VIRTUAL SERVER SELF-SERVICE PROVISIONING

Juraj SucikSystem Architect

CERN

CERN

What means « »?

European

Organization for

Nuclear

Research

European

Council for

Nuclear

Research

CERNCERN

Conseil

Européen pour la

Recherche

Nucléaire

1952

1954

Organisation

Européenne pour la

Recherche

Nucléaire

European Laboratoryfor Particle Physics

The largest particle physics lab in the world

Twenty Member StatesAustria, Belgium, Bulgaria, Czech Republic, Denmark, Finland, France, Germany, Greece, Italy, Hungary, Netherlands, Norway, Poland,Portugal, Slovakia, Spain, Sweden, Switzerland, United Kingdom

Annual budgetin 2007982 MCHF (610 MEUR)

External fundingfor experiments

Eight Observer StatesEuropean Commission, USA, Russian Federation,

India, Israel, Japan, Turkey, UNESCO

People2415 Staff730 Fellows and

associates200 Students9133 Users2000 External Firm

Do fundamental research

By answering questions like the structure of matter…

4th - 5th

centuryBC

End of

19th

century

Beginning of

20th

century1960s

Checking existing theories: the standard modelLEPTONS QUARKS

ORDINARYMATTER

GLUONS

Strong Force

PHOTONS

Electro-Magnetic Force

BOSONS

Weak Force

GRAVITONS

Gravity

Images: www.particlezoo.net

ELECTRONELECTRONNEUTRINO

MUONMUONNEUTRINO

TAUTAUNEUTRINO

UP DOWN

CHARM STRANGE

TOP BOTTOM

4forces

Answering fundamental questions…• How to explain particules have a mass?

Newton could not explain, neither can we…

• What is 96% of the Universe made of ?We can only see 4%of its estimated mass!

• Why isn’t there antimatterin the Universe?Nature should be symetric…

• What was the state of matter justafter the « Big Bang » ?Travelling back to the earliest instants of the Universe would help…

HiggsBoson

Bringing nations together and educate• Biggest international scientific

collaboration

• Various students programmes

• Over 100 countries

• Hundreds of physics institutes

• Half of the world’s particle physicists

By accelerating and colliding objects…

At incredible levels of energy!

E=mc2

The largest particle accelerators17 miles (27km)long tunnel

Thousands of superconducting magnets

Ultra vacuum:10x emptierthan on the Moon

Coldest placein the Universe: -271° C

In safe conditions!

The biggest and most sophisticated detectors

13

Cathedrals of science100m underground

600 million collisionsper second detectedby hundreds ofmillion sensors

Thousands of collaboratorsfor each detector

In safe conditions!

Practical applications: the World Wide WebWas developed in the frame of the LHC project in 1989!

Freely given to the World!

Practical applications: cancer treatmentFor both detection and cure of cancers

PET Scans

Hadron Therapy

Practical applications: detectorsScanning trucks in less than one hour without unloading them!

Practicalapplications:using the Grid

Ultra high-speed processingof satellite imagery in thecase of natural disasters

And of course… some Nobel prizes!

George Charpak

“for his invention and development of particle detectors, in particular the multiwire proportional chamber”

Carlo Rubbia(with Simon van der Meer)

“for their decisive contributions to the large project, which led to the discovery of the field particles W and Z, communicators of weak interaction”

GRID

IT Infrastructure at CERN

• General Purpose Computing Environment• Administrative Computing Services• Physics and engineering computing• Consolidation, coordination and standardization of computing

activities• Physics applications

(e.g., for data acquisition& offline analysis)

• Accelerator design and operations

LHC Data Every Year

• 40 million collisions per second• After filtering, 100 collisions of interest per second• > 1 Megabyte of data digitized per collision recording rate > 1 Gigabyte / sec• 1010 collisions recorded each year stored data > 15 Petabytes / year

• analysis requires a computing power equivalent to ~ 100,000 of today's fastest processors

Computing power available at CERN

• High-throughput computing based on reliable “commodity” technology

• More than 35’000 CPUs in about 6000 boxes (Linux)

• 14 Petabytes on 14’000 drives (NAS Disk storage)

• 34 Petabytes on 45’000 tape slots with 170 high speed drives

Nowhere near enough!

