We have a dedicated server running a number of smaller guest VMs containing for example, HTTP server, MySQL and some real-time applications.
However we find that JBD inside the guests are often spiking to 100% disk util approximately every 5-10 seconds. This causes the real-time applications to hang for ~500-1000ms when it attempts to perform disk accesses because it is too busy.
There are no significant amount of activity in log files and using noatime, nobarrier and we are unable to disable the journal. MySQL and the HTTP server are doing some IO activity, but not 100%.