Prof. Dr. Nane Kratzke What to do if your cloud is burning? Well, be prepared ...
ESCAPE ROUTE (aka Agenda)
2
Burning cloud? What does that mean?
How long is your escape route in cloud computing?
Who takes care for escape routes in cloud computing?
How can escape routes look like in cloud computing?
Why to take the risk?
Cloud TRANSIT (a research project to get you out of a cloud)
8
• There are a lot ofapproaches to get intoa cloud.
• But almost no(pragmatic) approaches exist toleave a cloud or movebetween clouds.
• But: If you know howto get out, you aremore willing to take therisk to go in.
What does it mean? My cloud is burning ...
Prof. Dr. rer. nat. Nane KratzkePraktische Informatik und betriebliche Informationssysteme 9
• Provider is insolvent ...
• Provider rises prices ...
• Provider reduces resource limits ...
• Provider terminates your contract ...
• Provider has availability problems ...
• Changing laws (data protection) ...
• Other governance/compliance reasons (data hosted on US territory, NSA?)
There are a lot of (hardly predictable) reasons toleave a cloud service provider.
ESCAPE ROUTE (Agenda)
10
Burning cloud? What does that mean?
How long is your escape route in cloud computing?
Who takes care for escape routes in cloud computing?
How can escape routes look like in cloud computing?
Why to take the risk?
Example: Instagram
Prof. Dr. rer. nat. Nane KratzkePraktische Informatik und betriebliche Informationssysteme 11
• Worldwide social network for image sharing
• 20 employees
• Hosted by Amazon Web Services
• Net asset value of 1 Bill. USD (that paid Facebook)
• No noteworthy IT assets or datacenters (just 20 laptops)
YearsIt took only
Example: Instagram
Prof. Dr. rer. nat. Nane KratzkePraktische Informatik und betriebliche Informationssysteme 12
• Approximately 1 year for analysis and
• development of toolings (especially IP Collision Handling)
• About 4 to 8 weeks for all migration steps (inlcuding severe outages)
Question: How long does it take to transfer all Instagram services anddata into Facebook datacenters?
???
This was no ad-hoc transfer! This was a major project.
ESCAPE ROUTE (Agenda)
14
Burning cloud? What does that mean?
How long is your escape route in cloud computing?
Who takes care for escape routes in cloud computing?
How can escape routes look like in cloud computing?
Why to take the risk?
Did you know ...More than 95% of all enterprises are small enterprises?
Prof. Dr. rer. nat. Nane KratzkePraktische Informatik und betriebliche Informationssysteme 15
micro enterprisessmall enterprisesmedium enterpriseslarge enterprises
Category Employees Turnover
Micro enterprises < 10 < 2 Mio. €
Small enterprises < 50 < 10 Mio. €
Medium enterprises < 250 < 50 Mio. €
Large enterprises >= 250 >= 50 Mio. €
Distribution of ICT enterprises in the European Union (2014), EUSTAT
Current Cloud Computing Research ...
Prof. Dr. rer. nat. Nane KratzkeComputer Science and Business Information Systems 16
• Has often implicit assumptions:
• Arbitrary companies with
• large IT-staffs providing the capability to handle
• arbitrary complexity of tools and methods.
• These “Super Tankers“ do not have to be afraidof inconviences like vendor lock-in. They are bigenough to solve the problem ...
Our target group is different ...
Prof. Dr. rer. nat. Nane KratzkeComputer Science and Business Information Systems 17
• Small sailing boat vs. Supertanker (weather)
• Small and medium sized enterprises (SME)
• 1 person IT-staffs
• Public and private cloud computing
• We analyze
• Container technologies (like Docker)
• Container cluster (like Kubernetes, Swarm, Mesos)
According to that ....
Cloud fire protection for(not just) smallenterprises (that means95% of all enterprises) looks like that ...
Prof. Dr. rer. nat. Nane KratzkeComputer Science and Business Information Systems 18
ESCAPE ROUTE (Agenda)
19
Burning cloud? What does that mean?
How long is your escape route in cloud computing?
Who takes care for escape routes in cloud computing?
How can escape routes look like in cloud computing?
Why to take the risk?
Good News ...
Prof. Dr. rer. nat. Nane KratzkePraktische Informatik und betriebliche Informationssysteme 20
2006
2 cloud servicesreflected by cloud
standards
2016
11 cloud servicesreflected by cloud
standards
5 times more standardization than 10 years before !!!
