'The History of Metrics According to me' by Stephen Day

Post on 16-Apr-2017

5031 Views

Category:

Technology

2 Downloads

Preview:

Click to see full reader

Transcript

The History of Metrics According to Me

Stephen DayDocker, Inc.Docker Meetup, SFDecember 2016

v1

Stephen DayDocker, Inc.stephen@docker.comgithub.com/stevvooe@stevvooe

MetricsWhy?

The Early Years

5

Titration

By UCL - Flickr, CC BY 2.0, https://commons.wikimedia.org/w/index.php?curid=45263025

Music

Electrical Engineering

CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=392484

Microcontrollers

Human Powered Submarine

Graphs

Software Engineering

Software Engineering is way behind!Sort of.

See http://www.itl.nist.gov/div898/handbook/ for an example

13

Observability and ControllabilityThe Problem

Low Observability High Observability

Guessing Informed

Models

INFO[0012] response completed

go.version=go1.6.3

http.request.host=localhost:5001

http.request.id=4db26096-a417-48e8-a759-ddd8ea66c3a0

http.request.method=GET

http.request.remoteaddr=172.17.0.1:57406

http.request.uri=/v2/

http.request.useragent=curl/7.49.1

http.response.contenttype=application/json; charset=utf-8

http.response.duration=5.4388ms

http.response.status=200

http.response.written=2

instance.id=7926875b-d3fb-41e4-bd75-2370a3e8a03c version=v2.5.1

Record-based

ModelsSample-based

t

Why aren’t metrics built into everything?

Excuses!

Why aren’t metrics a part of Docker?

The Goal

Prometheus

https://prometheus.io/

Types of Metrics

- Counters: cumulative number that only increases

- Gauges: value that can go up or down- Summary/Histogram: Sample of observations

https://prometheus.io/docs/concepts/metric_types/

Format

Configuration

Architecture

Demo

What’s next?

So much.

Road Maphttps://github.com/docker/docker/issues/27307

- For Contributors and Maintainers to better understand performance- For Users to better understand behavior of the daemon in their

infrastructure- Implemented in 1.13

Docker Engine Metrics

Externally Observable Metrics

- CPU, memory usage, etc.- Replaces docker stats

Integrated Target Discovery

- Prometheus automatically discover engine and container metric targets

Application Metrics

- Directly proxy your metrics out of the Docker daemon

- Leverage built in integrated target discovery

THANK YOU

top related