the delay equation. For systems with a large number of source symbols, the complex delay calculation can be simplified to a linear approximation: $\tilde{D}_1 = \frac{n l_{RC}}{r}$. This simplification allows cloud providers to predict performance and manage resources in real-time without intensive computation. Furthermore, the GA optimization demonstrates that improving the mathematical structure of data at the storage level can be as effective as optimizing network paths.

Mathematical derivation of the delay equation and its linear approximation.

Results and Discussion

The RCCS was evaluated against the Luby Transform Code System (LTCS) using CloudSim 4.0.1. Key performance metrics include:

  • One-Stage Delay Reduction: RCCS achieved a 23% reduction in access delay. For a 100-packet workload, delay decreased from 1.75s (LTCS) to 1.31s (RCCS).
  • Two-Stage Efficiency: In multi-stage scenarios, RCCS maintained a 4% speed advantage over LTCS.
  • Resource Utilization: The GA-optimized fragments reduced storage redundancy while maintaining high data protection levels. CPU decoding time was also significantly lower than standard implementations.

Conclusion

The Raptor Code System (RCCS) provides a robust framework for reducing file access delays in cloud environments. By combining efficient coding, genetic optimization, and strategic retrieval models, it outperforms traditional LT-based systems. Future work will focus on analyzing the energy consumption of GA optimization and adapting the system for Platform as a Service (PaaS) and Software as a Service (SaaS) environments.

Summary updated

BibTeX
@article{sharma2020optimized,
  author = {Avinash Sharma and Mithilesh Sagar},
  title = {{Optimized Delay of File Accessing using Raptor Code System in Cloud Computing}},
  journal = {International Journal of Trend in Scientific Research and Development},
  year = {2020},
}
Figure