Query Performance

Typical business intelligence and big data anlaytics issues:

• Queries take too long to finish or don't finish at all
• Performance degrades as data increases and queries increase

Causes:

• Concurrency
  • Too many users hitting the system at once
  • Queries being run are too complex
• Data volume chokes system
  • Rich queries require computations that a single node vertically-scaled database cannot handle
  • Overloaded system causes errors to return an error to the user, or hours/days of waiting
  • Can store the data but can't analyze – data is outpacing the fixed amount of CPU, memory, and network resources
• Network becomes bottleneck
  • Network interconnect choked due to the sheer amount of data
  • Queries run inefficiently, requiring multiple passes over data or materialized views

Remedies:

• Parallel processing architecture
• Support high user concurrency
• Scale out using inexpensive commodity nodes
• Embedded Analytics
• SQL-MapReduce®
• Expressive analytic flexibility
• High-performance analytic processing where the data resides
• Reusability of SQL-MapReduce® components
• Hybrid Row/Column Database
• Unified SQL-MapReduce computation layer can access data optimized for performance through a row store, column store, or a combination of both
• Both row and column stores are first class citizens in the database architecture, sharing a unified set of data services for ease of management and seamless data access