Iorate version 3.05 inclues the ability to skew the performance workload presented to the test devices. A skew percentage is used to control the level of skew - at 93%, then 93% of the I/O will happen against 7% of the capacity. The distribution is fixed for a given seed (to the random number generator), so that all tests with the same skew and seed will have matching I/O profiles. This allows for testing of automated tiering functions within the storage array under test. Note that while the distribution will be the same across the capacity of the device, the exact I/O profiles are random for each test run. Applying a shift allows for the simulation of the workload profile changing the skew focus areas partially over time.
Details on using the new parameters are included in the sample I/O profiles provided. The new options include:
- Area - the size of each portion of the target devices that will get the same level of workload (default is 4 MB)
- Correlate - a rough idea of how closely hot areas should be to hot and cold to cold (99% is almost always next to each other, 1% is highly random)
- Seed - the random number seed used in calculating the placement of the skew areas
- Shift - the shift applied to the base skew (shifts some capacity from hot to medium, medium to cold, and cold to hot)
- Skew - the skew percentage, expressed as the percentage of work done (against an area of 100% - skew)
There is no updated version of the User Guide yet for 3.0. That is in process. A general README is included to provide an overview of some of the scripts distributed with the program. The man page has had syntax updates made for the sun-LUN skew testing.
Version 3.05 adds the -u option to use direct I/O. It also has a fix to align the I/O created with the sub-LUN skew testing, which was causing I/O to fail on HP-UX systems.
Beta testing appears to have fixed the major challenges. Please report any bugs or issues with operations of the program. Thank you.