Servers, Storage and Virtualization · #50

Online Server Consolidation Planner

Turns a physical server inventory into virtual requirements (vCPU, RAM, disk, host count).

  • Data is processed in your browser and not sent to a server
  • Export: CSV

Tool

cpu_peak = busiest period percentage; if ram_used_gb is empty installed RAM is used.

Expected increase over the planning period.

Physical cores per target host.

What does this tool do?

When virtualizing physical servers, adding up nameplate values (cores, installed RAM) overstates host needs two- to threefold, while average usage alone causes bottlenecks at peak times. The tool derives requirements from peak CPU and used RAM, adds 25 % CPU and 10 % RAM headroom, sizes servers without measurements with conservative assumptions (CPU 100 %, RAM = installed) and flags them clearly.

How to use it

  1. CSV header: server, cores, cpu_peak, ram_gb, ram_used_gb, disk_gb, group (server, cores and ram_gb required).
  2. Enter growth and the target host's cores, RAM, reserve and vCPU ratio.
  3. Press Calculate; download the table as CSV.

Fields

Server inventory (CSV)
cpu_peak = busiest period percentage; if ram_used_gb is empty installed RAM is used.
Growth
Expected increase over the planning period.
Host cores
Physical cores per target host.
vCPU:core ratio
Host RAM
Host reserve

Worked example

InputValue
Server inventory (CSV)
server,cores,cpu_peak,ram_gb,ram_used_gb,disk_gb,group
DC01,4,15,8,5,100,ad
DC02,4,12,8,5,100,ad
SQL01,16,70,128,110,2000,erp
ERP01,8,55,64,48,500,erp
FILE01,4,20,16,,4000,
RDS01,16,85,96,80,300,
Growth20
Host cores32
vCPU:core ratio4
Host RAM384
Host reserve16

Expected result: 6 servers fit on one host (2 with N+1): 49 vCPU and 352 GiB RAM against 128 vCPU / 368 GiB per host; FILE01 has no RAM measurement, so installed 16 GB was used.

RAM determines the host count and a single host is 96 % full; once the growth headroom is used a second host becomes mandatory. With a FILE01 measurement the requirement may drop. The ad and erp groups are listed as dependencies: move servers of the same group together and place DCs on different hosts.

Second scenario

InputValue
Server inventory (CSV)
server,cores,cpu_peak,ram_gb,ram_used_gb,disk_gb
WEB01,8,,32,20,200
WEB02,8,,32,22,200
MAIL01,8,40,64,60,1500
Growth0
Host cores16
vCPU:core ratio3
Host RAM256
Host reserve12

Expected result: WEB01 and WEB02 have no peak CPU measurement; 100 % was assumed and both rows are marked amber.

Method and assumptions

Required vCPU = ceil(cores × peak % × (1 + growth) × 1.25), at least 1. Required RAM = ceil(used RAM × (1 + growth) × 1.10).

Hosts = max(ceil(total vCPU / (host cores × ratio)), ceil(total RAM / (host RAM − reserve))). One host is added for N+1.

Using the result and troubleshooting

Collect at least two weeks of performance data for servers without measurements (perfmon, sar or your monitoring system). Validate the result with the VM vCPU/RAM and N+1 tools.

Limits and common mistakes

  • Disk IOPS, network traffic and licensing constraints are not considered; disk is shown as a total only. Performance differences between CPU generations are not normalized.

Frequently asked questions

Can I use average CPU instead of peak?

Not recommended. Averages drop with night hours; size by the 95th percentile or business-hours peak.

Related tools and service

Related service: Virtualization and Data Management

Systems online tools · See all online tools