Grids and Clouds:Tackle the Limits of

ICT

Simon C. Lin (林誠謙 )Academia Sinica Grid Computing, Taipei,

Taiwan 11529Simon.Lin@twgrid.org

28 April 2009, TrendMicro Clouds Seminar

Future of Computing?

•Computing is about the application of Information and Communication Technology

• “Cambrian explosions” in ICT, rich diversity in services, middleware, commodity hardware, devices and sensors

There are Limits to Everything ...

Induction: What are Grids and e-Science? –May 18th, 2004 -

9

Exponential Growth World

Gilder’s Law(32X in 4 yrs)

Storage Law (16X in 4yrs)

Moore’s Law(5X in 4yrs)

Triumph of Light – Scientific American. George Stix, January 2001

Pe

rfo

rman

ce p

er D

olla

r S

pen

t Optical Fibre(bits per second)

Chip capacity(# transistors)

Data Storage(bits per sq. inch)

Number of Years0 1 2 3 4 5

9 12 18

Doubling Time(months)

• Rock's Law: the cost of leading-edge plant doubles every three years (now at 3-4 billion USD)

11

The Data Deluge• A large novel: 1 Mbyte; The Bible: 5 Mbytes

• A Mozart symphony (compressed): 10 Mbytes

• A digital mammogram: 100 Mbytes

• OED on CD: 500 Mbytes, Digital movie (compressed): 10 Gbytes

• Annual production of refereed journal literature (∼20 k journals; ∼2 M articles): 1 Tbyte

• Library of Congress: 20 Tbytes

• The Internet Archive (10 B pages) (From 1996 to 2002): 100 Tbytes

• Annual production of information (print, film, optical & magnetic media): 1500 to 3000 Pbytes. From 2008, annual production will be greater than total available storage space!

• All Worldwide Telephone communication in 2002: 19.3 ExaBytes. Global Internet traffic will reach EB order next year!

• Moore’s Law enables instruments and detectors to generate unprecedented amount of data in all scientific disciplines

<<< Limit and Challenge!

Moore’s Law in HPC Moore’s Law in HPC MachinesMachinesLimit Soon! >>>

13

100 -1000x Limit! >>>

Power Consumption Challenge for More Computing Power

<<< Limit!

<<

< A

rch.

Lim

it!

EGEE Training Materials, 2004

16

Enabling Grids for E-sciencE

INFSO-RI-508833

Tera → Peta Bytes

• RAM time to move– 15 minutes

• 1Gb WAN move time– 10 hours ($1000)

• Disk Cost– 7 disks = $5000

(SCSI)

• Disk Power– 100 Watts

• Disk Weight– 5.6 Kg

• Disk Footprint– Inside machine

• RAM time to move– 2 months• 1Gb WAN move time– 14 months ($1 million)• Disk Cost– 6800 Disks + 490 units + 32 racks = $7 million• Disk Power– 100 Kilowatts• Disk Weight– 33 Tonnes• Disk Footprint– 60 m2

May 2003 Approximately Correct Distributed Computing Economics

Jim Gray, Microsoft Research, MSR-TR-2003-24

<<< Limit to Scaling!

17

From Optimizing Architecture to Optimizing Organisation

• High-performance computing has moved from being a problem of optimizing the architecture of an individual supercomputer to one of optimizing the organization of large numbers of ordinary computers operating in parallel.

Scott Kirkpatrick, SCIENCE Vol. 299, 2003, p668

18

Simulated Time Can Get Rough

<<< Proc. # Limit !

How Large the Computing Could

Be?

21

Enabling Grids for E-sciencE

EGEE-III INFSO-RI-222667 EGEE - Bob Jones - OGF25/User Forum - 2-6 March 2009

~280 sites45 countries>80,000 CPUs>25 PetaBytes>14,000 users>250,000 jobs/day

• Quality:

– Monitoring via Nagios - distributed via official releases, configured through YAIM, integrated with other tools

– Gradual implementation of Service Level Agreement with sites

– Result: 85% of sites are now above the 75% availability threshold

• Geographical expansion:

– now have production sites all across Asia: Australia, China, India, Japan, Korea, Malaysia, Pakistan, Taiwan, Thailand

In certification: Indonesia, Philippines, Vietnam

SETI@home: 1,000,000 CPUs“Poor man’s Grid”

BOINC (Berkeley Open Infrastructure for Network Computing) platform launched in 2003.General-purpose open-source platform for client-server model of distributed computing.

