ISQL Performace Optimization

I have a rather general question about query optimization and performance benchmarking with Informix ISQL/4GL 4.1 running with the online engine. In Paul Mahler's book, INFORMIX 4GL, he briefly mentions an indexing scheme called multilevel indexing where one can use indexes on indexes to gain a little on performance with respect to lookup. He frames his discussion around B-Trees which is straightforward on the theory, yet vague on how one can implement this in ISQL. I am sitting here with the reference manual and see no way syntactically to do this. Yet, Mahler's reference almost implies that this is implicit in building any indexing scheme with informix. Is this possible in ISQL? Secondly, my development team has been spending a good deal of effort lately looking for bottlenecks in our code that, when optimized, can yield a substantial speedup in time. Thus far we've been able to squeeze about 12% out of our payroll accounting package but are looking for more. We've gone through and looked for poorly formed joins where the filter was performed after the join itself; or reworked poorly designed queries to avoid multiltable joins that were only necessary a small amount of the time where independent table selects were applicable; and of course, a great deal of work has gone into rethinking our indexing strategy. We've used the SET EXPLAIN option to study the query plan for most of the processor intensive joins butr have found it a bit vague in determining from where some extra time savings could be found. Can someone point me in the direction of some literature discussing Query optimization / database optimization and tuning with respect to informix products? Or, perhaps provide some useful feedback on some known methods of optimal indexing strategies (multilevel indexing perhaps). ---------------------------------> Jeff Dirks Software Engineer/Developer Fourgen Software Technologies, Inc Seattle, WA uunet!4gen!jeffd OR jdirks@wolf.cs.washington.edu <--------------------------------