Storage Issues Replica Placement Most existing works focus

Storage Issues

Replica Placement �Most existing works focus on how to place replica with low cost. �Maybe it is safer that we separate the replicas as far as possible? ◦ In same server => server crash ◦ In same rack = > rack failure ◦ In same datacenter = > earthquake or other cataclysms �Consider both distance and cost.

Data Deduplication �Data deduplication is a specialized data compression technique for eliminating coarse-grained redundant data. ◦ Improve storage utilization. �Issues: ◦ How to improve the duplication detection and chuck existence querying efficiency. �Efficient chunking, faster hash indexing, localitypreserving index catching, and efficient bloom filters …etc. ◦ Compressing the unique chunks and performing (fixed-size) large writes through containers or similar structures.

Read Performance of Deduplication Storage �Publication of David H. C. Du, HPCC’ 11. �Read performance is critical to reconstruct the original data stream.

Read Performance of Deduplication Storage(Cont. ) �One example is to store images of VMs(process/memory/disk) to shared network storage. ◦ VM images of idle desktops are migrated to network storage for energy saving.

Benchmarks �Filebench �http: //sourceforge. net/apps/mediawiki/filebench /index. php �Phoronix Test Suite – disk test suite �http: //www. phoronix-test-suite. com/ �Bonnie++ �http: //freecode. com/projects/bonnie