On 8/26/05, Josh Berkus <josh@agliodbs.com> wrote:
> ## November 2005: The PostgreSQL Global Development Group announced today
> the release of PostgreSQL 8.1. The new version, containing many advanced
> database features as well as performance enhancements, makes PostgreSQL
> the reference platform for high-volume, high-peformance open source data
> centers.
"... continues PostgreSQL position as the reference platform..."
Never admit you weren't the best. It's silly, and it's all marketing
fluff, but it's about positioning the product as the best, and as
something that has been the best for a while.
> "This is our second major release in 2005," said _____, PostgreSQL major
> contributor, "It's very exciting. Thanks to all of the new programmers
> and companies contributing to the project, as well as automated testing,
> development is faster than every before."
"... development is faster than ever before, without sacrificing the
stability and quality-control that PostgreSQL is known for."
> Bitmapscan: Some indexes will be automatically converted to bitmaps in
> memory, giving up to 20x faster index performance on complex queries
> against very large tables. Bitmapscan also greatly reduces the need for
> multi-column indexes.
"... against very large tables, such as those found in data marts and
data warehousing applications."
> Shared Row Locking: we have continued to improve on the "better than
> row-level locking" which Multi-version Concurrency Control (MVCC) provides
> by supporting shared locks for relational references.
Can we give an example of how this increases performance, or reduces
deadlock opportunities? I'm not sure how to explain this to the
"average DBA," who sometimes misses the point of MVCC.
> Table Partitioning: in version 8.1, the query planner's ability to select
> the correct table partitions for each query (called Constraint Exclusion)
> is expanded, making PostgreSQL's table partitioning useful to a broader
> range of applications.
I'd move this up near bitmap indexing, and play the two together, as
they're really big deals for data mart and data warehousing
applications. Somehow tying the two together would be useful, I
think.
> "Blah new features blah blah," remarked some high-profile postgresql
> contributor. "Blah coolest yet blah blah blah."
Needs more foo, less bar.
Great work, as always.
Chris
--
| Christopher Petrilli
| petrilli@gmail.com
