Academic Systems-Oriented Database Research

1
Academic Systems-Oriented Database Research

• Betty Salzberg
• Northeastern University
• HPTS 1999 presentation

2
A report from the October 1998 NSF workshop on
industrial/academic cooperation in database
systems

• Speakers were from industry
• Audience was from academia
• Web site http//www.ccs.neu.edu/groups/IEEE/ind-a
  cad

3
System-Oriented Database Research

• What mistakes are common on academic papers?
• What are good research problems?
• What is the value-added of a Ph.D.?
• How can industry help get the students and the
  research it would like?

4
What mistakes are common on academic papers?

• Wrong Assumptions Example Jeff Vitters 1985
  TODS paper where he assumed you could add bits to
  a block on a WORM disk. But 1000 bytes is the
  smallest writable unit with a 300-byte checksum
  burned into the disk.
• No Consideration of Performance Costs Example
  Optimistic concurrency. An abort is much more
  expensive than a wait if there has been any
  writing. In the no-conflict case, the expense of
  keeping track of and comparing lists of reads and
  writes at commit is worse than the expense of a
  lock table.

5
What are some good research problems? (excerpt
from workshop)

• Schema integration
• version management
• comparative studies
• zero administration
• tool for compensation action creation
• data cleansing
• cache consistency in three tier systems

• Distributed application management
• performance modeling
• tools for back-tracing
• support disconnected clients
• tools to enable exploitation of new hardware and
  software

6
What is the value-added of a Ph.D.?

• Older and more mature
• Can read technical papers critically
• Can write technical papers
• Can give technical talks
• Has completed a long project (Persistent!)
• Has been vetted (passed qualifying exam)
• More likely than bachelors student to understand
  basics
• May have done a lot of programming for the Ph.D.
  project
• Does the Ph.D. thesis subject matter?

7
If you want more relevant research papers and new
employees with system sense, please help us.
8
How can industry help get the students and the
research it would like? (suggestions from
workshop)

• Collaborate
• Ph.D. student internships
• Visits ---(both ways)
• sabbaticals
• summers
• short visits/talks
• write papers for major conferences
• send contributions to IEEE Data Engineering
  Bulletin or SIGMOD Record

• Let an academic interview you for an article
• serve on program committees
• review papers for journals
• write textbooks or help write them
• take an academic job (wouldnt it be nice if
  academics were industry professionals as well?)