Came across this blog post by Michael Stonebraker on ACM communications site. He is predicting that we are reaching an era where traditional relational DBMS systems will perish unless they adopt to the challenges ahead. He says, today:
- In data warehouse market, column store beats a row store (RDMBS) by a factor of 50 in typical query performance
- In OLTP market, in-memory databases beat row store systems by a factory of 50
and so on…
Few years back, there were not many applications that dealt with giga bytes of data and thousands of users. “Web search” was the only application that one can easily recall where probably RDBMS was not used (or didn’t fit the bill). And that was the period we saw dozens of new open source/commercial RDBMS implementations.
But today, we see lot more:
- applications dealing with giga bytes and peta bytes of data.
- web sites exploring horizontal database scaling options and column-store databases
- open source projects implementing column-store (and like) databases.
I am sure there are thousands of applications for which RDBMS is still the best choice, but there are sure signs that a big pie from RDBMS market will go away…
Blogosphere is amazing… couple of related links I found:
Relational Databases: Are they obsolete ?
RDBMS: going the way of the mainframe?
Pioneer calls RDBMS technology obsolete (another one by Stonebraker 2 years back)
Are RDBMS obsolete?