In the AS/400 system, how does data storage and retrieval operate?
The AS/400 system uses multiple methods to store and retrieve data. The most common method is through what we refer to as the "Physical" and "Logical" file system. OS/400 contains DB2/400, which is the database management system. This system builds an "Access Path," which is a road-map to each record stored on the system.
Generally speaking, when the file description is created (through DDS or SQL), the system will build an index to the records based on the physical location on disk for each record. When a program, query or SQL asks for a specific record, or set of records, the system will use the access path to retrieve those records. If the request is not using the key of the description, the system will search for a matching key access path. If it doesn't find one, it will create it in order to satisfy the request. In addition, hardware and software at the disk subsystem level handles RAID and mirroring processes to further reduce the retrieval time.
There are at least 12 other methods that are much more complex, and IBM won't tell us all of the details. The last known full-system testing that I am aware of is a TPC-C test performed in July of 1999 using the Baan method of testing. In that test, 12,000 users were simulated with a sub-second response time. The closest platform at that time in the Unix, Windows NT and other midrange market was 3,325 users with a 1.98 second response time, making DB2/400 a little over 2.5 times faster on scale.
To let you know how fast that would be: A single 12-way processor AS/400 could handle 720,000 passengers per hour without the "I'm waiting on the computer to update your ticket" comment coming out once. Almost every airline I know of uses AS/400s for exactly that purpose, in addition to cargo and baggage handling. Additionally, Wall-mart, K-Mart, Sears, Mastercard, Visa, American Express and Microsoft all use AS/400s for everything from point-of-sale to credit-card authorizations to accounting.