After fifteen years on this platform, the thing I keep coming back to is “boring uptime,” and I mean that as the highest compliment. I’ve had production systems run for years without a restart. When you’re supporting financial workloads, that level of reliability stops being a feature and becomes the baseline for how you operate.
What genuinely shifted my thinking more recently is the on-chip AI inferencing with Telum II. We pushed fraud-detection models directly onto the processor, no GPU cluster and no network hop. Sub-millisecond inference happens right where the transaction occurs. I honestly didn’t expect it to work that cleanly, but it did, and it’s changed how I architect solutions.
The virtualization density is something x86 folks often don’t fully appreciate until they see it in practice. Running hundreds of Linux guests on z/VM with rock-solid isolation makes the usual noisy-neighbor problem basically disappear. And when we started running OpenShift on LinuxONE, the integration with our existing DevOps pipelines was smoother than I expected. Red Hat tooling, Ansible, Terraform, it all works, and the ecosystem has clearly matured.
Security also feels like it’s built into the silicon rather than bolted on later: pervasive encryption, Secure Execution, and confidential computing. In regulated industries, that ends up being the entire conversation with auditors.
The ROI discussion used to be the hardest part. Not anymore. Once you factor in consolidation, licensing reduction, energy savings, and the compliance overhead you avoid, the numbers start to speak for themselves. I’ve seen organizations save millions over five years compared to an equivalent x86 sprawl.
Fifteen years in, I’m still finding new things to appreciate. That doesn’t happen with much technology. Review collected by and hosted on G2.com.
Look, I genuinely love this platform but I'd be doing nobody any favors pretending it's perfect.
The biggest pain point, even after all these years, is the initial onboarding complexity. When we bring in new Linux or cloud engineers onto a LinuxONE project, there's a learning curve that's steeper than it should be. Concepts like LPARs, z/VM, HMC... they're not intuitive if you're coming from a pure x86 or cloud background. IBM has improved the HMC web interface considerably over the years, but honestly it still feels like it was designed by engineers for engineers. A more modern, simplified UI experience would go a long way in reducing that ramp-up time.
The pricing and licensing model is another area where I've lost sleep. The initial acquisition cost is significant and justifying it to finance teams requires a detailed TCO exercise every single time. IBM has gotten better at providing tools and reference architectures to support that conversation, but the sticker shock is real for first-time buyers. A more transparent, flexible consumption-based pricing option would make the sales cycle much less painful.
Skills availability in the market is also a genuine concern. Finding people who know z/VM, zSecure or even basic s390x Linux administration is harder than it should be. IBM and the community need to invest more aggressively in training and certification accessibility.
And while the AI capabilities on Telum II are impressive, the tooling around model management and MLOps on the platform is still catching up to what data scientists expect from more established AI infrastructure. The AI Toolkit is improving but needs to move faster.
None of these are dealbreakers for me. But they're real friction points worth being honest about. Review collected by and hosted on G2.com.