Data-driven Innovation - Speaker Deck

Slide 1

Slide 1 text

Data-driven innovation [email protected] Dr. Matt Wood @mza

Slide 2

Slide 2 text

Hello

Slide 3

Slide 3 text

Hello

Slide 4

Slide 4 text

Data

Slide 5

Slide 5 text

DNA

Slide 6

Slide 6 text

Chromosome 11 : ACTN3 : rs1815739

Slide 7

Slide 7 text

Chromosome X : rs6625163

Slide 8

Slide 8 text

Chromosome 19 : FUT2 : rs601338

Slide 9

Slide 9 text

Chromosome 2 : rs10427255

Slide 10

Slide 10 text

TYPE II Chromosome 10 : rs7903146

Slide 11

Slide 11 text

+0.25 Chromosome 15 : rs2472297

Slide 12

Slide 12 text

I know this, because...

Slide 13

Slide 13 text

No content

Slide 14

Slide 14 text

A T C G G T C C A G G

Slide 15

Slide 15 text

A T C G G T C C A G G A G C C A G G U C C Transcription

Slide 16

Slide 16 text

A T C G G T C C A G G A G C C A G G U C C Translation Ser Glu Val Transcription

Slide 17

Slide 17 text

No content

Slide 18

Slide 18 text

No content

Slide 19

Slide 19 text

Chromosome 11 : ACTN3 : rs1815739

Slide 20

Slide 20 text

Chromosome X : rs6625163

Slide 21

Slide 21 text

Chromosome 19 : FUT2 : rs601338

Slide 22

Slide 22 text

Chromosome 2 : rs10427255

Slide 23

Slide 23 text

TYPE II Chromosome 10 : rs7903146

Slide 24

Slide 24 text

+0.25 Chromosome 15 : rs2472297

Slide 25

Slide 25 text

I know all that, because...

Slide 26

Slide 26 text

Human Genome Project

Slide 27

Slide 27 text

40 species ensembl.org

Slide 28

Slide 28 text

Compare

Slide 29

Slide 29 text

Change

Slide 30

Slide 30 text

Less

Slide 31

Slide 31 text

No content

Slide 32

Slide 32 text

No content

Slide 33

Slide 33 text

Compare

Slide 34

Slide 34 text

Transformative

Slide 35

Slide 35 text

No content

Slide 36

Slide 36 text

Data generation costs are falling everywhere

Slide 37

Slide 37 text

Customer segmentation, ﬁnancial modeling, system analysis, line of sight, business intelligence.

Slide 38

Slide 38 text

Opportunity

Slide 39

Slide 39 text

Transformation

Slide 40

Slide 40 text

Innovation

Slide 41

Slide 41 text

Generation Collection & storage Analytics & computation Collaboration & sharing

Slide 42

Slide 42 text

Generation Collection & storage Analytics & computation Collaboration & sharing lower cost, increased throughput

Slide 43

Slide 43 text

Generation Collection & storage Analytics & computation Collaboration & sharing lower cost, increased throughput highly constrained

Slide 44

Slide 44 text

Barrier

Slide 45

Slide 45 text

Data generation challenge X

Slide 46

Slide 46 text

Analytics challenge

Slide 47

Slide 47 text

Accessibility challenge

Slide 48

Slide 48 text

Enter the AWS Cloud

Slide 49

Slide 49 text

Utility

Slide 50

Slide 50 text

Remove constraints

Slide 51

Slide 51 text

Data-driven innovation

Slide 52

Slide 52 text

Distributed

Slide 53

Slide 53 text

Slide 54

Slide 54 text

2 Software for distributed storage & analysis

Slide 55

Slide 55 text

2 Software for distributed storage & analysis Infrastructure for distributed storage & analysis

Slide 56

Slide 56 text

Software Frameworks for data-intensive work loads. Distributed by design.

Slide 57

Slide 57 text

Infrastructure Platform for data-intensive work loads. Distributed by design.

