Cost-Performance Optimization for Long-Term Sensor Data Retention in Intercloud Object Storage

Abstract

The exponential growth of sensor deployments in smart cities, healthcare, and environmental monitoring requires scalable and cost-efficient long-term data storage. While cloud-based storage offers high availability and elasticity, dependence on a single provider introduces risks such as service outages, vendor lock-in, and data loss. Existing fault-tolerant models like DepSky are not optimized for high-frequency, real-time sensor data or long-term cost efficiency. This paper presents an enhanced intercloud storage framework that extends the DepSky-A and DepSky-CA models with streaming capabilities, erasure coding, compression, and client-side confidentiality mechanisms. The proposed architecture supports real-time ingestion, block-level verification, and Byzantine fault tolerance across untrusted cloud providers. We implemented and deployed the system across four major commercial cloud platforms and evaluated it using a simulated workload of 150TB/year of sensor data. Results demonstrate that the enhanced Streaming DepSky-CA model achieves a sustained throughput of ~1.45 GB/sec, over 99.95% availability, up to 20% cost savings compared to traditional replication, and an average 17% reduction in storage size due to compression. These findings position the proposed model as a practical and efficient solution for long-term, privacy-preserving, and resilient sensor data storage in intercloud environments.