Hacker News new | past | comments | ask | show | jobs | submit login
If new DBMS technologies are so scalable, why is Oracle still on top of TPC-C? (dbmsmusings.blogspot.co.uk)
13 points by willvarfar on May 18, 2012 | hide | past | favorite | 9 comments



The actual title is >80 chars:

"If all these new DBMS technologies are so scalable, why are Oracle and DB2 still on top of TPC-C? A roadmap to end their dominance."

The "roadmap to end their dominance" is very interesting: they have a new database model and prototype called "Calvin" that scales well with distributed transactions...



Oracle and DB2 are at the top because no other solution is mature, it does not mean the ACID transaction approach is the best.

Can they make ACID transactions scale? They make bold claims and I'm afraid it will be hard to live up to it.

Is it worthwhile to make ACID transactions scale? Perhaps it's more efficient to switch to a new model for massively parallel processing?


For the kind of application that TPC-C models ACID is non-negotiable, plain and simple.

It's one thing for there to be minor blips in data integrity to happen in the server's handling Google's search results or a social network's user data. But in a system that processes financial transactions, a blip in data integrity is an accounting error. An accounting error that could potentially expose the operator of the database system to serious financial harm or legal liability.


You assume that a single blip in data integrity will cause functional errors.

I think you can design systems that can fail "in theory", but don't in practice.


That. Why do some programmers --even on HN-- think that storing BS data for a social network has the same needs as storing life-and-death financial, medical, judicial, millitary etc data?

As far as data stability and accuracy needs go, social networks are at the very bottom of the pyramid...


NoSQL is fool's gold


I'm convinced that "NoSQL" really means "We found out how flipping difficult it is to write a decent query optimizer, so we didn't write one. Your on your own."

Otherwise, I cannot think of a reason to eschew the Standard Query Language as an interface to one's database product.


I'm skeptical about the NoSQL hype, too, but I don't think one can really write the whole thing off as fool's gold. If only because of examples like Google's success with BigTable.




Consider applying for YC's Fall 2025 batch! Applications are open till Aug 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: