If 1. New or different files each run and 2. Neither is larger than mid GiB, sort'em both and sequential match merge. Put one or both as VIO.
If there are many larger files, or the smaller key file does not vary greatly from run to run, set it up as VSAM, DB2, Adabas, SAS or whatever and pass the other file, sorting before or after the processing pass into the final desired order. IMO, YMMV, see Payne's disclaimer. :) > -----Original Message----- > From: IBM Mainframe Discussion List [mailto:IBM-MAIN@LISTSERV.UA.EDU] > On Behalf Of David Speake > Sent: Thursday, October 04, 2012 4:48 PM > To: IBM-MAIN@LISTSERV.UA.EDU > Subject: Design of Key Matching application > > One of my application programmers asked for information on design > decision. > > Given two files that are not in order by the match key. > Assuming the smaller file has at most one record per possible key > value. > Larger file may have multiple records for some keys and may also have > values not found in the smaller file. > Assume optimal blocking, buffering, both files, both solutions. > Which is more efficient? > > Sort the smaller by that key, load to VSAM, then pass the unordered > larger file and do random retrievals from the VSAM file. > > Sort both PS files by key in question and pass sorted files for > matching by key. > > We both (sorta) think the answer is ... "It depends" but on what > criteria? > > Number of records in each file? Which is more important? And by how > much? > Record lengths? Same as above? > Length of key field in question? (Within SORT and VSAM length > restrictions of course). > Key bias in larger file? > Ratio of hits/non hits? > > Anyone have a nice formula? :-) > > David Speake > > ---------------------------------------------------------------------- > For IBM-MAIN subscribe / signoff / archive access instructions, send > email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
