Introduction to Cloud Computing

http://net.pku.edu.cn/~course/cs402/2009/

pb@net.pku.edu.cn 6/30/2009

(Cloud Computing)?

(Cloud Computing)

What is Cloud Computing?

1.

2.

3.

First write down your own opinion about cloud computing , whatever you thought about in your mind. Question: What ? Who? Why? How? Pros and cons? The most important question is: What is the relation with me?

Cloud Computing is

No software access everywhere by Internet power -- Large-scale data processing Appeal for startups

Cost efficiency Software as platform

Security Data lock-in
5

SaaS PaaS Utility Computing

Cons

Software as a Service (SaaS)

a model of software deployment whereby a provider licenses an application to customers for use as a service on demand.

Platform as a Service (PaaS)

Web ApplicationServicesPaaS Internet Multitenant architecture platform

7

Utility Computing

pay-as-you-go Microsoftpay lessutility computing 500 use less pay lesscloud computing
8

Cloud Computing is

Key Characteristics
illusion of infinite computing resources available on demand; elimination of an up-front commitment by Cloud users; ability to pay for use of computing resources on a very large datacenters short-term basis as needed billing large-scale software infrastructure utility computing operational expertise

10

Why now?

very large-scale datacenter Business

pay-as-you-go computing

11

Key Players

Amazon Web Services Google App Engine Microsoft Windows Azure

12

Key Applications

Mobile Interactive applications, Tim OReilly Mobile datacenter mashup Parallel batch processingCloud ComputingMapReduceHadoop /cloud Amazonhost large public datasets for free The rise of analyticstransaction based analytics Extension of compute-intensive desktop application matlab, mathematicacloud computingwoo~
13

Cloud Computing = Silver Bullet?

Google37 Google

Problem of Data Lock-in

14

Challenges

15

Some other Voices

The interesting thing about Cloud Computing is that weve redefined Cloud Computing to include everything that we already do. . . . I dont 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

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

16

Whats matter with ME?!

What you want to do with 1000pcs, or even 100,000 pcs?

17

Cloud is coming
Google alone has 450,000 systems running across 20 datacenters, and Microsoft's Windows Live team is doubling the number of servers it uses every 14 months, which is faster than Moore's Law

Data Center is a Computer

Parallelism everywhere

Massive Scalable Reliable Resource Management Data Management

Programming Model & Tools

18

20

Happening everywhere!
microarray chips Molecular biology (cancer) fiber optics Network traffic (spam)

microprocessors

Simulations (Millennium) particle colliders

300M/day

Particle events (LHC)

1B
21

1M/sec

22

Maximilien Brice, CERN

23

Maximilien Brice, CERN

24

Maximilien Brice, CERN

25

Maximilien Brice, CERN

How much data?

Internet archive has 2 PB of data + 20 TB/month Google processes 20 PB a day (2008) all words ever spoken by human beings ~ 5 EB CERNs LHC will generate 10-15 PB a year Sanger anticipates 6 PB of data in 2009
640K ought to be enough for anybody.

26

NERSC User George Smoot wins 2006 Nobel Prize in Physics

Smoot and Mather 1992 COBE Experiment showed anisotropy of CMB

Cosmic Microwave Background Radiation (CMB): an image of the universe at 400,000 years

27

The Current CMB Map

source J. Borrill, LBNL

Unique imprint of primordial physics through the tiny anisotropies in temperature and polarization. Extracting these Kelvin fluctuations from inherently noisy data is a serious computational challenge.
28

Experiment

Evolution Of CMB Data Sets: Cost > O(Np^3 ) Limiting

Nt Np Nb
Data

Notes

COBE (1989) BOOMERanG (1998) (4yr) WMAP (2001)

2x109

6x103

3x101

Time

Satellite,

Workstation

3x108

5x105

3x101

Pixel

Balloon, 1st HPC/NERSC

7x1010

4x107

1x103

Satellite, Analysis-bound Satellite, Major HPC/DA effort Ground, NGmultiplexing Satellite, Early planning/design

Planck (2007)

5x1011

6x108

6x103

Time/ Pixel

POLARBEAR (2007)

8x1012

6x106

1x103

Time

CMBPol (~2020) data compression

1014

109

104

Time/ Pixel

29

Example: Wikipedia Anthropology

Kittur, Suh, Pendleton (UCLA, PARC), He Says, She Says: Conflict and Coordination in Wikipedia CHI, 2007

Increasing fraction of edits are for work indirectly related to articles

Experiment

Computation

Download entire revision history of Wikipedia 4.7 M pages, 58 M revisions, 800 GB Analyze editing patterns & trends
30

Hadoop on 20-machine cluster

Example: Scene Completion

Hays, Efros (CMU), Scene Completion Using Millions of Photographs SIGGRAPH, 2007

Image Database Grouped by Semantic Content

Computation

30 different Flickr.com groups 2.3 M images total (396 GB).

