Mythbusters dutch vmug_2012


Published on

Some things never change. Or do they? VMware vSphere is getting new and improved features with every release. And these features change the characteristics and performance of the virtual machines. If you are not up to speed, you will probably manage your environment relying on old, no-longer-accurate information. The vMythbusters have collected a series of interesting hot topics that we have seen widely discussed in virtualization communities, on blogs and on Twitter. We’ve put these topics to the test in our lab to determine if they are a myth or truth.

  • Be the first to comment

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Mythbusters dutch vmug_2012

  2. 2. MYTHBUSTING  GOES  VIRTUAL  Eric  Sloof   Ma3as  Sundling  VMware  CerHfied  Instructor   Evangelist  NTPRO.NL   Dell  So@ware  @esloof   @msundling  
  3. 3. INTRODUCTION  •  VMware  vSphere  evolves  with  every  release  •  Things  that  used  to  be  true  arent  true  anymore  •  Engage  in  virtualizaHon  communiHes  and  social     media  to  get  up  to  speed  
  4. 4. AGENDA/MYTHS  1.  VMware  HA  works  out-­‐of-­‐the-­‐box  2.  VMware  snapshots  impacts  performance  3.  Disk  provisioning  type  doesn’t  affect  performance  4.  Always  use  VMware  tools  to  sync  the  Hme  in  your  VM    
  5. 5. MYTH  1  VMware  HA  works  out-­‐of-­‐the-­‐box  
  8. 8. HOST  FAILURES  A  CLUSTER  TOLERATES    ESX01 ESX02 ESX03 Shared storage – vm.vmdk
  9. 9. DEFAULT  MINIMUM  SLOT  SIZE   VM1 VM2 VM3 VM4 VM..n 32 MHz 69 MB•  If  you  have  not  specified  a  CPU   reservaHon  for  a  virtual  machine,  it   is  assigned  a  default  value  of   32MHz.      •  When  the  memory  reservaHon  is  0,   the  slot  size  equals  the  virtual   machine  overhead.  
  10. 10. SLOT  SIZE  BASED  ON  RESERVATION   VM1 VM2 VM3 VM4 VM…n512 MHz1093 MB •  vSphere  HA  calculates  the  CPU  and   memory  slot  size  by  obtaining  the   largest  CPU  and  memory   reservaHon  of  each  powered-­‐on   virtual  machine.  
  11. 11. HA  ADVANCED  SETTINGS  •  das.slotmeminmb Memory•  das.vmmemoryminmb reservation SLOT•  das.slotcpuinmhz  •  das.vmcpuminmhz   CPU reservation SLOT
  13. 13. VMS  REQUIRING  MULTIPLE  SLOTS   VM1 VM2 VM3 VM4 VM5 VM6 512 MHzSlot size 512 MB Reservation•  You can also determine the risk of resource fragmentation in your cluster by viewing the number of virtual machines that require multiple slots.•  VMs might require multiple slots if you have specified a fixed slot size or a maximum slot size using advanced options.
  14. 14. FRAGMENTED  FAILOVER  CAPACITY   ESX1 ESX2 ESX3 Shared storage – vm.vmdk
  15. 15. WORST  CASE  SCENARIO    ESX01 3.6 GHz ESX02 3.6 GHz ESX03 3.6 GHz 16 GB 16 GB 32 GB Shared storage – vm.vmdk
  16. 16. KEEP  HOSTS  THE  SAME  SIZE  Host memory: 3 * 16 GB Host memory: 2 * 16 GB 1 * 32 GB
  17. 17. PERCENTAGE  OF  CLUSTER  RESOURCES  RESERVED  ESX01 ESX02 ESX03 Shared storage – vm.vmdk
  19. 19. ADMISSION  CONTROL  BASED  ON  RESERVATIONS  •  vSphere HA uses the actual individual reservations of the virtual machines.•  The CPU component by summing the CPU reservations of the powered-on VMs.
  20. 20. COMPUTING  THE  CURRENT  FAILOVER  CAPACITY  •  If you have not specified a CPU reservation for a VM, it is assigned a default value of 32MHz
  21. 21. RESOURCES  RESERVED  IS  NOT  UTILIZATION  •  The Current CPU Failover Capacity is computed by subtracting the total CPU resource requirements from the total host CPU resources and dividing the result by the total host CPU resources.