Example:
But ...
Prof. Dr. rer. nat. Nane KratzkePraktische Informatik und betriebliche Informationssysteme 21
2 2
2 4 6 77
7 7 11 11
1 1
2 4 7 1014
21 26 42 44
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Relationofconsid
eredservices
consideredbyCIMI,OCCI,CDMI,OVF,OCI,TOSCA notconsidered
The relation of services reflected by cloudstandards to cloud services not reflected bycloud standards decreased over the last 10 years!
Analyzed using over 2300 offical release notes of Amazon Web Services (AWS). Data for other providers like Google, Azure, Rackspace, etc. not presented. Basic conclusions for theseproviders are the same.
Cloud-native applicationsare vulnerable for vendorlock-in. That is especially
true for SMEs.
Cloud-native Applications
Cloud native applications are often characterized bya highly implicit technological dependency on
hosting cloud infrastructures. The project Cloud TRANSIT investigates how to design cloud-native applications and services to reduce technologicaldependencies on underlying cloud infrastructures.
DEFINITION: A cloud-native application is a (micro)service-based, elastic and horizontal scalable application where each self-containeddeployment unit of that application is designedaccording to cloud-focused software design patternsand operated on a self-service agile elastic platform.
The Cloud-Native Reference Model (ClouNS)
Prof. Dr. rer. nat. Nane KratzkeComputer Science and Business Information Systems 23
Popular Container-based Cluster Platforms ...
Prof. Dr. rer. nat. Nane KratzkeComputer Science and Business Information Systems 24
Docker SwarmSwarm Mode (sinceDocker 1.12) Clones
Kubernetes-like controlprocesses but integrates
them in just onecomponent. Secure by
default (control and dataplane). Hides operation
complexity.
GoogleControl processes that
continuously drive current stateof container based applicationstowards a defined desired state. Makes Google‘s experience ofrunning large scale productionworkloads available as open
source.
MesosphereApache Mesos based
datacenter operating systemfor fine grained resource
allocation. Frameworks tooperate containers and data
services. Datacenter focused. Mesos operates successfullylarge scale datacenters since
years (Twitter, Netflix, ...)
Practitioners ask for simple solutions (elastic platforms) ...
Avoid Vendor Lock-In using alreadyexisting Container-Technologies
Prof. Dr. rer. nat. Nane KratzkePraktische Informatik und betriebliche Informationssysteme 25
Operate application on current provider.
Scale cluster into prospective provider.
Shutdown nodes on current provider. Cluster reschedules lost container.
Migration finished.
Pets
Cattle
It is all about pets vs. cattle!http://www.theregister.co.uk/2013/03/18/servers_pets_or_cattle_cern/
ESCAPE ROUTE (Agenda)
26
Burning cloud? What does that mean?
How long is your escape route in cloud computing?
Who takes care for escape routes in cloud computing?
How can escape routes look like in cloud computing?
Why to take the risk?
Prof. Dr. rer. nat. Nane KratzkePraktische Informatik und betriebliche Informationssysteme 27
Kostenassoziativät
New Business Models
e.g. cost associativity
e.g. unpredictable workloads
Berkley View of Cloud Computing, 2009: Cost associativity in Cloud Computing
28
It cost the same to operate ...
... 720 machines
for one hour
or one machine for
720 hours.
We are afraid of peak loads, but why?
29
„In other words, even if cloud services cost, say, twice as much, a pure cloud solution makes sense forthose demand curves where the peak-to-average ratiois two-to-one or higher.“Weinman, Mathematical Proof of the Inevitability of Cloud Computing, 2011
http://www.joeweinman.com/Resources/Joe_Weinman_Inevitability_Of_Cloud.pdf
Analyzed use case
• Web technology lecture/practical course forcomputer science students (bachelor) in summer2011 and summer/winter 2012.
• Projects: Development of web informationsystems (Drupal based)
• All groups were assigned cloud service accountsprovided by Amazon Web Services (AWS).
• Analysis of billing as well as usage data providedby AWS.
