Researchers Unveil New Data Replication Method for Disaster-resilient Information Platforms
Japanese researchers have developed a way to protect data and enable information services to continue working in disaster-affected areas, even when there is a network disruption over a wide area preventing access to remotely located back-ups. In recent years, disaster recovery services are available to provide service continuation even in the event of a wide-area disasters, such as those that back-up data to remote locations via the Internet.
The Research Institute of Electrical Communication at Tohoku University (RIEC), Hitachi, and Hitachi Solutions East Japan, developed a new technique, "cost/risk-aware data replication." This method achieves a balance between the cost of additional storage and data protection by reducing the number of replicas in servers which have a lower risk of data loss and increasing the number of replicas in servers with a higher risk. Further, a method was developed to solve the optimization problem of determining the optimal backup server each time the data is replicated.
The cost/risk-aware data replication technique illustrated in the above figure with 4 servers and an average of 1.5 replicates, selects appropriate servers which reduce the risk of data loss compared to the original server, for first replication. After first replication, it determines the data loss risk level of that server, and if the risk is still high, selects another server for second replication. As the average number of replications is fixed at 1.5 times, servers C and D which have a higher risk of data loss at 50%, are selected as the servers that require second replication.
The team conducted simulations on a configuration of 24 servers at various municipal locations at both the city and prefectural levels.
When the data was replicated between locations unlikely to be simultaneously affected during a disaster, researchers found that as much as 94% of the data could be protected even when half of the servers were damaged.
By adjusting the number of data replicas according to the risk of data loss, the same amount of data can be protected with half the number of replicas compared to conventional methods. This technology should enable disaster-resilient information platforms at an affordable cost.
The next step in this research will be to conduct verification tests using servers loaded with the cost/risk data replication technique placed in multiple locations throughout the three campuses of Tohoku University to create a large scale test environment. In the experiment, the data on the servers will be accessed by the electronic prescription notebook system developed with the Miyagi Prefecture Pharmaceutical Association.