  22. 22. PERCENTAGE  RESERVED  ADVANCED  SETTING  •  The default CPU reservation for a VM can be changed using the das.vmcpuminmhz advanced attribute•  das.vmmemoryminmb defines the default memory resource value assigned to a VM
  24. 24. SPECIFY  FAILOVER  HOSTS  ADMISSION  CONTROL  POLICY  ESX01 ESX02 ESX03 Shared storage – vm.vmdk
  25. 25. SPECIFY  FAILOVER  HOSTS  ADMISSION  CONTROL  POLICY  •  Configure vSphere HA to designate specific hosts as the failover hosts
  26. 26. THE  FAILOVERHOST  To ensure that spare capacity is available on a failover host, you areprevented from powering on virtual machines or using vMotion tomigrate VMs to a failover host.Also, DRS does not use a failover host for load balancingIf you use the Specify Failover Hosts admission control policy anddesignate multiple failover hosts, DRS does not attempt to enforce VM-VM affinity rules for virtual machines that are running on failover hosts.
  27. 27. STATUS  OF  THE  CURRENT  FAILOVER  HOSTS   Green - The host is connected, not in maintenance mode, and has no vSphere HA errors. No powered-on VMs reside on the host. Yellow - The host is connected, not in maintenance mode, and has no vSphere HA errors. However, powered-on VMs reside on the host. Red - The host is disconnected, in maintenance mode, or has vSphere HA errors.
  28. 28. MYTH  BUSTED  •  VMware  High  Availability  needs  to  be  configured  •  Be  careful  with  reservaHons  •  Always  check  run-­‐Hme  informaHon  
  29. 29. MYTH  2  VMware  snapshots  impacts  performance  
  30. 30. WHAT  IS  A  SNAPSHOT?  •  Preserves  state  and  data  of  a  VM  at  a  specific  point  in   Hme  •  Data  includes  virtual  disks,  se3ngs,  memory  (opHonally)  •  Allows  you  to  revert  to  a  previous  state  •  Typically  used  by  VM  admins  when  doing  changes  and   by  backup  so@ware  •  ESX3,  ESX(i)4  had  issues  with  deleHng  snapshots  •  ESXi5  improved  snapshot  consolidaHon  
  31. 31. WHAT  IS  A  SNAPSHOT?   File   Descrip<on   .vmdk   Original  virtual  disk   delta.vmdk   Snapshot  delta  disk   .vmsd   DB  file  with  relaHons  between   snapshots   .vmsn   Memory  file  •  Snapshot  grows  in  16MB  chunks   –  Requires  locking      
  32. 32. LOCKS  •  Locks  are  necessary  when  creaHng,  deleHng  and   growing  snapshot,  power  on/off,  create  VMDK  •  ESX(i)4  used  SCSI-­‐2  reservaHon   –  Locks  enHre  LUN    
  33. 33. LOCKS  •  ESXi5  uses  Atomic  Test  &  Set  (ATS)  VAAI  primiHve   –  Locks  only  individual  VM   –  Requires  VAAI  enabled  array  and  VMFS-­‐5      
  34. 34. PERFORMANCE  •  Locking   –  ATS  increase  performance  up  to  70%  compared  to   SCSI-­‐2  reservaHon    •  Normal  operaHons   –  Snapshot  age   –  Number  of  snapshots   –  Snapshot  size  •  Be  careful  with  snapshots  in  produc<on!    
  35. 35. MYTH  NOT  BUSTED  •  Improvements  to  snapshots  management  and  locking  •  Snapshots  sHll  have  impact  on  performance  
  36. 36. MYTH  3  Disk  provisioning  type  doesn’t  affect  performance  
  37. 37. DISK  TYPES  
  38. 38. BLOCK  ALLOCATION  Wrimen  Blocks   Block   Block   Block   Thick Provision Lazy ZeroedVMDK  File  Size   VMDK  Wrimen  Blocks   Block   Block   Block   Thin ProvisionVMDK  File  Size   VMDK   VMDK   VMDK  Wrimen  Blocks   Block   Block   Block   Thick Provision Eager ZeroedVMDK  File  Size   VMDK  
  39. 39. THE  ISCSI  LABORATORY  •  Iomega  StorCenter  px6-­‐300d  with  6  SATA  7200  Disks    •  Windows  2008  R2   4096  MB  –  1  vCPU   Hardware  Version  9  •  VMware  vSphere  5.1  •  Single  Intel  1GB  Ethernet  •  Cisco  2960  switch   MTU  Size  1500  
