Log Buffer #11: A Carnival of the Vanities for DBAs
Welcome to this week's Log Buffer, where we'll take a look at some of the hot topics and blog postings over the last few days.
As it's Friday and we're all starting to wind down for the weekend, we'll start off with Tom Kyte on The Daily WTF who has come up with an example of the ultimate extensible database - as long as your users never want to report off of it. On a similar topic, Jeff Smith's SQL Server Blog goes through his "Top 10 Things I Hate About SQL Server", with #2 being "Datatype conversions ...What is the deal with data types? I just use VARCHAR for everything, since that's the only way to avoid stupid error messages when I add data to my tables. Works great, until I want to get the month of a date or add two numbers -- stupid SQL Server makes me convert() the value first. How dumb is that? SQL doesn't know how to add "12" and "5" -- it thinks it's "125" !!!! To make things even worse, SQL Server gives me errors half the time saying it CAN'T convert! Uh, you are a computer; you should be able to figure this stuff out."
Jonathan Lewis warns us about downloading and running scripts you find on the internet, whilst Jeff Moss writes about helping the Oracle CBO by adding constraints, and then realises that it's actually all in Jonathan's CBO book. Jeff's presented a couple of good papers at the UKOUG on data warehouse tuning and partitioning and compression which are well worth reading if you've got a spare half an hour.
Jeff Hunter
actually judging database costs on annual TCA per user amusing,
thinking it's better to perform real-world tests using real
applications rather than rely on artificial benchmarks, whilst Pete
Finnigan celebrates two
years of Bogging (is that anything like dogging, Pete?).
Guy Bowerman from Informix talks about Finite Field arithmetic and Communitive Division Rings, and subsequently risks a breach in the space-time continuum, whilst Curt Monash asks whether data warehousing is all now about sequential access (perhaps we should ask Fabian?).
Daniel Schneller presents a preview of the new mySQL Index Analyzer, whilst Eddie Awads comes up with a way of making NULL=NULL (and does cause a breach in the space-time continuum).
"Does anyone still do logical data models" asks Ronald (depends if your client is paying fixed price or time and materials, in my experience...) and thinks about the difference between hobbyists and professionals. Duncan Lamb points to an article on Oracle Rootkits 2.0 (available shortly with Oracle 11g?) whilst the Oracle WTF guys post instructions on how to capture spiders humanely using a one of Tom Kyte's books - the net surely closes on Mr. Burns; how long can this man stay free?
David Wheeler talks about Batch Updates with PL/pgSQL (zzzzzzzzzzzz....) whilst (Groundskeeper?) Willie talks about the secrets of a successful upgrade to DB2 for z/OS Version 8. Time for some data warehousing articles I think...
... Andy Hayler, one of the founders of Kalido and a writer on data warehousing and enterprise software, takes a look at data warehousing architectures over time and concludes that whilst separate data marts and techniques such as EII have their place, if you want proper reporting against multiple transaction systems you really need a proper data warehouse, and an approach to deal with the issue of master data management. Although Andy typically pushes the idea of packaged data warehousing, he makes some good observations in his blog about the enterprise software market and in particular the efforts of vendors such as Oracle and Microsoft to eat into the market previously held by the specialist vendors. Check out "Data Warehousing Dead" by William McKnight and "Logical Data Modelling: A Key to a Successful Data Warehouse Implementation" whilst you're there.
Lucas Jellema from AMIS
tuning
SQL statements using execution plans, Neshan Bardolliwalla
talks about SAP's
new in-memory BI accelerator that requires no tuning or need
to build aggregates (as my son says, "yeah...right."), whilst David Walker's Data Management
Site has published and overview
architecture for Enterprise Data Warehouses.
Gabriel Fuchs talks about Business Intelligence & Different Cultures, and concludes that in a BI&W project, cultural considerations may equal technical ones in terms of importance. Finishing off on data warehousing, I've just had an article published on Oracle Technology Network on Data Profiling and Automatic Cleansing using Oracle Warehouse Builder 10gR2, and I've also come up with some interesting issues around Warehouse Builder's handling of slowly changing dimensions that you might find useful - if you can make it through to the end of what's a fairly lengthy investigation.
That's it for this week, thanks to the team over at Pythian for inviting me to do this Log Buffer, and have a great weekend.