It is complicated to troubleshoot a performance issue caused by a bursty workload. As most performance statistics are the calculated averages in a period of time, mostly 1 or 5 minutes in our existing performance monitoring and troubleshooting tools. These averages can not represent the real resource bottlenecks during the bursts, thus in some cases, a more granular data collection is needed to better understand the throughpout, CPU scheduling and CP behaviors.
sysstat is a great tool to collect some fundamental statistics at a much more granular level, seconds or even milliseconds.
sysstatcommand, especially those short bursts that only last for hundreds of milliseconds, or even just tens of milliseconds.