  40. 40. 3  DIFFERENT  DISKS   •  Thick Provision Lazy Zeroed •  Thin Provision •  Thick Provision Eager Zeroed
  41. 41. THICK  PROVISION  LAZY  ZEROED   Average Write 13.3 MB/s - Access time: 44.8 ms
  42. 42. THIN  PROVISION   Average Write 13.7 MB/s - Access time: 46.8 ms
  43. 43. THICK  PROVISION  EAGER  ZEROED     Average Write 86.6 MB/s - Access time: 9.85 ms
  44. 44. COMPARISION  THICK  PROVISION  LAZY  ZEROED  Average Write 13.3 MB/s - Access time: 44.8 msTHIN  PROVISION  Average Write 13.7 MB/s - Access time: 46.8 msTHICK  PROVISION  EAGER  ZEROED  Average Write 86.6 MB/s - Access time: 9.85 ms
  45. 45. MIGRATION  •  Storage  vMoHon  is  able  to  migrate  the  disk  format  of  a   Virtual  Machine  
  46. 46. MYTH  BUSTED  •  Thin  and  Lazy  Zeroed  disks  have  the  same  speed  •  Once  allocated,  these  disks  are  as  fast  as  Zeroed  disks  •  Thick  Provision  Eager  Zeroed  offer  best  performance   from  first  write  on  
  47. 47. MYTH  4  Always  use  VMware  tools  to  sync  the  Hme  in  your  VM  
  48. 48. TIME  SYNC  PROBLEMS  •  VMs  have  not  access  to  naHve  physical  HW   Hmers  •  Scheduling  can  cause  Hme  to  fall  behind  •  CPU  /  Memory  overcommit  increases  risk  •  People  are  mixing  different  Hme  sync  opHons    
  49. 49. VMWARE  TOOLS  •  ESX(i)  4  and  prior  –  not  possible  to  adjust  Hme   backwards  •  ESXi  5  –  Improved  Hme  sync  to  be  more  accurate   and  can  also  adjust  Hme  backwards    •  Enable/Disable  periodic  sync  in  VMware  Tools  GUI,   vCenter  or  VMX  file  
  50. 50. VMWARE  TOOLS  •  Default  periodic  sync  interval  is  60  sec  •  Sync  is  forced  even  when  periodic  sync  is  disabled:   –  Resume,  Revert  Snapshot,  Disk  Shrink  and   vMoHon  •  In  order  to  disable  completely  configure  vmx  file   –  TesHng  scenarios     tools.syncTime  =  FALSE   Hme.synchronize.conHnue  =  FALSE   Hme.synchronize.restore  =  FALSE   Hme.synchronize.resume.disk  =  FALSE   Hme.synchronize.shrink  =  FALSE  =  FALSE  =  FALSE      
  51. 51. GUEST  OS  SERVICES  •  Windows  (W32Time  service)   –  Windows  2000  uses  SNTP   –  Windows  2003+  uses  NTP  and  provides  bemer  sync   opHons  and  accuracy   –  Domain  joined  VMs  sync  from  DC   –  Use  Group  Policy  to  control  se3ngs  •  Linux  (NTP)   –  Configure  ntpd.conf   –  Start  ntpd   •  chkconfig  ntpd  on   •  /etc/init.d/ntpd  start      
  52. 52. BEST  PRACTICES  •  ESX(i)  hosts:   –  Configure  mulHple  NTP  servers   –  Start  NTP  Service  •  Virtual  Machines:   –  Disable  VMware  Tools  periodic  sync   –  DC:  Configure  mulHple  NTP  servers  (same  as  ESX(i)   host)   –  Domain  joined  will  sync  with  DC   –  If  not  domain  joined  then  configure  W32Time  or  NTP   manually  •  Do  not  use  both  VMware  Tools  periodic  sync  and  Guest   OS  <me  sync  simultaneously!  
  53. 53. MYTH  BUSTED  •  Use  W32Time  or  NTP    •  Do  not  use  VMware  Tools  period  sync  
  54. 54. SUMMARY  •  Myth  1:  VMware  High  Availability  needs  to  be   configured,  be  careful  with  reservaHons  and  always   check  run-­‐Hme  informaHon  •  Myth  2:Improvements  to  snapshot  management  and   locking  but  sHll  performance  impact  •  Myth  3:  Use  Thick  Eager  Zeroed  disks  for  best  I/O   performance  •  Myth  4:  Use  W32Time  or  NTP  to  sync  Hme  instead  of   VMware  Tools      
  55. 55. VMWORLDTV  •  hmp://  
  56. 56. QUESTIONS  Eric  Sloof   MaJas  Sundling  VMware  CerHfied  Instructor   Evangelist  NTPRO.NL   Dell  So@ware,  @esloof,  @msundling