A good answer is better high-level insights, aberration detection, etc. But I wonder if we could invent a single, scalar, measure of dynamism in large systems, akin to temperature in thermodynamic systems. 2/n
-
-
Show this thread
-
Imagine being able to look at a large system under stress; an overload event, or a capacity loss, and assign it a temperature. Or at an inactive, or constantly performant system, and see that it's cool. Imagine being able to see temperature increase during a deployment. 3/n
Show this thread -
That would be so neat, because it would give people a macro signal of when to be careful and when to apply cooling. It seems like it would be crazy and heuristically, but I wonder if it could be defined statistically based on state exchanges. 4/n
Show this thread -
This is, after all, how temperature is derived from molecular momentum. If such a measure could be invented, it would be a huge breakthrough. Anyway, that's my thought. 5/5
Show this thread
End of conversation
New conversation -
-
-
I have one boring answer to this myself: I want to see a standard way to publish metrics, so I can get useful metrics from third party libraries and binaries. Sidecars are an answer for latency, errors etc but not internal metrics.
-
Have you seen
@opentelemetry (formerly@opencensusio &@opentracing)? An SDK for creating metrics, and a collection of exporters to destinations like Prometheus, StackDriver, DataDog, Cloud Watch. Try the quick start: https://opencensus.io/introduction/ - 4 more replies
New conversation -
-
-
In my opinion, the biggest challenge is to enable observability into systems with minimal effort by developers, yet provide end to end problem detection, tracing, proactive actions while providing the right level of detail with an option to dig in as needed
#distributedsystemsThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
infrastructure. Being able to ask questions to your infra. Especially networking. Also dynamic tracing live integrated in platforms like the BEAM. Having historical traces is good. Being able to ask ad-hoc questions live is better.
-
Distributed tracing on demand is amazingly useful.
- 1 more reply
New conversation -
-
-
...that's a little creepy, actually, I have been talking and thinking about a temperature based approach to ambient awareness for ~2 years now, but only just recently started prototyping.
- 2 more replies
New conversation -
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.