Interesting. The more frequent comment in the comments of the article is “if you use swap, you’re doing it wrong”. What is the current status on performance engineering? Is swap commonly disabled? It does help failing faster, may lower latency, but is it actually disabled in production systems?