Recent searches
Search options
the sort of thing I wish more books stated is what ends up being the sort of unspoken thing that a lot of #DistributedSystems veterans do, which is: re-frame your system a different way until you do not need the fancy-ass distsys theory to make it work and it just survives common issues by being the right shape.
It is generally easier to alter your roadmap than it is to engineer a novel distributed system that works the way you need it to.
@mononcqc i recall a boss once saying "so what if things are stateless? we can make them stateful to take advantage of our process migration system"
@tef ouch. Though I guess if I want to give the most benefit of the doubt, the next best thing after not having a problem is having a problem you know how to solve comfortably :/
@mononcqc oh no doubt here, it was "my magic tech is amazing and we'll poison people to sell the antidote"
@tef it’s funny how the village’s window maker has such an extensive collection of rocks just the right size to be thrown around sometimes
@mononcqc but think of all the elaborate cache invalidation schemes you won't get to design* that way!
* and debug, forever
A related tool is to re-work your system until you can use another platform or data store to solve the problem for you.
E.g., “ok, this other team provides Zookeeper as a service, can we make sure all the data that needs strong consistency goes there?”
@ajdecon yeah if you gotta have problems it’s nicer to choose problems you’re already equipped to solve
@mononcqc any recommendations for existing books that do this?
@vincentlu unfortunately not. Either I don’t know about them or they need to be written at some point.
@mononcqc @vincentlu I mean... any talk by Kyle Kingsbury should count no? Or Jepsen report :D