Monitoring Challenges - Monitorama 2016 - Monitoringless

Monitoring Challenges

Monitorama June 2016 Adrian Cockcroft

@adrianco

What does @adrianco do?

@adrianco

Technology Due Diligence on

Deals

Presentations at Companies and

Conferences

Tech and Board Advisor

Support for Portfolio

Companies

Consulting and Training

Networking with Interesting PeopleTinkering with

Technologies

Vendor Relationships

Previously: Netflix, eBay, Sun Microsystems, Cambridge Consultants, City University London - BSc Applied Physics

Monitorama 2014…

Monitorama 2016

What problems does monitoring address? Why isn’t this a solved problem already?

Who gets disrupted by what? Stuff I’ve been tinkering with

Measuring business value

Problem detection and diagnosis

“Ultimately business value is what the business values, and that is that.”

Mark Schwartz CIO DHS/DCIS

Business Value of Monitoring

Customer happiness Cost efficiency

Safety and security Compliance

Business Value of Monitoring

Customer happiness Cost efficiency

Safety and security Compliance

Customer Happiness

Time to value Availability

Response time

Cost Efficiency

Utilization Optimization Automation

Why isn’t this a solved problem already?

Why isn’t there one standard for monitoring?

Why isn’t there one standard for monitoring?We tried that once, immediately obsoleted by rise of Windows NT

X/Open Universal Measurement Architecture - 1997 http://pubs.opengroup.org/onlinepubs/009657299/c427-1/front.htm

http://pubs.opengroup.org/onlinepubs/009657299/c427-1/front.htm

Monitoring Evolution Challenges

Platform - Entities - Hierarchy Interfaces - Metrics - Schema

Scale - Ephemerality

Different vendors and tools in each generation…

1970’s Mainframes





1970’s Mainframes

1980’s Minicomputers





1990’s Unix Servers

1970’s Mainframes







1970’s Mainframes

2000’s Windows on x86







1970’s Mainframes



2000’s Linux on x86






1970’s Mainframes




2000’s VMware on blades






1970’s Mainframes





2010’s Public cloud






1970’s Mainframes





2010’s Public cloud2010’s Containers






1970’s Mainframes





2010’s Public cloud2010’s Containers

2010’s Serverless





Why don’t monitoring vendors adapt and survive?

Cost per node drops Revenue opportunity decreases

Waves of disruption

New vendors have new schema’s, an order of

magnitude lower cost per node, and many more shorter lived

nodes to monitor

$Millions (illustrative order of magnitude costs)


Waves of disruption



nodes to monitor


$1M


Waves of disruption



nodes to monitor

$100K


$1M


Waves of disruption



nodes to monitor

$100K


$10K

$1M


Waves of disruption



nodes to monitor

$100K


$10K

$1M

$5K


Waves of disruption



nodes to monitor

$100K


$10K

$1M

$5K

$1K per core


Waves of disruption



nodes to monitor

$100K


$10K

$1M

$5K

$1K per core

$100’s per month


Waves of disruption



nodes to monitor

$100K


$10K

$1M

$5K

$1K per core

$100’s per month$10’s per month


Waves of disruption



nodes to monitor

$100K


$10K

$1M

$5K

$1K per core

$100’s per month$10’s per month

$1’s per month


Waves of disruption



nodes to monitor

Vendor Landscape

A Tragic Quadrant

Ability to scale

Ability to handle rapidly changing microservices

In-house tools at web scale companies

Most current monitoring & APM

tools

Next generation APM

Next generation Monitoring

Datacenter

Cloud

Containers

100s 1,000s 10,000s 100,000s

Lambda

A Tragic Quadrant

Ability to scale

Ability to handle rapidly changing microservices

In-house tools at web scale companies

Most current monitoring & APM

tools

Next generation APM

Next generation Monitoring

Datacenter

Cloud

Containers

100s 1,000s 10,000s 100,000s

Lambda

Vendors - tell me where you belong on this plot…

Tinkering

Simulated MicroservicesModel and visualize microservices Simulate interesting architectures Generate large scale configurations Stress test real monitoring tools

Code: github.com/adrianco/spigo Simulate Protocol Interactions in Go Simian Army Visualizations

