Software studio: first stress tests

Before recording the proof of concept that I published I did some stress tests on the Studio setup.

Test 1

4 inputs, 2 outputs + recording to disk:


-> Line-in music from my HiFi set through Garageband with 2 filters applied.

-> Quicktime Player playing a 60 minute music track.

-> iTunes playing songs continuously.

-> iChat session with continuous music coming from the other end.

<- iChat session (same as above) continuously streaming the final mix to the other end.

<- Output to Headphones.

- Recording straight to disc.


This test was run using my normal user account with several apps running in the background.

Result: the maximum recording time before it would stop was about 20 minutes. Usually it stopped after about 6-10 minutes. Under this load it seems to be impossible to keep recording for a long time.

(SoundTrack allows you to restart the recording quickly and all that was previously recorded is not lost.)



Test 2

3 inputs, 2 outputs + recording to disk:


-> Line-in music from my HiFi set through Garageband with 2 filters applied.

-> iTunes playing songs continuously.

-> iChat session with continuous music coming from the other end.

<- iChat session (same as above) continuously streaming the final mix to the other end.

<- Output to Headphones.

- Recording straight to disc.


After taking away the Quicktime player I could pretty much record for an hour straight. After an hour I decided to stop the test. Note that the computer wasn't touched during these tests. Rather unlike a real recording where there is a lot of user activity. However, in a real recording you also don't have 4 sources playing full-time.

During the tests I had the 'Activity Monitor' utility open and watched the graph of disk activity. Invariably, at the moment that the recording would get in trouble there was a lot of disk activity.

For the proof of concept recording I decided to create a new user account without any startup items, set the energy saving settings to maximum performance and unchecked the 'put hard disk to sleep when possible' option.

This worked without a glitch but it must be said that I didn't do anything other than record the show: No switching to the Finder, Browser or other applications than the ones used for the recording. Plus: the recording only took 5:21 minutes.

Conclusion: More tests are needed for more precise results. Stay tuned.

Some preliminary ideas: Writing to and reading from disk at the same time may be a problem. The disk itself may not be the only bottleneck here: if applications don't have enough time or don't get enough processing cycles to deal with their data in time, buffers may fill-up that in the end will have to be written to disk in a big chunk. This can cause the recording process to choke.