It is always good to learn from others' experiences. The following are the use cases published by some famous organizations that can give us some idea of the hardware specification.
In 2012, Netflix had the following configuration:
2x quad core x86_64 for the Jenkins master with 26 GB RAM
1 Jenkins master with 700 engineers using it
Elastic slaves with Amazon EC2 + 40 ad-hoc slaves in Netflix's data center
Work load:
1,600 Jenkins jobs
2,000 Builds per day
2 TB of build data
In 2013, Yahoo! had the following configuration:
2 x Xeon E5645 2.40GHz, 4.80GT QPI (HT enabled, 12 cores, 24 threads) with 96 GB RAM, and 1.2 TB of disk space
1 Jenkins master with 1,000 engineers using it
48 GB max heap to JVM
$JENKINS_HOME*
lives on NetApp20 TB filer volume to store Jenkins job and build data
50 Jenkins slaves in three data centers
Workload: