I wrote a new blog post, about monitoring, and when averages are useful: http://brooker.co.za/blog/2017/12/28/mean.html …
-
Show this thread
-
Replying to @MarcJBrooker
Temperature is probably the most successful statistical measurement ever, used and understood many times a day by almost everyone in many contexts, and it's a mean average!
1 reply 0 retweets 1 like -
Replying to @colmmacc @MarcJBrooker
We still have a lot to learn from temperature too: using a mean makes it easy to measure, and we use scales (C/K,F) that are tolerant to a "close enough" precision. A lot of engineering systems could benefit from a similar measure. We just have to invent them.
1 reply 1 retweet 3 likes -
Replying to @colmmacc @MarcJBrooker
How much planning impact do the higher percentile temperature measurements? Temperature variance isn’t usually at a level that it impacts behavior (although given climate change that’s a very iffy statement)
1 reply 0 retweets 0 likes -
Replying to @mndoci @MarcJBrooker
Standing in my kitchen I can know the room, oven, kettle, fridge and freezer temperatures separately and easily. Variance is there, it's just been systemized and compartmentalized very successfully, and we know to measure each mean separately.
1 reply 0 retweets 0 likes -
Replying to @colmmacc @MarcJBrooker
Now if only we could regulate service behavior the way we can regulate temperature for appliances and homes. I suppose what you are saying is that in the computational/service world, we haven’t done the systemization/compartmentalization
1 reply 0 retweets 0 likes -
Replying to @mndoci @MarcJBrooker
Yep! A lot of robust designs are very simple and low temperature at the outermost layers, and then increase in complexity and entropy towards their core. Problems often manifest as an inversion of those states.
1 reply 0 retweets 0 likes -
If we measured dynamic Kolmogorov complexity, in terms of the the number of branches actually executed in running code, I have a hunch that we'd identify "hot" and "cool" systems more easily and learn to insulate more organically.
1 reply 0 retweets 0 likes
We'd also see big changes (deployments, new features) in behavior as a big temperature change, naturally associate it with thermal stress, and so manage it appropriately; keep the cooling on standby.
-
-
I like this idea of building systems that are easily monitorable, not only because they can be measured, but because their emergent properties can be understood with a relatively simple mental model.
0 replies 0 retweets 0 likesThanks. Twitter will use this to make your timeline better. UndoUndo
-
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.