But I don’t agree that the benefits of multi-core databases are limited to handling large volumes of data in in query-intensive apps. Nor is multi-threading necessarily the best answer for heavy/extreme OLTP workloads. Multi-core DBMS appear to enable higher transaction throughput in OLTP workloads (and better scalability). This is due in part to the elimination of multi-threading logic in the DBMS (latching, et al), which can compose ~14% of the overhead associated with the processing of a transaction.

Check out this paper from the 2008 SIGMOD conference: http://cs-www.cs.yale.edu/homes/dna/papers/oltpperf-sigmod08.pdf

It does a great job explaining the DBMS overhead associated with the execution of a transaction and also asserts that 90% of it is a by-product of legacy OLTP requirements and no longer necessary.

One of the co-authors of the paper, Mike Stonebraker, has co-founded a company developing a multi-core OLTP DBMS called VoltDB.

I posted a comment earlier asking if you had gotten permission to post the code you’d developed or could offer any insight to someone looking to do the same thing, but it appears my comment got deleted. Anyway, I am re-iterating my previous question about the status of the code.

Thanks!

Row-oriented storage engines have lots of issues at scale. What is needed is a column oriented storage engine which can process a single query in more than one thread.

But you don’t have to wait for the future. The The Kickfire Analytics Appliance features a column store and a unique patented SQL chip with which to process queries in parallel.

Good post.
Allow my to link to MySQL’s character sets and collations demystified, a post I wrote which explains how character sets and collations work in MySQL, as well as some best practices to use.

“I find it quite puzzling that in this day and age of globalization that many of the tools don’t default to UTF8.”
I couldn’t agree more. For some years now I’m thinking “it will all be over with withing a few years”, but so many websites are still working with locale specific charsets, I now feel les optimistic.

Regards

I did use an approach similar to Mike Hillyer’s methods. Basically, there is a side table that has a list of parents and children that link back to the node table, allowing me to have children with multiple parents as well as parents with multiple children, so we can group arbitrary sets of locations into “regions” or other types of groups.

Good on you for verifying obscurities! That’s hard to do, and it shows dedication that you’re intent on making sure the data is correct — very important for a DBA.

How did you deal with the schema for the tree heirarchy? Using Mike Hillyer’s method, as in here:

http://dev.mysql.com/tech-resources/articles/hierarchical-data.html

??