isn't this the entire goal of distributed computation? (to avoid needing to look at the entire system at once)
This Tweet was deleted by the Tweet author. Learn more
Conversation
Replying to
That's a really good question. I would say yes, although in practice complex distributed designs often don't design or handle for all the failure scenarios.
Eg: cascading performance problems, related failure domains, etc...
Elegant designs are always best!
2
1
(Remember that CAP theorem for example, doesn't guarantee you'll never be down... In fact, if you *roughly* fail nodes at a rate N that is larger than the rate M that you add new ones, then eventually you'll be definitely down!)
1
Replying to
That is true, but the failure of one node shouldn't *cause* the failure of ten more (necessarily) right?
2
Not that I have any insight into this particular failure