From that graph, disk latency is horrendous.
Hello Eric,
There is no time axis on that graph so cannot tell over what period the graph you posted shows but this doesn't look like one application having "always performans problem" but something gone seriously wrong and fast - either Guest-OS has crashed or is barely functional or your environment is experiencing a storage or networking issue. First step is to determine are all/subset of VMs having issues or is just one VM. If just one VM then check if the Guest-OS is still functional and utilisation metrics from inside the VM and outside (e.g. horrendous SCSI-level vmdk disk latency), if you don't think the cause is at this level (e.g. poorly configured applications in Guest-OS) then you need to go to lower levels e.g. are your NICs experiencing driver crash or flapping (/var/log/vobd.log, vmkernel.log and esxtop 'n' option for dropped packets etc.), is the DAVG to the backing device through the roof (esxtop 'u' option).
Bob