Monthly Archives: December 2010

Server Consolidation and Virtualization

These are the just outlined discussion regarding Server Consolidation and Virtualization steps.

1 -Existing Environment Understanding like what is current environment and setup in terms of servers like Linux & windows.
2 -Assess Existing Environment with Virtual Capacity Planner for collecting and analysis data based on Processor speed/CPU frequency/CPU utilization, RAM for highest used/lowest used and hard disk capacity based on IO operation like read/write/TPS etc and compare with industry driven standard.
3 -Calculate Power and cooling cost for existing environment hardware in term of WATTS/AMP/BTU and once the no. of ESX serer determined compare this with the virtualized environment.
4 -Prepare ESX server list to host the no of physical server to be consolidate and virtualized like 12 esx server required for 134 physical server along with HA {N+1 Approach} ,DRA and VMotion etc if we have proper license for the same.
5 – This configuring is purely based on HA N+1 approach, we can go for N+2 approach as well but it all depends on budget and planning
Subsequently this HA N+1 ( Failover apporach) is based on HA SLOT configuration like if we know how many vm we can run on one esx server then what happened if one of esx server goes down in cluster, will the remaining esx servers can host the existing vm on themselves without compromising the performance.
6 – Slot is just a combination of whole CPU +RAM on the cluster. Pick the esx host which has the least amount of RAM in my case this is 24GB and find out the highest usage of RAM of physical server { you can get it with the virtual capacity planner result } that is plan for virtualization like 2GB. Divide the least amount of ESX RAM/ highest RAM of VM i.e 24/2 = 12Slot
7 – So the total ESX server required to host the no of VM is = Total no of VM (total physical hosts that needs to be virtualized) / no of Slot

For example –

Total physical hosts that need to be virtualized          = 134VM
Highest usage of RAM of physical server                  = 2GB
Planned ESX server RAM size                                   = 24GB

Slot required to maintain the HA                               = 24GB/2GB = 12

Total ESX servers required to handle this                  = 134/12 = 11.16 (Round figure 11)

Note – This is without N+1 architecture , this configuration require 11 ESX servers but we would like to have one more esx server to support incase one esx server goes down
Total ESX server with N+1 Architecture is         = 12

8 – After working on N+1 architecture, we can setup the Cluster as per the plan but I loved the following way of doing the things to identify the VM based on Critical and Non-Critical for Performance issue.
(a) – Identify the Critical VMs and plan HA+DRS cluster for the same.
(b) – Identify the Non-Critical VMs and plan DRS cluster for the same.

Why I am saying is that  for Planned maintenance, Vmotion work fine but for unplanned the HA only support 🙂

9 -LUNS architectural Approach for shared storage for the VMs.


Total no. of VMs per ESX                             =       11.16 (Round 11)
Avg. size of VMDK disk                                 =       40GB
Total Storage required/Host                      =       440GB
Total no. of VMs                                               =       134
Total Storage required                                  =       5360GB
Total no of ESX servers                                 =       12

No. of form required                                     =       1 (Due to less than 16 ESX host – Being conservative)
No. of Servers /cluster                                 =       12
Total VMs/Cluster                                         =       134
Total LUNS for disk-intesive                     =       134/32 = 4.1875
Total no of LUNs required                         =       4
Size per LUNS (with + 10% Growth)      =       1474GB (1.5T)

There is other way around to configure storage is by the following formula

((Avg. size of VMDK disk * Total no. of VMs) * (LUNS + 10% Growth))


– Its upto you what growth rate you want as I took 10% growth per annual
10 -Rest is Environment specific.

I have tried to cover few things; I know there are a lot of things to keep in mind before planning and implantation.


Why one should go for Server consolidation and Virtualization – Hard Fact


IDC researched the following compelling reasons for consolidation


• For every dollar spent on hardware, 50 cents is spent on power and cooling.
This number has increased consistently over the last several years, and
this is no doubt one of the primary drivers for virtualization and consolidation.

• The average data center is 52% utilized and 1/3 of data centers are
approaching maximum floor space capacity.

• The average rack is 56% utilized and the 1/3 of data centers are approaching
maximum rack space density.

The following are some power and cooling-related facts:

• Power and cooling shortages cause server or storage downtime in 49% of
data centers.

• According to American Power Conversion (APC), server and storage consume
50% of power and air conditioning consumes 34% of power.

• According to McKinsey, deploying virtualization saves 25-30% of power.
• According to the Department of Energy (DOE) data center electricity has
doubled since 2000, costing about $4.1billion.