Cloud Data Services: SelfManageability and other Challenges

1
Cloud Data Services Self-Manageability and
other Challenges

• Surajit Chaudhuri
• Microsoft Research

2
Pillars of Self-Managing Technology for Databases
2
3
Self-Manageability is hard

• Contributing Factors
• Richness of SQL
• Strict separation between execution and
  optimization
• Expectations of backward compatibility make
  sweeping changes difficult
• Chaudhuri, Weikum VLDB 2000

4
Impact of Cloud Data Services

• Example Amazons SimpleDB, SQL Server Data
  Services
• Self-Manageability is a must!
• Opportunity to rethink
• Server API
• Data-driven 3-tier applications, Large scale data
  analysis (Map-Reduce/Parellel DBMS)
• Re-architecting of the backend
• Utilities
• A great opportunity to shake out the past
• Backward compatibility less of an issue

5
Cloud Data Services New Challenges

• Highly distributed
• Workload unpredictable
• Self-managing - No maintenance window
• Fine Grained Access Control, Data Privacy
• Effective use of resources
• Adaptive techniques required
• Impact of Multi-tenancy

6
Other Notable Areas

• Search
• Over structured data
• Exploiting structured data for text search
• Changes in Hardware
• SSD, Multi-core
• Mobile devices
• How do we address data management/sync
  challenges?