::: Virtual Aleph ::: Virtualization Techniques: MA7528 - VMWare vCenter Server: Operational Best Practices in the Datacenter

VMworld 2016 Banner

VMworld 2016 Banner
VMworld 2016 Barcelona

12 October, 2010

MA7528 - VMWare vCenter Server: Operational Best Practices in the Datacenter

In this session
Ashawin Kotian and Praveen Kannan (both VMWare product manager for vCenter product) describe operational best practice in the DC for vCenter.
All is in function of levereging resources pooling on-demand
they have done a quick overview on how to get to the Cloud and the three ways are:
    enterprise private cloud
    service provides public cloud
    hybrid cloud bridging then two
the datacenter of the future is an aggregation of resources shared among business units and among users on-demand.
vCenter is and will be the private cloud management platform (foto slide)
New in vCenter 4.x,  64 bits support tobe more scalable and to minimize the limitations that we encounter with the 32 bit version.
We can execute vCenter on more powerful hardware using  all the memory above 4 GBytes
compared with the 32 bits releases we can support 4x hosts pe DC, 5x VMs per DC, 3x Powered on VMs per DC (now 10000 vms per vCenter powered on). Also we have some hosts profile enahancements and vCenter update manager support EMC powerpath
We can import Microsoft Hyper-V with vCenter Converter and ESX ESXi integration in Active directory has been added.
Virtual serial ports concentartors has been added too

the four phases of vCenter management lifecycle are
   plan
   deploy
   operate
   protect

plan:
think about the workload types
do size the infrastructure in a good way
think about geographical location and about the QOS of the whole infrastructure

deploy:
keep attention on how much statistics you re collecting: default is fine
database sizing calculator online on vmware site

linked mode, based on adam. Keep in mind to establish trust relationship between domains, dns must be perfect, leverage NTP to ensure <5 minutes drift between all the vCenter.

the deployment choice must not be lead by bandwidth consideration but by latency. (64kbps fare well for most operations)

keep in mind that the below operation put large payload on vCenter Server
adding hosts
patching with vCenter Update Manager

if possinble use rdp or ssh to consume less bandwidth while connecting to Virtual Machine instead of using console (unless necessary)

MONITORING vCenter HEALTH
the core process is vpxd
then tomcat
then DB
then ESX hosts side hostd and vpxa
learn how to interpret vpxd.log: timestamp, thread id, log level, task id,

PROTECT:
vCenter is, and for sure will be, critical part of the virtual infrastructure.
it depends on rto/rpo and how many time you can afford have no management on your infrastructure: vCenter Heartbeat is a solution for enterprise class environment and for very important infrastucture

cheers
\mf