Prof. Dr. rer. nat. Nane KratzkeComputer Science and Business Information Systems 30
Usage Analysis
31Prof. Dr. rer. nat. Nane KratzkeComputer Science and Business Information Systems
13 14 15 16 17 18 19 20 21 22 23 24 25
Average Box UsageMaximum Box Usage in an hour
(A)Maximum and Average Box Usage
Calendar Week
Use
d S
erve
r Box
es
010
2030
4050
13 14 15 16 17 18 19 20 21 22 23 24 25
(B)Accumulated Processing Hours per Week
Calendar Week
Pro
cess
ing
Hou
rs
0500
1000
1500
2000
14 16 18 20 22 24
0.0
0.2
0.4
0.6
0.8
1.0
(C)Average Box to Maximum Box Ratio
according to Weinman
Calendar Week
Avg
to M
ax B
ox U
sage
Rat
io
Training
Project 24x7 Migration
Average to Peak Ratio per week
32Prof. Dr. rer. nat. Nane KratzkeComputer Science and Business Information Systems
13 14 15 16 17 18 19 20 21 22 23 24 25
Average Box UsageMaximum Box Usage in an hour
(A)Maximum and Average Box Usage
Calendar Week
Use
d S
erve
r Box
es
010
2030
4050
13 14 15 16 17 18 19 20 21 22 23 24 25
(B)Accumulated Processing Hours per Week
Calendar WeekP
roce
ssin
g H
ours
0500
1000
1500
2000
14 16 18 20 22 24
0.0
0.2
0.4
0.6
0.8
1.0(C)
Average Box to Maximum Box Ratioaccording to Weinman
Calendar Week
Avg
to M
ax B
ox U
sage
Rat
io
Cloud computing iseconomical reasonable
Cloud computingmight be reasonable
Cloud computing iseconomical not reasonable
Economical Decision AnalysisA four step process to decide for or against cloud based virtual labs
Prof. Dr. rer. nat. Nane KratzkeComputer Science and Business Information Systems 33
A cloud based solution provides a morethan 25 times cost advantage.
The measured ATP ratio of 0.035 means in fact a 1/0.035 == 28.57 times cost advantage.
This means for the presented use case:
Compared to necessary investment efforts for a classicaldedicated system implementation.
Why this big cost advantage?
Prof. Dr. rer. nat. Nane KratzkeComputer Science and Business Information Systems 34
13 14 15 16 17 18 19 20 21 22 23 24 25
Average Box UsageMaximum Box Usage in an hour
(A)Maximum and Average Box Usage
Calendar Week
Use
d S
erve
r Box
es
010
2030
4050
13 14 15 16 17 18 19 20 21 22 23 24 25
(B)Accumulated Processing Hours per Week
Calendar Week
Pro
cess
ing
Hou
rs
0500
1000
1500
2000
14 16 18 20 22 24
0.0
0.2
0.4
0.6
0.8
1.0
(C)Average Box to Maximum Box Ratio
according to Weinman
Calendar Week
Avg
to M
ax B
ox U
sage
Rat
io
How to dimensionize the data center? Hmm, peak load ...peak load
averageload
And the delta?
Measures the overdimension of a data center
ESCAPE ROUTE (Agenda)
35
Burning cloud? What does that mean?
How long is your escape route in cloud computing?
Who takes care for escape routes in cloud computing?
How can escape routes look like in cloud computing?
Why to take the risk?
Summary
• You want to adopt cloud computing?
• Think about your escape strategy FIRST!
• Support research focussing small andmedium sized enterprises (it does not costsooo much)
• That supports 95% of all enterprises
• (and not only 5% supertankers)
• New (maybe disruptive?) business models ...
• Cost associativity
• Cost advantages for non-static ofunpredictable workloads
Acknowledgement
• All Pictures taken from Pixabay.com (CC0 Licence)
Prof. Dr. rer. nat. Nane KratzkeComputer Science and Business Information Systems 37
Our research is funded by German Federal Ministry of Education and Research
(Project Cloud TRANSIT, 03FH021PX4). We thank fat IT solution GmbH (Kiel)
for their support of Cloud TRANSIT.
Picture ReferencePresentation URL
About
Prof. Dr. rer. nat. Nane KratzkeComputer Science and Business Information Systems 38
CoSA: https://cosa.fh-luebeck.de/en/contact/people/n-kratzke
Blog: http://www.nkode.io
Twitter: @NaneKratzke
GooglePlus: +NaneKratzke
LinkedIn: https://de.linkedin.com/in/nanekratzke
GitHub: https://github.com/nkratzke
ResearchGate: https://www.researchgate.net/profile/Nane_Kratzke
SlideShare: http://de.slideshare.net/i21aneka
Prof. Dr. rer. nat. Nane Kratzke