Friday, March 16, 2007

CMDB - Not a DB, not 'definitive master'

Ok - this was a real insight to me. Hope it's not Old News :-)
[25-Jan/15-Feb-2007. This PDF from the "CMDB Federation [BMC, CA, Fujitsu, HP, IBM and Microsoft] supports a similar view - "Federated CMDB's".]

Was talking to a friend today who does config, change & release mgt for a large, secure Government Department.
He's struggling with Data Quality issues - surprise, surprise.

They import equipment details from a logistics system, they do a *lot* of shipping to/from 100+ overseas offices.
It doesn't enforce strict product id's - there could be 20 different versions of a single PC product name.

But talking to him further, it's axiomatic that CMDB's will not and cannot ever be the master or definitive source of all information.

CMDB's are, by definition, a join of existing, disparate databases mastered by other products/applications.
The only thing they can be is an 'intelligent collective repository [just made that up].

It reminds me of comments by Jerry Gregoire

When he was CIO of Dell he said :
"you don't want to use an ERP" - takes away your distinctive business process, costs to get in and out of, and ties you into one monster database. And slows you down - you can't change your business quickly if the ERP vendor doesn't support your new function/process.

Dell used an object broker and connected into existing apps & databases.
They implemented a brand-new Inventory system with *zero* new databases. The place has gone downhill since he's left - I sometimes wonder if that's linked.

Jerry also pens pretty straightforward advice which I heartily agree with.


Neil Gunther said...

There's another wrinkle wrt Dell that you may be missing. It's no longer about whether they have an efficient DBMS to support their supply chain or not. The fact that they have ANY assembly line at all, has become their boat anchor. In particular, a major competitor no longer rolls their own PCs. Components are gathered (randomly?) via on-line auctions (huh?) and the offshore mfg lines (say, in China or Malaysia) are also selected dynamically according to quoted cost and time-to-market. This is like SOA taken to some kind of asymptotic limit. You might wonder if this produces a Forrest Gump "box of chocolates" PC. The answer is yes, more than likely. Standards just went out the window. There's no way to back-track where the components came from, etc. But, you protest "That's no way to run a railroad!" Isn't it? Because the cost of production is so low in this approach, it is more cost-effective for the mfr to simply replace any PC (possibly N times, I guess). So, where do IT service standards fit into that kettle of fish? :-)

SteveJ said...


Slam-dunk comments as usual. Thanks.

This is not a situation that the owner of any large fleet of PC's wants... If you buy 5,000 PC of the one config, you really want them identical.

Otherwise your SOE gets to be very hard. The whole point is to limit variability... Random variations with different bug sets are a nightmare. "Same Same" does not fly with users - but if you want cheap maybe you gotta be prepared to deal with the consequences.

Another of my speculations about Dell computers falling profits:
- laptops are becoming as popular as desktops. There are only a few laptop manufacturers - so Dell buys whole units, maybe at a price dictated by the builder, rather than a production line that Dell controls more.

I know that I want to see, touch, feel a laptop before buying it - whereas desktops and servers I'll buy from the spec sheets...