To: sadsack who wrote (324 ) 10/5/1999 1:46:00 PM From: treetopflier Read Replies (1) | Respond to of 717
20 fraudulent transactions out of a billion. Let's stick to the claim and examine what this would really take. First, as you correctly state, the criteria for determining a fraudulent transaction would have to be codified in the form of one or more SQL queries. Most likely multiple queries to suite the various transaction types. These each of those billion transactions would have to pass through MSTR software to generate an appropriate query and return a result set to the MSTR software. Since your forte is enterprise software, why don't you share your MSTR based design, including Microsoft NT hardware/software and MSTR products for performing 1B+ transactions in a timely manner against a dataset of approximately 500GB+ in size, while the OLTP engines are ALSO performing those 1B transactions. Soon as you articulate the answer to this one and supply a reference site or two in production, you can raise your rates. I am of the opinion after years of working with datasets and transaction volumes of the sizes Saylor claims he can handle that the most efficient way to detect a fraudulent transaction is not via an external database tool. Each OLTP transaction, in concert with internal database software (e.g. PL/SQL triggers, procedures) is the only viable way to perform transaction handling (e.g. fraud detection) at the volume levels Saylor claims to be able to handle. You don't look at transactions after they have happened, e.g. Business Intelligence, you examine them while they are occurring by building business applications with appropriate safeguards. Much of this BI bullshit is just marketing jargon created to sell superfluous products like MSTRs when the basic componentry is already owned by the customer. ttf