ELB Load Balancer

ZuulAPI Proxy

KaryonBusiness Logic

StaashData Access Layer

PriamCassandra Datastore

ThreeAvailabilityZones

DenominatorDNS Endpoint

http://github.com/adrianco/spigo

Zipkin Trace for one Spigo Flow

Response Times

See http://www.getguesstimate.com/models/1307 Guesstimate

http://www.getguesstimate.com/models/1307

memcached hit %

memcached response mysql response

service cpu time

memcached hit mode

mysql cache hit mode

mysql disk access mode

Hit rates: memcached 40% mysql 70%Guesstimate

Spigo Histogram Results name: storage.*.*..load00...load.denominator_serv quantiles: [{50 47103} {99 139263}] From To Count Prob Bar 20480 21503 2 0.0007 : 21504 22527 2 0.0007 | 23552 24575 1 0.0003 : 24576 25599 5 0.0017 | 25600 26623 5 0.0017 | 26624 27647 1 0.0003 | 27648 28671 3 0.0010 | 28672 29695 5 0.0017 | 29696 30719 127 0.0421 |#### 30720 31743 126 0.0418 |#### 31744 32767 74 0.0246 |## 32768 34815 281 0.0932 |######### 34816 36863 201 0.0667 |###### 36864 38911 156 0.0518 |##### 38912 40959 185 0.0614 |###### 40960 43007 147 0.0488 |#### 43008 45055 161 0.0534 |##### 45056 47103 125 0.0415 |#### 47104 49151 135 0.0448 |#### 49152 51199 99 0.0328 |### 51200 53247 82 0.0272 |## 53248 55295 77 0.0255 |## 55296 57343 66 0.0219 |## 57344 59391 54 0.0179 |# 59392 61439 37 0.0123 |# 61440 63487 45 0.0149 |# 63488 65535 33 0.0109 |# 65536 69631 63 0.0209 |## 69632 73727 98 0.0325 |### 73728 77823 92 0.0305 |### 77824 81919 112 0.0372 |### 81920 86015 88 0.0292 |## 86016 90111 55 0.0182 |# 90112 94207 38 0.0126 |# 94208 98303 51 0.0169 |# 98304 102399 32 0.0106 |# 102400 106495 35 0.0116 |# 106496 110591 17 0.0056 | 110592 114687 19 0.0063 | 114688 118783 18 0.0060 | 118784 122879 6 0.0020 | 122880 126975 8 0.0027 |

Normalized probability

Response time distribution measured in nanoseconds using High Dynamic Range Histogram

:# Zero counts skipped|# Contiguous buckets

Median and 99th percentile values

service time for load generator

Cache hit Cache miss

Serverless

Serverless AWS Lambda - lots of production examples

Google Cloud Functions Azure Functions alpha launched

IBM OpenWhisk - open source

Startup activity: iron.io , serverless.com, apex.run toolkit

http://iron.io

http://serverless.com

http://apex.run

Monitorless ArchitectureAPI Gateway

Kinesis S3DynamoDB


Kinesis S3DynamoDB


Kinesis S3DynamoDB

Monitorable entities only exist during an execution trace

AWS Lambda Reference Arch http://www.allthingsdistributed.com/2016/05/aws-lambda-serverless-reference-architectures.html

http://www.allthingsdistributed.com/2016/05/aws-lambda-serverless-reference-architectures.html

Serverless Programming Model

Event driven functions Role based permissions

Whitelisted API based security Good for simple single threaded code

Serverless Cost Efficiencies

100% useful work, no agents, overheads 100% utilization, no charge between requests

No need for extra capacity for peak traffic Anecdotal costs ~1% of conventional system

Ideal for low traffic, Corp IT, spiky workloads

Serverless Work in Progress

Tooling for ease of use Multi-region HA/DR patterns

Debugging and testing frameworks Monitoring, end to end tracing

Using AWS Lambda to monitor AWS

DIY On-Premise Serverless Operating Challenges

Scheduling and startup latency Execution and monitoring overhead

Charging model Capacity planning

Monitoring Challenges

Too much new stuff Too ephemeral

Price disruption

Thanks!

Thanks!

Also speaking at: Docker Portland Meetup Wednesday Evening @Puppetlabs - Microservices: Whats Missing

Security

Visit http://www.battery.com/our-companies/ for a full list of all portfolio companies in which all Battery Funds have invested.

Palo Alto Networks

Enterprise ITOperations & ManagementBig DataCompute

Networking

Storage

http://www.battery.com/our-companies/

Monitoring Challenges - Monitorama 2016 - Monitoringless

Software