April 17, 2026
Pods, plots, and popcorn
Precision over perception: Why architecture matters in benchmarking
5.6x or 5.6x the spin? Users say it’s 300 minivans vs 4 buses
TLDR: VMware touted a “5.6x” win after testing 300 tiny virtual nodes against OpenShift’s 4 big ones, using pods that did no real work. Commenters call it apples-to-orchards math, demanding apples-to-apples tests and mocking the result as queueing theater rather than true speed or efficiency.
VMware bragged that its cloud stack hit “5.6x” more tiny app containers (called pods) than Red Hat’s OpenShift, citing a Principled Technologies study. The comments? On fire. The loudest voices say the headline hides a simple trick: VMware spread the load across 300 small virtual machines on the same four servers, while OpenShift ran on just 4 big machines. Cue the memes: “It’s not faster cars, it’s more parking spots.”
Armchair referees and seasoned admins piled on. One camp calls it “benchmarketing,” pointing out that the pods in this test did nothing—no real work—so it’s just measuring how fast you can check empty boxes. Others note the configuration mismatch: VMware capped at 200 pods per node; OpenShift cranked its cap way up to 5,000, yet the test only bragged about the total number, not efficiency per machine. The “faster startup” claim? Commenters say that’s just queueing theory—300 lanes will beat 4 lanes every time. Several begged for a fair fight using OpenShift Virtualization to mirror VMware’s setup. A few defenders shrugged, “Marketing gonna market,” but the chorus shouting “apples vs orchards” drowned them out. Links to the VMware post and the YAML config flew around like receipts.
Key Points
- •VMware cited a Principled Technologies study claiming VCF 9.0 with VKS delivers 5.6x higher pod density than Red Hat OpenShift.
- •The test used four identical Dell PowerEdge servers but compared 300 VKS virtual worker nodes (75 per host) against four OpenShift bare-metal worker nodes.
- •Reported totals were 42,000 pods for VKS (~140 per node) and 7,400 for OpenShift (~1,850 per node), indicating higher per-node density for OpenShift.
- •MaxPods settings differed: VKS was configured at 200 pods/node (per YAML), OpenShift at 5,000 pods/node (above its 250 default), yielding theoretical ceilings of 60,000 vs 20,000.
- •“Faster pod readiness” for VKS is attributed to reduced scheduling contention from many nodes and the use of empty kube-burner pods, not to workload performance; the article suggests comparing OpenShift on OpenShift Virtualization for a fairer test.