New Seminar Dates, New Articles and Odds & Ends

Mark Rittman

Nov 26, 2006 • 4 min read

A quick round-up of some odds and ends. First of all, a BIG thank you to Anders Holbøll who put together a script to gather all my previous blog articles from the Google cache, format them into an RSS feed compatible with Wordpress' import facility, then worked with me to use them to restore all the old blog postings for this site. Thanks to Anders, and Andy C before him, together I'd say we've recovered almost all that was lost and for this I'm eternally grateful. Thanks Anders and Andy C, much appreciated.

Last week involved trips to Dublin and then to Reading to run the BI Seminar, with this week's new material being a short section on Sunopsis Data Conductor and some new material on BI, BPEL and SOA. There were some familiar and friendly faces at both events, with Mick Lunan, an old colleague from the Plus Consultancy days, at the event in Reading. In common with the Dutch, Danish and London events, we had some pretty smart cookies in the audience and I picked up a few tips myself that I'll add into future events. Anyway, here a picture of some of the Reading audience on the first morning...

Next week I'm off to Croatia for Wednesday and Thursday, then I've got a break from travelling until after Christmas, when I'll be visiting Slovenia, Austria, Denmark (again) and Sweden - details on the new dates are on this page. I'm also doing a new seminar, on Data Warehousing using Oracle Database 10g, more details of which I'll post later this week.

Nick Goodman, now at Pentaho but previously a bit of an expert on Oracle Warehouse Builder, has released a set of OMBPlus utility scripts over at Sourceforge. OMBPlus is the scripting language for OWB and comes in useful when you want to automate the deployment of a project, promote a project from dev to test to prod, make mass-changes to objects in the repository and so on. Nick built them using OWB 10.1 and tested them on Paris Beta 3, so they should work on 10.2 but you'll need to test them out yourself first to be on the safe side. Thanks Nick.

Peter Scott posted a short article on his blog a week or so ago on whether, with technology such as EII and Oracle BI EE now available, data warehouses are still relevant or are in fact "dinosaurs". Pete, and David Aldridge in the comments afterwards, raised what I thought were a number of good arguments in favour of the traditional data warehouse approach. A couple of good points from Pete:

"OLTP sources such as ERP systems are optimised to give fast and robust access to single row data. The systems are not designed to trawl through masses of data
Retrieving large data sets and then aggregating them away in a user query tool can waste precious network bandwidth - is it not better to move and aggregate just once?
Many operational systems have only a sense of ‘now’ - the amount of history is limited
Furthermore, legislation (SOx, accounting rules, national security etc.) may mandate long-term storage away from the transactional systems"

whilst David made a couple of perceptive points in the comments:

"I’ll give special support to your third bullet point because the amount of analysis required to determine a historical point-in-time status for OLTP data is definitely non-trivial.
I’d add a fourth point — that different systems almost inevitably cannot be directly joined without some form of code translation or worse still data cleansing, and embedding such logic in a query tool ranges from the extremely difficult to the not possible.
There are many more of course — are we going to start running parallel queries during regular work hours against OLTP tables, with the resultant checkpointing every time a table is queried? Are we happy to share our sensitive OLTP system’s shared pool with non-sharable DW/DSS queries? Are we going to throw away partition pruning?"

I'll have to remember to ~~shamelessly plagiarise~~ quote these sagely words of advice when I come to do the DW seminars next January. Good points, especially the checkpointing caused by PQ and the difficulty of including code translations in query tool joins.

Syed Jaffar Hussain posted a good article a few days ago about the OS-level statistics that are available with Oracle Database 10g. Like most people, I use utilities such as vmstat, iostat and sar to assess the load on the server, wheras Syed points out that 10g has views such as DBA_HIST_OSSTAT which do the same thing. I'll make a point now of bookmarking Syed's blog. Also, in terms of new blogs, I noticed that the Hotsos education team have started a new blog, as has Doug Burns' breakaway collection of soft toys - go and read it before Doug gets sectioned.

Finally, I was browsing around the Oracle website and noticed a few new interesting articles. I've just skim-read them for now, but I'll make a note here as they look rather interesting:

Anyway, that's it for me now. Tomorrow I've got a rare day working from home in Brighton, then it's off to Zagreb. Hopefully I'll get a chance to do an update posting later in the week.

Sign up for more like this.