This paper presents a source-side backup scheme with low-resource usage through collaborative deduplication and approximated lazy deletion when frequent virtual machine snapshot backup is required in a large-scale cloud cluster. The key ideas are to orchestrate multiround duplicate detection batches among machines in a partitioned asynchronous manner and remove most unreferenced content chunks with approximated snapshot deletion. This paper discusses the challenges, main design and strategies, and evaluation results.
Tuesday June 21, 2016 2:25pm - 2:50pm MDT
Denver Marriott City Center1701 California Street, Denver, CO 80202