Select Candidate Images Most Suitable for Filling Hole

Classify images with gist scene detector [Torralba] Color similarity Local context matching

Index images offline 50 min. scene matching, 20 min. local matching, 4 min. compositing Reduces to 5 minutes total by using 5 machines Flickr.com has over 500 million images

Extension

31

Example: Web Page Analysis

Fetterly, Manasse, Najork, Wiener (Microsoft, HP), A Large-Scale Study of the Evolution of Web Pages, Software-Practice & Experience, 2004

Experiment

Systems Challenge
Moreover, we experienced a

Use web crawler to gather 151M HTML pages weekly 11 times Generated 1.2 TB log information Analyze page statistics and change frequencies

catastrophic disk failure during the third crawl, causing us to lose a quarter of the logs of that crawl.

32

GATGCTTACTATGCGGGCCCC CGGTCTAATGCTTACTATGC

?
Subject genome

GCTTACTATGCGGGCCCCTT AATGCTTACTATGCGGGCCCCTT TAATGCTTACTATGC AATGCTTAGCTATGCGGGC AATGCTTACTATGCGGGCCCCTT AATGCTTACTATGCGGGCCCCTT CGGTCTAGATGCTTACTATGC AATGCTTACTATGCGGGCCCCTT CGGTCTAATGCTTAGCTATGC ATGCTTACTATGCGGGCCCCTT

Reads

Sequencer
33

DNA Sequencing

Genome of an organism encodes genetic information in long sequence of 4 DNA nucleotides: ATCG

Bacteria: ~5 million bp Humans: ~3 billion bp

Current DNA sequencing machines can generate 1-2 Gbp of sequence per day, in millions of short reads (25-300bp)

Shorter reads, but much higher throughput Per-base error rate estimated at 1-2% (Simpson, et al, 2009)

ATCTGATAAGTCCCAGGACTTCAGT GCAAGGCAAACCCGAGCCCAGTTT TCCAGTTCTAGAGTTTCACATGATC GGAGTTAGTAAAAGTCCACATTGAG

Recent studies of entire human genomes have used 3.3 (Wang, et al., 2008) & 4.0 (Bentley, et al., 2008) billion 36bp reads

~144 GB of compressed sequence data

34

Subject reads CTATGCGGGC A T CTATGCGG TCTAGATGCT GCTTA T CTAT A T CTATGCGG A T CTATGCGG A T CTATGCGG TTA T CTATGC CTATGCGGGC GCTTA T CTAT CTAGATGCTT Alignment CGGTCTAGATGCTTAGCTATGCGGGCCCCTT

Reference sequence

35

Subject reads

ATGCGGGCCC CTAGATGCTT CTATGCGGGC TCTAGATGCT ATCTATGCGG CGGTCTAG ATCTATGCGG CTT CGGTCT TTATCTATGC CCTT CGGTC GCTTATCTAT GCCCCTT GCTTATCTAT CGG GGCCCCTT CGGTCTAGATGCTTATCTATGCGGGCCCCTT

Reference sequence

36

Example: Bioinformatics
Michael Schatz. CloudBurst: Highly Sensitive Read Mapping with MapReduce. Bioinformatics, 2009, in press.

Evaluate running time on local 24 core cluster

Running time increases linearly with the number of reads

37

Example: Data Mining

Haoyuan Li, Yi Wang, Dong Zhang, Ming Zhang, Edward Y. Chang: Pfp: parallel fp-growth for query recommendation. RecSys 2008: 107114

del.icio.us crawl->a bipartite graph covering 802739 Webpages and 1021107 tags.

38

An Example

Try on these collection:

2006870 Million , 2 TB. Google, Yahoo100+ Billion pages

40

Divide and Conquer

Work

Partition
w3
worker

w1
worker

w2
worker

r1

r2

r3

Result

Combine

41

Whats Mapreduce

Parallel/Distributed Computing Programming Model

Input split

shuffle

output

42

Typical problem solved by MapReduce

: key/value Map: extract something

map (in_key, in_value) -> list(out_key, intermediate_value) input key/value pair key/value pairs

Shuffle:

key
reduce (out_key, list(intermediate_value)) -> list(out_value) keyvalues (usually just one)

Reduce: aggregate, summarize, filter, etc.

43

Word Frequencies in Web pages

one document per record map function

key = document URL value = document contents document<word, 1>

map (potentially many) key/value pairs.

