Background

In the era of precision medicine and big data, large population-cohort studies have become a main theme of epidemiology. Combining computer science and statistics with a given specialty, they uncover patterns and knowledge behind the data and provide new evidence for disease-prevention strategy. Peking University needed a distributed file system to unify storage and sharing of data and to efficiently support analysis, deep mining, and application of medical data.

Project Requirements

Large population cohorts are defined not only by sample size. Data quality is higher, coverage is fuller, supporting documents are richer, transfer and sharing are easier, and the bar for data standardization is higher. With large samples and many time points, such studies have unique strengths in etiology research and also raise the difficulty of data management and quality control. Peking University’s new distributed file system stores large-cohort research databases. Besides unifying storage and sharing, it must efficiently support analysis, deep mining, and application of medical data.

Solution

Sugon built the distributed file system for Peking University on ParaStor. Sugon ParaStor uses a 100 Gb/s EDR InfiniBand high-speed storage network to meet the platform’s need for high aggregate bandwidth on large medical files and high IOPS on small files. Data on the big-data platform grows extremely fast. ParaStor scales to the exabyte level, with linear growth in performance and capacity, meeting requirements for processing power and scalability.

ParaStor also uses multiple techniques to guarantee high reliability and high availability from the physical layer to the logical layer, protecting storage, management, application, and sharing of medical data.

ParaStor offers multi-dimensional resource management, multiple fault-warning mechanisms, and automatic detection and handling of dozens of sub-health states to keep the system stable. Visual operations and unified management of multiple clusters reduce operations pressure and cost.

Customer Benefits

With ParaStor, Peking University can use efficient cohort-data research to reveal disease causes, evaluate prevention, describe the natural history of disease, understand population health, and guide experimental design, while turning knowledge into early diagnosis and intervention for the clinic and the population—raising prevention and treatment and lowering the social health burden.