I am amazed at how many environments I see where expensive hardware resources are just plain wasted in the name of improving Vmware performance. The common areas of waste include:
-Storage. Often VMDK are allocated much bigger then they need be. This is especially painful to see on expensive fiber channel SAN storage at $30 to $50 per Gig!!
- Memory. The typical scenario is that when a server gets virtualized, admins allocate the the same amount of memory as it did when it was physical. This is a very common mistake especially if you are running similar workloads and are getting the benefit of memory reduction in ESX via TPS ( transparent page sharing)
- VMs that do nothing. They were deployed at one time and still occupy memory and storage but do nothing. Not a lot of talk about taking VMs down!
Now granted, Vmware perfomance is almost 100% dictated by available resource capacity in the core four: memory, cpu, network and storage, but blind overallocation of hardware resources is not going to help improve performance if for example you are having a disk I/O bottleneck.
Monday, November 10, 2008
Subscribe to:
Post Comments (Atom)
1 comments:
Reading this post prompted me to have a look at a clients site in regards to storage that was allocated but unutilised - I came up with this SearchMyVm query
vm.guestDisks.capacity>25Gb and vm.guestDisks.freeSpace>20Gb and vm.disks.file.datastore.freeSpace<15Gb
Try it, and see which VMs could be usefully shrunk to make more elbow room on your congested VMFS :)
Post a Comment