>50 volunteer computing projects today in a wide range of sciences. Not all use BOINC.

Volunteer Computing

Folding@home: >1 petaflop

Sony pre-installs folding@home on Playstation-III – user can choose to run in background. 50k PS3s make first distributed petaflop machine (Guiness World Record Sept 2007).

Often of Philanthropic Nature

LHC@home: >3000 CPU-years,>60K Volunteers

Fortran program Sixtrack simulates LHC proton beam stability for 105-106 orbits. Include real magnet parameters, beam-beam effects. Predict stable operation conditions.

Citizen Cyber-science Projects

Volunteer Thinking

Performance

• Current

– 500K people, 1M computers

– 6.5 PetaFLOPS (3 from GPUs, 1.4 from PS3s)

• Potential

– 1 billion PCs today, 2 billion in 2015

– GPU: approaching 1 TFLOPS

– How to get 1 ExaFLOPS:• 4M GPUs * 0.25 availability

– How to get 1 Exabyte:• 10M PC disks * 100 GB

History of volunteer computing

Applications

Middleware

1995 2005distributed.net, GIMPS

SETI@home, Folding@home

Commercial: Entropia, United Devices, ...

BOINC

Climateprediction.netPredictor@homeIBM World Community GridEinstein@homeRosetta@home ...

20052000 now

Academic: Bayanihan, Javelin, ...

Applications

34

ASGC Introduction

Large Hadron Collider (LHC)

Avian Flu Drug DiscoveryGrid Application Platform

A Worldwide Grid Infrastructure

Asia Pacific Regional Operation Center

>280 sites, 45 countries>80,000 CPUs, >25 PetaBytes>14,000 users, >200 VOs>250,000 jobs/day

Best Demo Award of EGEE’07

Lightweight Problem Solving Framework

1. Most Reliable T1: 98.83%

2. Very Highly Performing and most Stable Site in

CCRC08

Max CERN/T1-ASGC Point2Point Inbound : 7.3

Gbps

35

由歐洲到台灣每秒能傳送兩部大英百科全書 -- 歷史記錄

Max CERN/T1->ASGC Inbound : 7.3 Gbps, and Outbound : 5.9 Gbps

Cloud Computing

61

Is It Buzzword evolution?Is It Buzzword evolution?

One distributed computing buzzword per decade

Metacomputing (~1987, L. Smarr)

Grid computing (~1997, I. Foster, K. Kesselman)

Cloud computing (~2007, E. Schmidt ?)

62

Cloud hype 1Cloud hype 1

63

Cloud hype 2Cloud hype 2

It’s stupidity. It’s worse than stupidity: it’s a marketing hype campaign. Somebody is saying this is inevitable — and whenever you hear somebody saying that, it’s very likely to be a set of businesses campaigning to make it true.Richard Stallman, quoted in The Guardian, September 29, 2008

The interesting thing about Cloud Computing is that we’ve redefined Cloud Computing to include everything that we already do. . . . I don’t understand what we would do differently in the light of Cloud Computing other than change the wording of some of our ads.Larry Ellison, quoted in the Wall Street Journal, September 26, 2008

64

Cloud hype 3Cloud hype 3

--- Cloud computing--- Grid computing

65

Cloud hype 4Cloud hype 4

Illustration by David SimondsThe Economist April 09

The Open Cloud Manifesto

“The industry needs an objective, straightforward conversation about how this new computing paradigm will impact organizations, how it can be used with existing technologies, and the potential pitfalls of proprietary technologies that can lead to lock-in and limited choice.”

OpenCloudManifesto.org

66

March 2-6, 2009 OGF 25/EGEE User Forum, Catania Italywww.reservoir-fp7.eu

Why Cloud Computing ?

