Data Stream Management Systems Checkpoint

1
Data Stream Management SystemsCheckpoint

• CS240B Notes
• by
• Carlo Zaniolo
• UCLA CSD

2
Research Challenges from DBMS to DSMS

• Relational Data Models
• Relational Data Streams same as append only
  relations
• But plays a major role and so are
  timestampstemporal DBs?
• Windows as time-varying table on a data stream.
• Query Languages and Operators
• Some RA operators are no longer simple main
  memory, windows, timestamps and latency, require
  different execution model.
• Blocking operators are not for continuous
  queries expressive power is further compromised
• Complex aggregates with windows and slides,
  including UDAs, needed

3
from DBMS to DSMS (cont.)

• Query Plans and Optimization
• Scheduling for optimal response time, memory not
  for total execution cost
• Quality of Services (QoS) Approximation
• Load shedding, sampling, and other synopsesnot
  an issue in DBMS
• Prototypes and Systems
• Mature technology versus very new one, but
• Many research prototypes, and four startups (at
  least)
• Extensions proposed by DBMS vendors.
• Advanced Applications
• XML and XQuery important in both
• Better support for Data Mining needed in both.