OnLine vs SE Benchmark Test Set Request

I know this is a tired subject, so I'll make it brief. We've got SE, we're looking at OnLine. We've got a demo copy of OnLine I'm looking for a set of tests to compare the performance of the two engines. Here's what I've got so far: SETUP: AT&T StarServer E - Single Processor 486/33 No-one on the system. Unload our 3 largest tables. Create a new DB in SE and in OnLine. Load the data in. TESTS (time each one, compare performance): Queries: Single Table, Single row returned (test search time) -constraint indexed (done X times so some time elapses) -constraint nonindexed Multi table, Single Row returned -constraints indexed, join keys indexed -constraints nonindexed, join keys indexed Single Table, One Column, All rows returned (test thoughput) Multi Table, Several Columns from each, All Rows returned Updates: Update all rows of an unindexed column Update all rows of a well indexed column Update all rows of a poorly indexed column (set sex ="M") Delete: Delete all rows from a table without indexes Delete all rows from a table with good indexes Delete all rows from a table with bad indexes Insert: Insert many rows from a table without indexes Insert many rows from a table with good indexes Insert many rows from a table with bad indexes At the same time as I'm running some of these tests, I'll be running sar to see what this is costing the system. Is the CPU being hogged? Is the file system? How's the system acting under heavy Online Load? What I'm expecting to see is good performance on the single row queries since this is right up Informix's OLTP marketing pitch. I'm not expecting as good a performance on the reports (multitable/multirow) since they've never crowed about it. If it delivers substantially better (4-6X) performance in this one area, I probably won't go on. We're a read intensive site. We'll never have enough throughput (or budget) to make me happy. For the cost in both dollars and DBA maintenance, OnLine had better offer something more than 2X throughput. If anyone out there as gone through this before and can save me the trouble and tell me what I want to hear, do so. If there are any tests I'm overlooking, or if they are fundamentally flawed, I want to hear. If anyone's interested, I'll send along my findings. John White (white%elsouth@concert.net)