washingtonpost.com — Гейб Ньюелл из Valve рассказывает об управлении компанией — и, в частности, использует интересный подход к управлению крупной компанией как к задаче по масштабированию:
“Part of me as somebody with a computer software background thinks most systems that have trouble scaling it's because of internal communications bandwidth problems. If you look at the old way of doing things called master-slave computing systems, they tended to clog up around I/O if you really want to scale something to huge levels what you have to do is have as much local data as possible and minimize the amount of shared state between all of the different processors. That's a description of about how some engineer at Google who is thinking about how they do search, but it applies to organizations just as much as describing very, very large computing problems.
I actually think that it's testable, and we'll find out. I think our structure will work better than most of the older command and control type hierarchical systems that require a huge amount of shared state between everybody, and move very ponderously, and have to throw away huge amounts of data because otherwise the person at the top can't possibly know what every single person at the organization is doing. To me it always seems like a lot of those organizations, the person at the top seems incredibly ignorant of the actual experiences of their customers. And it's really that if everything has to go through that one person they have to trivialize and simplify everything down to stuff that is really sort of a caricature of what is really going on in their organization or for their customers.”
Казалось бы, при чём тут Путин.