views:

181

answers:

5

I'm in graduate school and about to start a project related to databases. The professor has provided a list of topics for us to choose from, but we are also free to suggest something else we're personally interested in. The basic format of the project is quite simple:

  1. I'll study 2-3 research papers related to the area I end up choosing
  2. I'll prepare a presentation (where "presentation" is somewhat loosely defined)
  3. I'll prepare a 4-5 page summary of what I have learned

I really want to find a topic that is exciting and that would ideally make a good presentation (I would love some form of a "working demo"). So far my search has left me intrigued about the Stanford Entity Resolution Framework. I figured StackOverflow would be an excellent resource to tap for this. What are some exciting areas of database research that you know of? Thank you for taking the time to help me out.

+6  A: 

NoSQL is getting a lot of attention. You can't seem to go on proggit or hacker news without some mention of MongoDB, CouchDB or something of the like.

http://en.wikipedia.org/wiki/NoSQL

Corey Sunwold
Nice. One idea I'm just had is I could combine this with ORM impedance mismatch (which is quite old as a research area) and try to argue some sort of transition in the developer community.
Lyudmil
I am not familiar enough with either NoSQL (key value, document type databases) or ORM impedance mismatch, but it would seem at first glance that these types of databases might actually solve that problem? I am hoping someone with more experience them me can weigh in on this.
Corey Sunwold
@csunwold I read a little more. They would not, I was babbling.
Lyudmil
+3  A: 

I think the main aim is in scalability.

Look at these industry real-life databases:

That should give you good background for presentation.

Viliam
Is there any background on scaling issues with relational DBs? I read a great blog post recently arguing that they were not "truly" scalable, but it would be nice to have a more academia-friendly source.
Lyudmil
@Lyudmil: This might be of interest for you, nice reading: http://www.xaprb.com/blog/2008/07/23/using-base-instead-of-acid-for-scalability/
Viliam
@Viliam: The article it links to is fascinating and well written. Thank you.
Lyudmil
+2  A: 

Might I suggest a couple:

  • Formal object-oriented entities. Entities (tables) that can naturally extend other entities,
  • Hash O(1) indexing,
  • Single-level-store (SLS) tablespaces
  • Kernel-level integration (think AS/400)
Xepoch
Could you provide some links? Some of these are difficult to google for.
Lyudmil
Xepoch
+3  A: 

Semantics-based storage issues (like the one SERF addresses, but also all kinds of semantic web related storage possibilities) are very hot in academia right now (a bit less so in industry, as the returns don't really promise to come anytime soon;-).

Non-relational, highly scalable databases are extremely hot in industry research, but not so much in academia. So, if you're looking mostly for academic successes, peer-reviewed journal publication opportunities, future thesis work, etc, they may not be your best bet (they might be in some cases, esp. if you're just going for a MS, not a PhD, and care about superb internship opportunities in industry research and development, hot near-future employment opportunities in the field, etc, etc).

A great place to research "what's hot right" in a mix of academia (mostly) and industry R & D (a reasonable amount) are the appropriate ACM SIGs. I'm sure you know about SIGMOD, but SIGIR, SIGKDD, and (for specialized areas of storage) several others of the SIGs, can very well be interesting, too.

Alex Martelli
+1  A: 

Negative Databases

In a negative database, the negative image of a set of data records is represented rather than the records themselves.

hiena
Interesting reads. I wonder what my effective `U` universe of fixed-length records would be on 4X trading :)
Xepoch