Slide 58

Slide 58 text

Support the data timeline

Slide 59

Slide 59 text

Generation Collection & storage Analytics & computation Collaboration & sharing highly constrained

Slide 60

Slide 60 text

Generation Collection & storage Analytics & computation Collaboration & sharing

Slide 61

Slide 61 text

Lower the barrier to entry

Slide 62

Slide 62 text

Agility

Slide 63

Slide 63 text

Responsive

Slide 64

Slide 64 text

Generation Collection & storage Analytics & computation Collaboration & sharing

Slide 65

Slide 65 text

Generation DynamoDB Analytics & computation Collaboration & sharing

Slide 66

Slide 66 text

Generation DynamoDB EC2, Elastic MapReduce Collaboration & sharing

Slide 67

Slide 67 text

Generation DynamoDB EC2, Elastic MapReduce S3, Public Datasets

Slide 68

Slide 68 text

Tools and techniques for working productively with data

Slide 69

Slide 69 text

Scale

Slide 70

Slide 70 text

Secure

Slide 71

Slide 71 text

2 Software for distributed storage & analysis Infrastructure for distributed storage & analysis

Slide 72

Slide 72 text

Amazon EC2

Slide 73

Slide 73 text

Scale out systems Embarrassingly parallel Queue based distribution Small, medium and high scale

Slide 74

Slide 74 text

High performance

Slide 75

Slide 75 text

High performance Compute performance

Slide 76

Slide 76 text

Cluster Compute Intel Xeon E5-2670 10 gigabit, non-blocking network 60.5 Gb Placement groupings

Slide 77

Slide 77 text

Cluster Compute Intel Xeon E5-2670 10 gigabit, non-blocking network 60.5 Gb Placement groupings +GPU

Slide 78

Slide 78 text

240 TFLOPS

Slide 79

Slide 79 text

High performance Compute performance IO performance

Slide 80

Slide 80 text

Unstructured

Slide 81

Slide 81 text

Variable

Slide 82

Slide 82 text

Amazon DynamoDB Predictable, consistent performance Unlimited storage Single digit millisecond latencies No schema. Zero admin.

Slide 83

Slide 83 text

...and SSDs for all

Slide 84

Slide 84 text

hi1.4xlarge 2 x 1Tb SSD storage 10 gigabit networking HVM: 90k IOPS read, 9k to 75k write PV: 120k IOPS read, 10k to 85k write

Slide 85

Slide 85 text

Netflix “The hi1.4xlarge configuration is about half the system cost for the same throughput.” http://techblog.netflix.com/2012/07/benchmarking-high-performance-io-with.html

Slide 86

Slide 86 text

Provisioned IOPS Provision required IO performance EBS optimized instances

Slide 87

Slide 87 text

Cost optimization

Slide 88

Slide 88 text

Reserved capacity

Slide 89

Slide 89 text

Reserved capacity On-demand

Slide 90

Slide 90 text

Reserved capacity On-demand

Slide 91

Slide 91 text

Spot instances

Slide 92

Slide 92 text

No content

Slide 93

Slide 93 text

$0.2530 vs $2.40

Slide 94

Slide 94 text

2 Software for distributed storage & analysis Infrastructure for distributed storage & analysis

Slide 95

Slide 95 text

map/reduce

Slide 96

Slide 96 text

Map. Reduce.

Slide 97

Slide 97 text

Write functions. Scale up.

Slide 98

Slide 98 text

Hadoop

Slide 99

Slide 99 text

Undi erentiated heavy lifting

Slide 100

Slide 100 text

Amazon Elastic MapReduce Managed Hadoop Clusters Easy to provision and monitor Write two functions. Scale up. Choice of Hadoop ﬂavors

Slide 101

Slide 101 text

Amazon Elastic MapReduce Integrates with S3 Analytics for DynamoDB Perfect for Spot pricing

Slide 102

Slide 102 text

Input data S3

Slide 103

Slide 103 text

Elastic MapReduce Code Input data S3

Slide 104

Slide 104 text

Elastic MapReduce Code Name node Input data S3

Slide 105

Slide 105 text

Elastic MapReduce Code Name node Input data S3 Elastic cluster

Slide 106

Slide 106 text

Elastic MapReduce Code Name node Input data S3 Elastic cluster HDFS

Slide 107

Slide 107 text

Elastic MapReduce Code Name node Input data S3 Elastic cluster HDFS Queries + BI Via JDBC, Pig, Hive

Slide 108

Slide 108 text

Elastic MapReduce Code Name node Output S3 + SimpleDB Input data S3 Elastic cluster HDFS Queries + BI Via JDBC, Pig, Hive

Slide 109

Slide 109 text

Output S3 + SimpleDB Input data S3

Slide 110

Slide 110 text

CDC Centers for Disease Control and Prevention

Slide 111

Slide 111 text

“BioSense 2.0 protects the health of the American people by providing timely insight into the health of communities, regions, and the nation by o ering a variety of features to improve data collection, standardization, storage, analysis, and collaboration”

Slide 112

Slide 112 text

Health data Collection & storage Analytics & computation Collaboration & sharing

Slide 113

Slide 113 text

Health data Collection & storage Analytics & computation Collaboration & sharing highly constrained