@avibryant @kamonteam But that's not an "interesting" problem area for measuring latency percentiles…
-
-
Replying to @giltene
@giltene@kamonteam yep. I'm coming from a machine learning context, where being able to eg find that "DC offset" on the fly is more useful.1 reply 0 retweets 0 likes -
Replying to @avibryant
@avibryant@kamonteam In this case it's trivial at the start: E.g. take the average of the first 100 results in the stream before feeding.2 replies 0 retweets 0 likes -
Replying to @giltene
@giltene@kamonteam bit trickier in a distributed context; each node will have a slightly different average, then you have to reconcile.1 reply 0 retweets 0 likes -
Replying to @avibryant
@avibryant@kamonteam Reconciling isn't hard. Each node can produce their own histogram, and they are additive later.1 reply 0 retweets 0 likes -
Replying to @giltene
@giltene@kamonteam right, but if they each have a different offset it gets harder.2 replies 0 retweets 0 likes -
Replying to @avibryant
@giltene@kamonteam this isn't an academic question - I'm thinking in particular about using it in MapReduce, where the nodes have no comm.2 replies 0 retweets 0 likes -
Replying to @avibryant
@avibryant@kamonteam You just need to communicate the offset along with each histogram, and adjust for it when you add.1 reply 0 retweets 0 likes -
Replying to @giltene
@giltene@kamonteam there's some devil in the details there, because the bins won't line up nicely.1 reply 0 retweets 0 likes -
Replying to @avibryant
@avibryant@kamonteam no bins (in the regular sense) to worry about. It's will all good the within the integer accuracy level you choose.1 reply 0 retweets 0 likes
@giltene @kamonteam I guess as long as you are always adjusting values up (possibly losing res) when you add, it should work out, yeah.
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.