Towards a Big Data system disaster recovery in a Private Cloud

Disaster recovery (DR) plays a vital role in restoring the organization's data in the case of emergency and hazardous accidents. While many papers in security focus on privacy and security technologies, few address the DR process, particularly for a Big Data system. However, all these studies that have investigated DR methods belong to the “single-basket” approach, which means there is only one destination from which to secure the restored data, and mostly use only one type of technology implementation. We propose a “multi-purpose” approach, which allows data to be restored to multiple sites with multiple methods to ensure the organization recovers a very high percentage of data close to 100%, with all sites in London, Southampton and Leeds data recovered. The traditional TCP/IP baseline, snapshot and replication are used with their system design and development explained. We compare performance between different approaches and multi-purpose approach stands out in the event of emergency. Data at all sites in London, Southampton and Leeds can be restored and updated simultaneously. Results show that optimize command can recover 1 TB of data within 650 s and command for three sites can recover 1 TB of data within 1360 s. All data backup and recovery has failure rate of 1.6% and below. All the data centers should adopt multi-purpose approaches to ensure all the data in the Big Data system can be recovered and retrieved without experiencing a prolong downtime and complex recovery processes. We make recommendations for adopting “multi-purpose” approach for data centers, and demonstrate that 100% of data is fully recovered with low execution time at all sites during a hazardous event as described in the paper.

http://dx.doi.org/10.1016/j.adhoc.2015.07.012