A recent study by Principled Technologies has shown that you can get greater Kubernetes pod density on VMware with Tanzu than with bare metal Kubernetes with Ubuntu Linux. Including comparable and sometimes better performance.
According to Vandana Venkatesan on the VMware vSphere Blog:
PT found that 13,700 pods were supported by vSphere with Tanzu versus only 2,150 pods supported by a Red Hat OpenShift bare metal deployment on identical hardware.
On the Red Hat OpenShift deployment, PT stopped scaling up the pods when they started seeing worker node instability, at 2,150 pods. Red Hat documents a default of 250 pods but allows this to grow to a maximum of 500 via additional configuration. For this testing, PT increased the OpenShift pod limit to 500 using the configuration knobs.
On the vSphere with Tanzu platform, PT stopped scaling up the pods at 13,700 pods, when the memory utilization reached 95% to avoid performance deterioration. vSphere with Tanzu was not only able to reach 6.3 times the number of pods, but also did not suffer node instability issues.
My first thought was the cost savings that this could bring to your IT team. More pods on fewer machines means you are spending less on hardware and software licenses. I know this will appeal to many IT departments but don’t let that be your driving factor. Let savings be a secondary win for the team. Always focus on the business outcomes and ensuring you are delivering a service without any shortcuts.
Are you running vSphere with Tanzu? Have you experienced any cost savings as a secondary benefit? What are your thoughts on this report, do you agree or disagree? How large is your Tanzu deployment? Let me know on Twitter and thanks for reading.