Nobody want to stop normal benchmarks like rendering, DBs, AI, etc... but I think of a more relevant use case such running tens of VMs, all at the same time with different workloads like webservers, DBs, AI, backups, mail servers, etc..., all running at full load and measuring metrics like throughput, latency, time taken to complete tasks and power usage.
I like you idea. Phoronix just installed Ubuntu 22.04 Server and ran each thing as an individual benchmark.
They also killed the free hypervisor, which as far as I know makes them the only player in the game without a free hypervisor. College kids with no money grow in to engineers with big budgets. Having exposure and goodwill across a large spectrum is one of those soft benefits that's hard to quantify, but I believe pays off big in the end.
I like that idea, by it seems like a nightmare to get consistency and comparable results between runs with a layout like that.
My only question is how long until they charge people for that again. You get people hooked on it and using it all the time and then once it becomes a requirement you start charging for it. Seems like something Broadcomm would love to do.
First taste is free, eh? It seems to me VMware's niche is shrinking, so maybe this is just a play to squeeze as much money as possible out of a dying market? Ramp prices quickly before folks are ready for a transition? I prefer VMware's ecosystem of management tools, but MS is building a compelling alternative for those that can shift to a different way of doing things. The totally free combo of Hyper-V server and WAC work very well for small, cost conscious deployments, and transitions pretty easily to Azure Stack HCI that can scale to massive deployments that seamlessly span on-prem and cloud and gain the entire management ecosystem of Azure.
Only reason VMware's niche is shrinking is because they are alienating anyone who isn't a massive corporation. When VMware went to 32 cores/socket licensing back in 2019, it wasn't that bad as most companies didn't have more than 32 cores/socket. If you had more than 32 cores you needed multiple licenses for that socket. With core counts maxing out at 64 cores at the time you could still increase your core counts and still be able to consolidate without extra costs (assuming you went from a quad socket 20c/40t host > dual socket 64c/128t host). After the Broadcomm purchase they went to a per core license on a subscription and got rid of the perpetual licenses. That alienated A LOT of companies. For where I work we cannot go to Hyper-V, one reason is cost but the other is they are a supported hypervisor for SAP HANA. We cannot even go to free ones like Proxmox due to this. Basically we can run SUSE hypervisor, Nutanix hypervisor, and Red Hat KVM. Sadly we cannot even go to basic KVM either. Overall VMware did screw over a lot of people. I see a lot of VMware Administrators (myself included) are going to be working on other hypervisors in the near future and having to learn an entire eco system without the same level of training that was available from VMware.
Facebook
Twitter
Instagram