- Curt Monash has a nice article on his website with an overview of highly-available database technologies, and some simple classifications of things like alternate approaches to replication, etc.: Uninterrupted DBMS operation — an almost-achievable goal.
Buried in the article is a pointer to a timestamping MVCC data store that was new to me: Metamarkets's Druid
According to Monash, Metamarkets intend to open source Druid, but that hasn't happened yet.
- Two nice posts about setting up a partitioned database on Postgres:
Unrelated, but also nice, a post about how and why transaction ID freezing occurs in Postgres: Freezing Your Tuples Off, Part 1
A nice detailed review of "the original NoSQL paper": Scalable distributed data structures for internet service construction
And, of course, the Brewer/Hellerstein/Gribble/Culler paper itself: Scalable, Distributed Data Structures for Internet Service Construction
- Nick McKeown's keynote address from this year's SIGCOMM is online.
He spends a bunch of time talking about TCP performance behaviors.
- Two weeks ago was the XLDB (Extremely Large Databases) conference held at the Stanford Linear Accelerator Center, and most of the presenters's slide decks are available at the conference website.
Don't miss the always-fun, always-controversial presentation from Mike Stonebraker
And the Facebook slides are fun, too; I love it when the bullet point listing the data store which is 'tens of petabytes' isn't even the largest one on the list...
- Last week was the "2012 Storage Developer Conference". This show is an industry conference, not an academic conference, so it's quite a bit different. Still, it's fascinating to look through the program and the abstracts for the various talks. It's not clear if any of the presentation materials will eventually end up online; they don't seem to be freely available right now...
That's it for now, have fun!