Hello there,
Trying to wrap my head around the Vmware Balloon driver. I think I understand, but hoping someone here can confirm or set me in the right direction. We had a performance issue recently, that turned out to be a storage issue, but the memory usage on our hosts caught my eye.
We have a cluster with eight(8) ESXi 4.0 U2 Hosts - dual Xeon 6 core procs with 72GB memory each. 32 Win 2k3 Guests spread evenly between the hosts (4 per) each configured to use 20GB memory, but on average only use 15-17GB. No reservations on CPU or Memory set. Standard shares equal across all guests. Memory usage on the hosts are consistenly below the physical memory in the server.
Once powered on the guests consistenly have 15-17Gb memory allocated to them, even though on the lowest usage times Windows reports only using 2.5GB memory. During lowest usage times Vmware reports around 1 GB active memory, but 15-17GB is still allocated to each guest.
Now my understanding of the balloon driver is that until there is resource contention, it will not invoke, which is why we're seeing 15-17GB memory always allocated to the hosts and it never comes down - because we never run out of physical resources. If I run performace reports/graphs, the Memory Balloon always show 0 on all the guests.
Am I understanding the balloon driver properly? Given the configuration, are the memory allocations from the hosts acting properly. And could/should I be thinking about other ways to optimize performance within the environment?
Thanks in advance.
-Adam
