@kellabyte Exactly. Or expect non-upgraded servers to simply continue old behavior and/or have clients check causality in the interim.
-
-
Replying to @kellabyte
@kellabyte Exactly! There are major opportunities for scalability by incorporating application semantics.1 reply 1 retweet 1 like -
Replying to @pbailis
@kellabyte In a sense, this is what NoSQL is about. The next step is to ask: if I *also* want to *guarantee* correctness, what's the API?2 replies 0 retweets 0 likes -
Replying to @kellabyte
@kellabyte Well, we used "fence" in the sense of "everyone should observe this write before each of the next writes I perform"1 reply 0 retweets 0 likes -
Replying to @pbailis
@kellabyte And, in multi-processor cache coherency, "fence" is used as similar explicit signal for synchronization.1 reply 1 retweet 1 like -
Replying to @pbailis
@kellabyte More on this here: http://en.wikipedia.org/wiki/Memory_barrier …; cache coherency is just about as confusing as distributed consistency!2 replies 2 retweets 2 likes -
Replying to @pbailis
@pbailis@kellabyte Modern multicores are distributed systems except partial failure will likely panic your OS not 'completed maybe' on you!2 replies 0 retweets 0 likes -
Replying to @darachennis
@darachennis@kellabyte Though "dark silicon" and projects like BubbleWrap (http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.175.2356&rep=rep1&type=pdf …) will change this!1 reply 1 retweet 1 like
@pbailis @kellabyte Disable power mgmt. Pop the bubble trouble. Win consistent latency. Blow bubbles for throughput... Nuance, trade off! :/
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.