We're mixing up apples and oranges with different operating systems, new versus existing and there is no discussion of runtimes.
Comments are my own opinion and like LLMs I might be hallucinating. My view (for good or bad). IBM Z a few segments. *z/OS is a core business and runs existing apps and data systems.* These include CICS, IMS, Db2, MQ, Datacom, IDMS, and a smattering of other software that exists in the z/OS ecosystem. Security is well defined and the processes for managing those workloads exist and are delivering value today. These workloads will continue to grow based on transaction volume, new features / applications and data requirements. The main strength of z/OS is a super tight multi-resource manager commit process that can't be duplicated in the cloud. (ACID versus BASE architectures). It is a massively solid I/O beast and runs mixed workloads with ease. It is solid, of great value and will grow based on its software stack. It continues to deliver superior value for the runtimes and workloads that run the business. Another significant benefit is reduced latency (better response time) when incorporating new capabilities like scoring (Telum) and LLMs (Spyre). It is a fantastic platform for workloads with many moving parts and the need for high volume. It is not easy to start from scratch in terms of Storage subsystems, network configuration, security mangement, and all the other components that interoperate to make this system highly scalable and efficient. I would acknowledge that once it's set up it can't be beat, but, starting from scratch is likely a multi-month effort from planning, to flooring, environmentals, storage, networking, and incorporating into the environment. You need people with IBM z Hardware and software skills. zTPF out there as well. I don't know a lot about it but financial institutions love it for high performance transactions z/Linux in the form of a Kubernetes infrastructure. This is the runtime every new person out of college knows and understands. Its container based, automation is done in a common way as part of K8s and network administration and coordination with the application deployments as Pods is standard. It very easy to standup an isolated application leveraging other applications via services. Even ISVs that support z/OS (including IBM) are deploying support software in Kubernetes because the HA characteristics can be leverage by the application through declaration and not configuration. This is where the cloud providers would benefit but like in the case of Google, they have their own proprietary clustering solutions and I suspect its the same for all cloud providers. The hardware is different, the architecture is different (meaning containers need to be built for the platform). Building these are almost impossible in the ecosystem because there are few open source projects or tools that have any access to an s390x system. Its just a reality. >From my perspective, there is a very low-likelyhood of a major cloud provider picking up s390x given they are already entrenched and the business value to them would have to be substantial beyond efficiency and energy consumption. On Fri, Oct 24, 2025 at 9:59 AM Radoslaw Skorupka < [email protected]> wrote: > That's wrong answer. > Yes, Google don't buy PC servers. > However they don't make anything. They buy CPUs. Including IBM POWER (a > lot of). > They produce their own *cheap servers* even with no enclosure. Obviously > (or not) they don't produce motherboards... but IBM don't do it as well. > They also buy hard drives, but they don't produce them. > Regarding mainframe - they simply don't need big reliable machine, > because they built their infrastructure on plenty of small, cheap and > not very reliable PCs. > > -- > Radoslaw Skorupka > Lodz, Poland > > > > > W dniu 24.10.2025 o 15:37, Peter Vander Woude pisze: > > On Thu, 23 Oct 2025 15:54:48 -0500, Jon Perryman <[email protected]> > wrote: > > > > The answer to the question below is that Google does not buy it's server > or storage hardware from others. The last I heard, they built their own > servers and unless they have changed their process, they build their own > disk storage subsystems also. > > > >> How is it that Google runs 5M servers yet not one of them is a z16 or > z17? Google provides cloud services on horribly designed software. Google > provides AI in a horribly configured environment when a z17 is technically > simalar to the El Capitan supercomputer. > >> > > ---------------------------------------------------------------------- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to [email protected] with the message: INFO IBM-MAIN > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
