VMware recently did something rare in enterprise tech: it admitted it overshot.
Specifically, the virtualization provider disclosed that it has, for years, overestimated the hardware provisioning requirements for its storage technology, vSAN. And now the Broadcom-owned company is walking back its guidance, with measurable implications for partners who build, size and support VMware Cloud Foundation environments.
VMware’s unexpected shift is rooted in a foundational truth: synthetic testing behaves nothing like real production workloads. And hardware recommendations for vSAN historically have come from this approach. That’s the word from Pete Koehler, product marketing engineer in the VMware Cloud Foundation division at Broadcom, in a Nov. 14 blog post. Although synthetic tests pushed vSAN hardware under extreme load, they did not reflect how the product actually behaves in the field, he wrote.
VMware then turned to its own data and discovered that its mandates were too onerous.
What VMware found
According to Koehler, VMware sifted through telemetry from thousands of production clusters and found that vSAN ESA is far more resource-efficient than the company’s older guidance suggested.
Here, Koehler said, VMware learned that vSAN clusters use much less RAM and fewer CPU resources than expected. That prompted company leaders to take two steps: reduce the minimum requirements for vSAN and make it easier to select the appropriate ReadyNode, Koehler said. Combined, these adjustments “offer significant cost savings for any VCF environment powered by vSAN … [and] can simplify the entire design and sizing experience,” Koehler wrote.
That’s huge. Since acquiring VMware in November 2023, chipmaker Broadcom has steadily increased prices and tightened licensing terms, squeezing out many longtime partners and customers. Downsizing vSAN’s hardware requirements marks a rare win for VMware partners who have been through the wringer over the last two years.
Koehler said vSAN storage clusters now need up to 67% less RAM and up to 33% fewer CPU cores. HCI clusters, meanwhile, require up to 50% less RAM. These parameters apply across vSAN ESA releases from 8.0 forward, including the latest VCF-aligned builds.
VMware also cut the ReadyNode lineup down to three simplified profiles for each deployment type, another nod to reducing architectural complexity.
Why VMware is shifting gears
The Register framed the matter bluntly: VMware “over-specced storage servers for years.” Again, the culprit was synthetic testing that didn’t match reality. Now, with data pointing to lower resource needs, VMware is adjusting its guidance and lowering minimum specs. The Register noted that this will be welcome news for customers dealing with bill shock under Broadcom’s licensing changes.
TechRadar took a similar tone, saying VMware’s old guidance was “pretty much unfounded.”
The new requirements bring some tangible perks to VMware partners:
- The hardware cost drop opens the door for stalled VCF deals. Koehler said the per-host savings could land “in the five figures” based on current street pricing. Multiply that across a cluster and customers who balked at VCF’s economics may reconsider now that hardware minimums are more forgiving. For partners, this can reopen opportunities that went cold when Broadcom narrowed licensing around VCF and vSAN.
- Fewer hosts might be needed. Because vSAN clusters need less RAM and fewer cores, VMware says customers could get by with fewer physical hosts. That translates into savings on network ports, power, cooling, and rack space, resulting in simpler designs and lower total cost of ownership. Partners who lead with assessment and right-sizing services stand to benefit here.
- Partners now have a stronger competitive posture. With lower hardware requirements and deeper VCF integration, VMware is strengthening its case against alternative virtual storage stacks that have gained traction due to price sensitivity. “It’s good news for VMware parent company Broadcom, which wants buyers to go all-in on its Cloud Foundation private cloud stack for virtual compute, networks, and storage, instead of using external arrays,” The Register wrote. All in all, the changes give partners a more cost-effective VCF narrative amid shakeups in the virtualization market.
- Storage-only clusters will see the biggest wins. VMware’s data drove the most dramatic cuts on storage-only ReadyNodes. For customers running storage clusters separately from compute, this creates room for higher capacity at lower cost and reduces or even removes the question of whether to leave vSAN, which is not easy to do. “[M]oving off vSAN brings considerations beyond simply choosing a new supplier,” as Computer Weekly pointed out. “Organizations must ensure any new storage architecture matches vSAN’s capabilities and plan a migration that minimizes downtime and the risk of data loss.”
On the whole, VMware partners now have a more compelling sales motion: Run vSAN ESA with fewer resources, spend less on hardware, reduce opex, maintain performance and stay aligned with Broadcom’s preferred architecture.
As Koehler put it, “vSAN lets you do more with what you already have.”