Computing for LHC

• Problem: even with Computer Centre upgrade, CERN can provide only a fraction of the necessary resources

• Solution: Data centers, which were isolated in the past, will be connected, uniting the computing resources of particle physicists worldwide

Users of CERN

Europe: 267 institutes4603 usersOut of Europe: 208 institutes1632 users

What is the Grid?

• The World Wide Web provides seamless access to information that is stored in many millions of different geographical locations

• In contrast, the Grid is an emerging infrastructure that provides seamless access to computing power and data storage capacity distributed over the globe

The most extensive scientific computing grid

15 Petabytes(15 millions of GB)of data every year

100’000 processors

200 computer centres around the planet

Should run 100 millions jobs

Used by 5000 scientists in 500 institutes

Virtualization

Why virtual?• Steady flow of requests for dedicated servers

in the CERN computer centre– Excellent network connectivity– Reliable power supply, cooling– 24x365 monitoring with operator’s presence– Daily tape backup– Use the hardware without owning the responsibility (maintenance,

procurement)– Focus on application without sharing the resources– Improve the CPU utilization of grid nodes– Optimize TCO

Buy and maintain physical hardware

(and computer centre)

Virtual infrastructure(IaaS)

• Extra effort to procure and maintain HW

• Delivery time in several weeks

• Lack of flexibility• Not easy to adapt to

dynamic patterns

Infrastructure as a Service (IaaS)

• Ready in ~ minutes• Highly flexible• Efficient capacity

planning

Experience since 2006• Server Self Service Center (S3C)

o Choose your server from a set of predefined imageso Take resources from the pool of available HWo Available within minutes

Requirements have evolved• New requirements identified

o Flexibility of resource allocationo Higher performanceo High-availability model

adapted to customers• Larger scale

o Efficient management

Source: Gartner (August 2008)

Why Hyper-V?• A built-in component of the operating system• Create powerful VM• 64-bit support for guests• Linux support• High availability• Quick migration• Manageability• High performance, reliability, security• VHD compatibility

Why SCVMM?• One solution to centrally manage all virtual infrastructure• Windows Powershell API• V2V and P2V capabilities• Web portal• Intelligent placement• Library• Templates• Delegated management roles• Job history• Support for highly available VM• VM Migration

Hyper-V Infrastructure

System Architecture

Microsoft Virtual Machine Manager

Windows Powershell

SOAP Services

Virtual Machine Manager

AdminConsole

CERN Virtual InfrastructureWeb Interface

Backups

OS Maintenance

LAN DB

Application Management

Web Portal

Challenges• Console access from Linux• Missing .Net API for SCVMM• Time sync issues in guests

Experiences• Cost efficient customized cloud computing infrastructure• Maintenance with limited downtime • Disaster recovery of VMs within minutes• Improved performance compared to Virtual Server 2005

Source: http://blogs.msdn.com/modonovan

Experiences• 172 running virtual machines• 17 templates• Scalable to large number of VM• Expiration handling• Green computing

January

Febru

aryMarc

hApril

MayJune

July

August

Septem

ber

October

0102030405060708090

Monthly increment

January

Febru

aryMarc

hApril

MayJune

July

August

Septem

ber

October

0

50

100

150

200

250

300

350

Number of guests

Real life use cases• Video streaming for LHC First Beam Day

6 virtual machines needed for ~1 week

• Terminal Servers for Engineering Apps A terminal servers installed with older version of the apps

• Oracle Application servers

Real life use cases• CERN Media Archive• CERN Alerter web server

Physical server with 2xCPU, 4GB RAM Upgrade necessary because of OS driver issue Virtual server set up “on demand” Resources limited to 1xCPU, 2 GB RAM

• Physics analysis running in VM• Etc, etc.

Future work• Upgrade to Hyper-V 2.0 & SCVMM 2008 R2• Use the new “Cluster Shared Volume” feature• Use the new “Rapid Provisioning” feature• VDI functionality

Conclusion• Innovative physics research laboratory• Pushing latest technology to its limits• Moving services to the cloud

Visit our websites:Informations: www.cern.chCERN TV: www.youtube.com/cernRecruitment: www.cern.ch/jobs