I saw a post on a Facebook group that linked to this article by Vito Botta who runs dynablogger (Check it out and support fellow entrepreneurs)!
Great article, lots of data. Love it!
Notes and Feedback
Different Test Candidates
There’s a mix of test candidates using different configurations. I think he should have bumped up all the VPS to 12 cores to match the bare metal instance, or removed the dedicated completely. But I don’t hate what he did, any review like this is great data to consider. I’m just selfish and wanted a review of VPS vs Bare Metal to show the cost and performance differences.
Also, sticking to either high frequency or low-frequency processors. The DO and Linode instances were probably general-purpose, and 2.3Ghz older Intel/AMD processors.
VPS Instances have Shared I/O
I’m not surprised at the I/O speeds on all the VPS options. This is typical, shared local storage with no real I/O guarantee. In fact, they’re probably limiting I/O to have some sort of guaranteed I/O for all instances on the node.
“Not surprisingly, the dedicated server’s multi-core score is higher than that of the Terrahost virtual servers, because it has 6 cores vs 4”
Both instances are using a lower clock Epyc processor. I’m assuming this is a hypervisor efficiency or performance issue. The shared instance will have access to all cores on the server, versus the dedicated being locked to four. Something is strange here.
It would have been great if he ran these tests over time to get a good idea of resource availability.
Definitely love those Terraform instances though with the Ryzen processors 🙂
Love this article and hope Vito does some more. No hate for anyone taking the time and effort to construct these tests.