Data Warehousing: A Technical Overview
P> Here are the nitty gritty innards of these monster systems. Start, for example, with the Pacificare Health System. Pacificare is a large database customer with 1200 VUPS and 1.5 terabytes of storage operating under various UNIX clusters.
According to Pacificare, this ocean of data serves a network of 2,500 users and 3,000 subscribers. Pacificare is mostly an all-Digital shop, the crux of the system contains two Digital Alpha server 8400s. These have eight gigabytes of RAM memory in each, and eight CPUs in each machine. Digital's 64-bit Alpha processing chip technology pushes speedy throughput. Key products and applications are: two VAX 7730s, 2 VAX 7640s, 4 VAX 7760s, 4 Alpha 7760s, 2 Alpha 8400s -- all in cluster. The operating system is open VMS. The Digital networking products are DECNet, Ethernet and Pathworks series. Digital peripherals include StorageWorks and MTI Storage. Digital software includes DECForms, DEC RDB for Open VMS, ACMS Series, DEC COBOL and All-in-1 series.
On the applications front, they serve HMO Information Systems Accounting, and the Digital services are taken care of by PC Utility, Break/fix and SI consulting.
That's the basic anatomy of one monster. There are, of course, others. But as you'll read later, it's significant that all this super high-tech hardware and software is being devoted to the marshaling and manipulating of medical records. And as with Digital's increasingly popular Internet search engine Alta Vista, the 64-bit processing takes search and retrieval into jet speed over the Pacific where others are still putting atop mere lakes of data.
Take a recent deal wherein Software AG Singapore won a huge data warehousing contract with Scotts Holdings Ltd., a $100 million multinational with interests in Indonesia, Thailand, the United Kingdom, India, Australia and Indochina. Scotts Holding's multifaceted bailiwick concentrates on property development, hospitality, center management and property investment.
Now, apply the data warehousing technology of Software AG Singapore to those multifold data sources and the true power of data warehousing becomes obvious. As the Scotts Holding model illustrates, data from diverse industries can be cross-pollinated to conceive valuable new databases and lists. For example, a list showing renters moving out, their income, and whether they would be amenable to certain approaches in certain areas from certain real estate agents -- and have they had kids recently, should we send them that catalog? In other words, through planning, capital and shrewd marketing, accurate data can create hybrid data that companies and governments will pay for. Data warehousing technology, in this respect, allows clever companies and governments to turn to what was a buzzing power boat (old databases) into huge aircraft carriers loaded with marketing and other firepower. Those companies with the capital to convert from power boats to aircraft carriers will find themselves masters of the international race to rule and exploit the ever-filling data oceans.
Like aircraft carriers, data warehouse technology has its own lexicon. Binary large objects (large pictures) are known as BLOBS; data warehousing can also store compressed video, sounds and other multimedia. They're working on smells. Perhaps next time, students will bury their time capsule inside some descendant of the Digital Alpha server 8400s. The sights, sounds, smells -- and don't forget text -- of a generation reduced and retrieved by 1s and 0s.
The demand for data warehousing has engendered a promiscuous stew called RAID, which is a peculiar acronym for redundant array of inexpensive disks. This RAID leads a handful of more promising methods of storing and accessing oceans of data. The squeaky but traditional technology such as sequential tape is on the wane. The once "gee whiz" optical disk storage is now being put in its place by RAID. Don't look for a shake-out to come any time soon. There's too much data in the old formats, and who knows what the next latest format for storage will be -- one that will make RAID look like the 8-track tape of data storage. Still, it may be well into the next century before inventory managers start shipping that old sequential tape to recycling centers. The essential fact to keep in mind is that the these new data warehouses do not have walls: You yourself will die, but the minutiae of your life will live on in 1s and 0s for as long as the laws of electricity hold.