langreiter.com plain, simple: comments-2008-04-18

LANGREITER.COM plain, simple

START

INDEX

Freitag, 18. April 2008

Two interesting disk-y posts:

James Hamilton: "How to ensure that data written to disk, is REALLY on disk? Yeah, I know, this shouldn't be hard but the I/O stack is deep, everyone is looking for performance, everyone is caching along the way, so it's more interesting than you might like."

Steve Dekorte: ACID Databases: Fact and Fiction

"I learned this by writing my own small database (SkipDB) and doing it wrong the first time."

WAL

Addendum: Paul McJones kindly let me know about the amazing technique of "group commits", in which mutually independent transactions (think ATM networks) are grouped into single disk write windows; that way, it's indeed possible to support thousands of (disjoint!) transactions per second while adding only negligible latency to each individual transaction. Paul also recommended Jim Gray's and Andreas Reuter's Transaction Processing: Concepts and Techniques: '[...] a good reference for just about all the "tricks of the trade"'. — Thanks!

2008-04-18-disky

no comments

Please log in (you may want to register first) to post comments!