44

Example continued:

MapReduce()key (shuffle/sort) reduce functionkeyvalues

sum

Reduce<key, sum>

45

MapReduce Runtime System

46

History of Hadoop

2004 - Initial versions of what is now Hadoop Distributed File System and Map-Reduce implemented by Doug Cutting & Mike Cafarella December 2005 - Nutch ported to the new framework. Hadoop runs reliably on 20 nodes. January 2006 - Doug Cutting joins Yahoo! February 2006 - Apache Hadoop project official started to support the standalone development of Map-Reduce and HDFS. March 2006 - Formation of the Yahoo! Hadoop team May 2006 - Yahoo sets up a Hadoop research cluster - 300 nodes April 2006 - Sort benchmark run on 188 nodes in 47.9 hours May 2006 - Sort benchmark run on 500 nodes in 42 hours (better hardware than April benchmark) October 2006 - Research cluster reaches 600 Nodes December 2006 - Sort times 20 nodes in 1.8 hrs, 100 nodes in 3.3 hrs, 500 nodes in 5.2 hrs, 900 nodes in 7.8 January 2006 - Research cluster reaches 900 node April 2007 - Research clusters - 2 clusters of 1000 nodes Sep 2008 - Scaling Hadoop to 4000 nodes at Yahoo!

47

From Theory to Practice

1. Scp data to cluster 2. Move data into HDFS

3. Develop code locally

4. Submit MapReduce job 4a. Go back to Step 3 You Hadoop Cluster

5. Move data out of HDFS 6. Scp data from cluster

48

MapReduce MapReduce MapReduce

50


LEC# TOPICS 1 - ABSTRACT MapReduce MapReduce Inverted IndexMapReduce PageRankMapReduce MapReduce ClusteringMapReduce MapReduce MapReduce 51

MapReduce
2 3 4 MapReduce Inverted Index PageRank MapReduce Clustering 5 6 7 8 MapReduce

Grading Policy
Lab Lab Lab Lab 1 2 3 4 - Introduction to Hadoop, Eclipse A Simple Inverted Index - PageRank over Wikipedia Corpus Clustering the Netflix Movie Data

Hw1 - Read - Intro Distributed system; Intro MapReduce Programming. Hw2 - Read MapReduce[1] Hw3 Read GFS[2] Hw4 Read Pig Latin[3]

30% Assignments 20% Readings 50% Course project

52

Programming Language

Lots of java programming practices

53

Teachers and Resources

http://net.pku.edu.cn/~cour se/cs402/2009/ http://groups.google.com/g roup/cs402pku http://hadoop.apache.org/c ore/ http://net.pku.edu.cn/~cour se/cs402/2008/resource.ht ml

54

Hadoop

Resources

Homework

http://net.pku.edu.cn/~course/cs402/2009/ 3-4project Lab 1 - Introduction to Hadoop, Eclipse Intro Distributed system; Intro Parallel Programming. http://code.google.com/edu/parallel/dsd-tutorial.html http://code.google.com/edu/parallel/mapreduce-tutorial.html

Lab1

HW Reading1

55

Summary

CloudComputing brings

Possible of using unlimited resources on-demand, and by anytime and anywhere Possible of construct and deploy applications automatically scale to tens of thousands computers Possible of construct and run programs dealing with prodigious volume of data

How to make it real?

Distributed File System Distributed Computing Framework

56

Q&A

[1] J. Dean and S. Ghemawat, "MapReduce: Simplified Data Processing on Large Clusters," in Osdi, 2004, pp. 137-150. [2] G. Sanjay, G. Howard, and L. Shun-Tak, "The Google file system," in Proceedings of the

nineteenth ACM symposium on Operating systems principles. Bolton Landing, NY, USA:

ACM Press, 2003. [3] O. Christopher, R. Benjamin, S. Utkarsh, K. Ravi, and T. Andrew, "Pig latin: a not-so-foreign language for data processing," in Proceedings of

the 2008 ACM SIGMOD international conference on Management of data. Vancouver, Canada:
ACM, 2008.
58

Google App Engine

App Engine handles HTTP(S) requests, nothing else

Think RPC: request in, processing, response out Works well for the web and AJAX; also for other services

App configuration is dead simple

No performance tuning needed

Everything is built to scale

infinite number of apps, requests/sec, storage capacity APIs are simple, stupid

59

App Engine Architecture

req/resp stateless APIs urlfech mail images R/O FS stdlib app

Python VM process

stateful APIs

memcache

datastore

60

60

Microsoft Windows Azure

61

Amazon Web Services

Amazons infrastructure (auto scaling, load balancing) Elastic Compute Cloud (EC2) scalable virtual private server instances Simple Storage Service (S3) Simple Queue Service (SQS) messaging SimpleDB - database Flexible Payments Service, Mechanical Turk, CloudFront, etc.

62

Amazon Web Services

Very flexible, lower-level offering (closer to hardware) = more possibilities, higher performing Runs platform you provide (machine images) Supports all major web languages Industry-standard services (move off AWS easily) Require much more work, longer time-to-market

Deployment scripts, configuring images, etc.

Various libraries and GUI plug-ins make AWS do help

63

Price of Amazon EC2

64