• Innovative business models require a utility-like intelligent infrastructure that embraces complexity to be successful in the competitive, fast-paced, services-based global economy. – ECONOMICS: Small up front investment and can be billed by

consumption. Reduction of TCO allows clients to pursue operational efficiency and productivity.

– RISK MANAGEMENT: Small up front commitment allows clients to try many new services faster and choose. This reduces big failure risks and allows clients to be innovative.

– TIME TO MARKET: Adopt new services quickly for pilot usages and scale quickly to global scale.

– INFORMATION SOCIETY: Value-added information generated by collection and analysis of massive amounts of unstructured data.

– UBIQUITOUS SOCIETY: Accessible via a heterogeneous set of devices (PC, phone, telematics..)

Because it makes sense !!!

SaaS PaaS IaaS

69

Definitions 1Definitions 1

Enterprise Grid CloudsService Grid

70

Definitions 2Definitions 2

Cloud computing is an on-demand service offering a large pool of easily usable and accessible virtualized resources (such as hardware, development platforms and/or software services) in a pay-per-use model. Clouds are usually commercial and use proprietary interfaces.

Service grids are systems that federate, share and coordinate distributed resources from different organizations which are not subject to centralized control, using standard, open, general-purpose protocols and interfaces to deliver non-trivial qualities of service. Service grids are used by Virtual Organisations, thematic groups of users crossing administrative and geographical boundaries.

71

Definitions 3Definitions 3

Contract harvesting(Cloud)

Cooperative harvesting(Service Grid)

72

Grids vs. Clouds 1Grids vs. Clouds 1

73

Grids vs. Clouds 2Grids vs. Clouds 2

74

Grids vs. Clouds 3Grids vs. Clouds 3

Cost of 1 Teraflop-year

Cloud: $1.75M Amazon EC2 rates

Cluster: $145KHardware (computing, network, storage); power; infrastructure; sysadmin

Volunteer: $1K - $10K

Server hardware; sysadmin; web development

Opportunities

79

Grid & Cloud: opportunities

Many cloud offerings = competitionAmazon Elastic Cloud Computing (EC2); IBM Blue Cloud; Microsoft Azure Services Platform; Sun Open Cloud Initiative; Google App Engine; Salesforce.com Force.com Cloud; GoGrid Cloud Hosting; RackSpace Cloud Hosting; Flexiscale Utility Computing on Demand…

Some academic/open-source variants emergingEucalyptus; Nimbus, OpenNebula,…

Experiments in Grid + CloudBalticCloud,StratusLab, VirtCloud, …some to be discussed in this session

80

Enabling Grids for E-sciencE

EGEE-III INFSO-RI-222667 EGEE - Bob Jones - OGF25/User Forum - 2-6 March 2009

Where would clouds come in?

• Majority of EGEE’s CPUs are at the sites• The interaction with site resources is ‘simple’

In: Authorization, Compute & StorageOut: Monitoring & Accounting

• Additional Site RequirementsVO specific site servicesVO specific applications deployed to the worker nodes

• Provide classical ‘cloud’ scale out capabilityPotentially, with VO specific worker nodes

81

Enabling Grids for E-sciencE

EGEE-III INFSO-RI-222667 EGEE - Bob Jones - OGF25/User Forum - 2-6 March 2009

Production & Volunteer grids

Conclusion

Peter Coveney (P.V.Coveney@ucl.ac.uk)

Grid ComputingMy preferred definition:

Grid computing is distributed computing performed transparently across multiple administrative domains

Notes:

Computing means any activity involving digital information -- no distinction between numeric/symbolic, or numeric/data/viz

Transparency implies minimal complexity for users of the technology

Grid - its really about collaboration!

• It’s about sharing and building a vision for the future

– And it’s about getting connected

• It’s about the democratization of science

• It takes advantage of Open Source!

Source: Vicky White

89

Conclusion

• Both Grids and Clouds are evolutions of Virtualisation

• Virtulaisation will provide enormous flexibility in both Grids and Clouds management

• Grids and Clouds are opportunity to each other!

• The ICT is always changing, nothing is guaranteed to prevail

• Taking advantage of your Core Competence is the most important issue!