DISEASE-TREATMENT RELATIONS USING MACHINE LEARNING APPROACH

  • Unique Paper ID: 146339
  • Volume: 4
  • Issue: 12
  • PageNo: 165-168
  • Abstract:
  • Based on the above analysis, it suffices to conclude that both N-way replication and erasure coding yield unbalanced tradeoff among read performance, write performance, and space efficiency. We argue and demonstrate through comprehensive experiments that, such imbalance causes two major problems.We design and implement disease, a hybrid storage scheme combining N-way replication and erasure coding, to provide high reliability, efficient I/O performance and flexible consistency simultaneously at low storage cost for object-based cloud storage systems.We propose MPL (Multi versioned Parity Logging) mechanism to facilitate efficient random write handling and read responding under both sequential and PRAM consistency.To verify the effectiveness of the proposed design, we evaluate disease on request-handling efficiency and the associated storage cost. We compare disease with 4-way CRAQ RS-code, two representative specific implementation of N-way replication and erasure coding. We set up wide range of system and workload configurations to investigate how the overall performance of a storage scheme is influenced when experiment parameters vary.We implement two consistency levels in the current disease system sequential consistency and PRAM (Pipeline RAM) consistency. These two consistency levels represent reverse consistency-performance tradeoff, and disease naturally supports both of them to meet different requirement in varied applications or services

Cite This Article

  • ISSN: 2349-6002
  • Volume: 4
  • Issue: 12
  • PageNo: 165-168

DISEASE-TREATMENT RELATIONS USING MACHINE LEARNING APPROACH

Related Articles