4 ms·
I think it's kind of the path of least resistance to improve hardware instead of software. A small group of individuals making incremental hardware improvements
by bostonsre 5y ago
I think it's kind of the path of least resistance to improve hardware instead of software. A small group of individuals making incremental hardware improvements is much easier to scale than having the entire software industry make optimizations to use less resources. It's also generally a lot cheaper to throw money towards new hardware instead of towards software optimizations at smaller organizations. For larger organizations, that calculus changes and the ROI of performance optimizations becomes more attractive. Some of those optimizations make it upstream to open source projects, but a lot of it is one off work that only benefits that company.
162 of the lex podcast with Jim Keller (a hardware god) dives into this well.
- rakoo 5y agoI think it's even simpler than that: making software leaner doesn't bring in much money. Making more features does. Selling more hardware does.
- bostonsre 5y agoRight, making software leaner for small organizations doesn't give a good return on investment for the effort applied. But for larger organizations that have a massive compute bill like say Netflix, the return on investment of software optimizations across that massive compute footprint saves millions of dollars. So it makes sense to have performance engineers like Brendan Gregg on the pay roll. His book on systems performance is ridiculously good and is useful for smaller orgs as well when systems performance degrades to the point where the customer is impacted.