Monitor and Manage Microservices

Preview:

Citation preview

Monitor and Manage Microservices

Instana, Inc. Proprietary and Confidential 3

Instana: Overview

• Founded March 2015, San Mateo

• Engineering in Germany

• $6M Series A led by Target Partners

• Private Beta launched end of 2015

• Gartner Cool Vendor 2016

Mirko Novakovic, CEO

• Founded codecentric and grew to > 350 consultants

• Founded CenterDevice SAAS service with > 6500 customers

Instana, Inc. Proprietary and Confidential 4

1990 1995 2005 2010 2015

Client/Server Linux Open Source

Web 3-Tier

Java EE

SOA Web 2.0

OSS Middleware

Cloud Big Data

µService Containers

Reactive Serverless

Tivoli BMC CA

Nagios

CA/Wily Dell/Quest

Precise HP/Mercury

dynatrace AppDynamics New Relic

APM 2.0APM 1.0

= Fundamental Architecture Shift

Next Generation

APM 3.0

2000

Microservices need new monitoring

Instana, Inc. Proprietary and Confidential 5

First generation - Web & Physical Datacenter

AppServer

Database

Agent

Deploy in month Lives for years

Instana, Inc. Proprietary and Confidential 6

Database

Agent

Second Generation - SOA and Virtual Datacenters/Cloud

Deploy in minutes Lives for months

Instana, Inc. Proprietary and Confidential 7

Next generation: µServices

Database

Agent

Instana, Inc. Proprietary and Confidential 8

Next generation: µServices

Database

Agent

8

Deploy in Seconds Lives for Minutes

Host Container Cluster

Changes

Middleware Services

Agile & Continuous Delivery

Deployment Resize

Restart Reallocate

Start Stop

Reconfigure

Redeploy Reconfigure

Scale

New Code Versions

Deployment

Host Container Middleware Cluster

CPU high Load to high GC Overhead (JVM) Re-Balancing

Issues

?

Problems occur at Any Layer

Code Exceptions/Errors

?

Services

Instana, Inc. Proprietary and Confidential 11

Need for Artificial Intelligence

„Humans cannot continuously monitor

the status of all of our systems.“

Instana, Inc. Proprietary and Confidential 12

Monitoring: A Tragic Quadrant

Next generation APM

In-house tools at web scale

companies

Most current monitoring & APM tools

Next generation Monitoring

100s 1.000s 10.000s 100.000s

Datacenter

Cloud

Containers

Lambda

Visual: Adrian Cockcraft - A Tragic Quadrant

Ability to scale

Ability to handle rapidly changing micro services

Instana, Inc. Proprietary and Confidential 13

µServices and WebScale architectures: https://github.com/adrianco/spigo

Instana, Inc. Proprietary and Confidential 14

µServices and WebScale architectures: https://github.com/adrianco/spigo

Instana, Inc. Proprietary and Confidential 15

µServices - Two Types of Graphs

• Graph of Transactions

• End-User

• Processes

• Graph of components

• Physical

• Logical

Instana, Inc. Proprietary and Confidential 16

Solution

Instana, Inc. Proprietary and Confidential 17

Logging - e.g. ELK Stack

Instana, Inc. Proprietary and Confidential 18

Timeseries Data Stores

Instana, Inc. Proprietary and Confidential 19

Metric Frameworks - Codahale

Instana, Inc. Proprietary and Confidential 20

Visual Engineering - Netflix Visceral

Instana, Inc. Proprietary and Confidential 21

Visual Engineering - Instana

Instana, Inc. Proprietary and Confidential 22

Visual Engineering - Instana

Instana, Inc. Proprietary and Confidential 23

Transaction - µService Up- and Downstreams

Instana, Inc. Proprietary and Confidential 24

Transaction - Flamegraph

Instana, Inc. Proprietary and Confidential 25

Transaction - Google Dapper Paper

Instana, Inc. Proprietary and Confidential 26

Transaction - Open Zipkin - https://github.com/openzipkin

Instana, Inc. Proprietary and Confidential 27

Transaction - Open Tracing

Instana, Inc. Proprietary and Confidential 28

Workaround: Logging + Metric + Tracing + Dashboard + Analytics

Lot of continuous development

effort

Lot’s of data - difficult to extract

information

Post-mortem analysis Massive CPU

and network overhead

Expensive

Operational survival Undifferentiated Heavy Lifting

Instana, Inc. Proprietary and Confidential 29

µServices - Two Types of Graphs -> Semantical Monitoring

• Graph of Transactions

• End-User

• Processes

• Graph of components

• Physical

• Logical

The Dynamic Graph, an Example

example

Search Product Trace

Index A

ES Cluster

Spring Boot

JVM

Process

Container

Host

ES Node

JVM

Process

Container

Host

ES Node

JVM

Process

Container

Host

ES Node

JVM

Process

Container

Host

ES Node

JVM

Process

Container

Host

Zone

Zone

App A

a business functionality

Curated Expert Knowledge:

Quality of Service - Automatic Management of Component Health

Automatic Service Discovery

‣ automatic and continuos discovery of Service Architecture

‣ health management KPI’s ‣ Throughput ‣ Latency ‣ Error Rate ‣ Saturation

‣ machine learning algorithms to learn performance patterns

Service Layer "above" components:

Quality of Service - Management by Incident

Automatically raise Incidents on breach of KPI’s:group changes and issues correlated by the dynamic graph

product search

What was the situation?

TimeShift:

Investigate your history of:

Architecture, Incidents, Issues, Changes, Health and Metrics.

Precisely explore the past

Distributed Tracing for Web-Scale

Tracing:

Zero Application Impact

No configuration

Error Detection

Asynch and Batch

Real Code visibility

Correlation to Services and Components

Open SDK

End-to-End Tracing build for the µService World!

Thank You!

Recommended