Here's a small set of changes to get riak-like tests running in wtperf. There are a couple minor things:
1. Right now populate is blazingly fast compared to riak because we only doing inserts instead of empty read followed by an update, and also because populate is sequential, and riak is not. It seems the random populate does not guarantee a full keyspace. We may want more of a random shuffle instead. For instance, the fruit config finished populating 25M items in 160 seconds rather than 45 minutes.
2. When running tests 2, 3, 4 after running test 1, I append to the existing test.stat. keithbostic I do not do the same for the latency log file, 'monitor' because having separate latency charts for each seems best. For now, the script I run just moves the 'monitor' file aside after the test completes.
3. It isn't clear we really need the random_value config, but given the bug found and different behaviors of random data versus all the same data, it seemed reasonable.