Wal-Mart and its "data warehouse"

greenspun.com : LUSENET : TimeBomb 2000 (Y2000) : One Thread

I saw an interesting article today on how Wal-Wart makes astute use of Information Technology to make all kinds of decisions. Then I saw this quote about their "data warehouse" at company headquarters:

"It is larger than any other commercial data warehouse, and perhaps second only to the Pentagon's, according to industry experts."


Can you say complexity? Also, Wal-Mart is one of 28 Fortune 500 companies that will enter a new fiscal year on February 1, 1999. Let's hope Wal-Mart will be compliant by the beginning of 1999. Some companies that start their fiscal year on February 1st:


-- Kevin (mixesmusic@worldnet.att.net), December 14, 1998


I like my profession, I really do. But if they don't quit making up more stupid buzzwords every day I'm gonna quit and take up tending bar. Data Warehouse - as if "really big database" wasn't good enough.

-- Paul Davis (davisp1953@yahoo.com), December 14, 1998.

Wal-Mart does indeed have the largest, outside of the Government (NSA). They use Teradata HW and RDBMS for massive parallel processing and used it to competitive advantage because they were able to capture so much data and 'mine' it. One example I remember is that they found out that when male parents went to buy diapers they also bought beer. No, I'm not kidding. So, they put them in close proximity in their stores to sell more. Regarding the 28 Fortune 500 companies, we'll see.

-- Rob Michaels (sonofdust@net.com), December 14, 1998.

Makes sense........why would anyone want to change a diaper when they're sober.

-- Craig (craig@ccinet.ab.ca), December 15, 1998.

I wouldn't worry about the data warehouse too much. The way it works is this: You have one big database which does all your transactions, realtime. This database is optimized to do all the updates really fast. But your manager types also want to run massive queries, eg How many people who bought diapers bought beer on the same trip? This takes a whole different architecture to do it efficiently, and if you do them both on the same database then both applications will be very slow. So every night you copy all the data from the day's transactions into the data warehouse, and you can run these queries without affecting operations.

So if the data warehouse goes down, it just means you can't figure out that the beer and diapers should go together. But you can still run your store.

All this is not to say that Walmart's operational database won't have problem, or that they'll be able to deliver their stock to their stores, etc. etc.

-- Shimrod (shimrod@lycosmail.com), December 15, 1998.

FWIW, many (not all) databases can be categorized as either On-line Transaction Processing (OLTP) or Decision Support Systems (DSS). An airline reservation system or bank database typically run OLTP real-time transaction 7x24, and also may have a separate DSS database with a feeder interface for management and customer reporting. Wal-mart is using their database as DSS, but may also have some OLTP applications (inventory?) that interface or feed it. It is more common to see a mix of feeds also... some daily, some weekly, some monthly...etc.

-- Rob Michaels (sonofdust@net.com), December 15, 1998.

Moderation questions? read the FAQ