In a 2008 study at the University of California, Santa Cruz funded by the National Science Foundation shows that more than 60% of data is inactive or static, but kept just in case it’s needed. Storing the inactive and active data in the same place is not cost-effective. Same problem also exist among cloud storage providers
- Dynamic file storage placement: To balance the performance with costs, there is need for cloud storage providers to adapt multi-tier storage systems and enable automatically moving inactive data and storing them in lower cost storage and retrieve it immediately. The data movement is accomplished in a manner that is transparent to the end users.
- File Access Pattern Recognition
- File Types: sequential access, random access
- Performance on-demand service: to enable users to reserve the storage and network I/O for specific files at a specific time in advance. This guarantees the performance of these files. Both users and providers benefit from this service. Users can pay and get better performance on demand without doing much effort. Providers can better schedule the resource based on the reservations.
- Storage system and network I/O scheduling
- Spot pricing, dynamic price over time
Existing Product and research:
Hitachi: Tiered Storage Manage: http://www.hds.com/products/storage-software/hitachi-tiered-storage-manager.html?WT.ac=us_mg_pro_tsm
Hystor: Making the Best Use of Solid State Drives in High Performance Storage Systems.
Energy Efficient Storage Management Cooperated with Large Data Intensive Applications
HybridStore: A Cost-Efficent, High performance Storage System Combining SSDs and HDDs
Cost Effective Storage using Extent based Dynamic Tiering