1

Where The Rubber Meets the SkyGiving Access to Science Data

Jim GrayMicrosoft Research

Gray@Microsoft.comHttp://research.Microsoft.com/~Gray

Alex SzalayJohns Hopkins University

Szalay@JHU.edu

Talk at 16th International Conference on

Scientific and Statistical Database Management

June 2004, Santorini, Greece

2

Promised Abstract: Historically, scientists gatherer and

analyzed their own data. But technology has created functional specialization where some scientists gather or generate data, and others analyze it. Technology allows us to easily capture vast amounts of empirical data and to generate vast amounts of simulated data. Technology also allows us to store these bytes almost indefinitely. But there are few tools to organize scientific data for easy access and query, few tools to curate the data, and few tools to federate science archives. Domain scientists, notably NCBI and the Virtual Observatory, are making heroic efforts to address these problems. But this is a generic problem that cuts across all scientific disciplines. It requires a coordinated effort by the computer science community to build generic tools that will help all the sciences. Our current database products are a start, but much more is needed.

Actual Abstract: I have been working with Alex Szalay and other astronomers for the last 6 years trying to apply DB technology to science problems.

These are some lessons I learned.

3

Outline• New Science

– X-Info for all fields X– WWT as an example– Big Picture– Puzzle– Hitting the wall– Needle in haystack– Mohamed and the mountain

• Working cross disciplines• Data Demographics and Data Handling• Curation

?Experiments &

Instruments

Simulationsfacts

facts

answers

questions

Literature

Other Archives facts

facts

4

New Science Paradigms• Thousand years ago:

science was empirical describing natural phenomena

• Last few hundred years: theoretical branch using models, generalizations

• Last few decades: a computational branch simulating complex phenomena

• Today: data exploration (eScience)unify theory, experiment, and simulation using data management and statistics– Data captured by instruments

Or generated by simulator– Processed by software– Scientist analyzes database / files

2

22.

3

4

a

cG

a

a

2

22.

3

4

a

cG

a

a

5

The Virtual Observatory• Premise: most data is (or could be online)• The Internet is the world’s best telescope:

– It has data on every part of the sky– In every measured spectral band:

optical, x-ray, radio..– As deep as the best instruments (2 years ago).

– It is up when you are up– The “seeing” is always great– It’s a smart telescope:

links objects and data to literature

• Software is the capital expense– Share, standardize, reuse..

6

Why Is Astronomy Special?• Almost all literature online and public ADS: http://adswww.harvard.edu/ CDS: http://cdsweb.u-strasbg.fr/

• Data has no commercial value– No privacy concerns, freely share results with others– Great for experimenting with algorithms

• It is real and well documented– High-dimensional (with confidence intervals)

– Spatial, temporal• Diverse and distributed

– Many different instruments from many different places and many different times

• The community wants to share the data• There is a lot of it (soon petabytes)

IRAS 100

ROSAT ~keV

DSS Optical

2MASS 2

IRAS 25

NVSS 20cm

WENSS 92cm

GB 6cm

7

The Big Picture

• Data ingest

• Managing a petabyte

• Common schema

• How to organize it?

• How to reorganize it?

• How to coexist with others?

• Data Query and Visualization tools • Support/training• Performance

– Execute queries in a minute – Batch (big) query scheduling

The Big Problems

Experiments &Instruments

Simulationsfacts

facts

answers

questions

?Literature

Other Archives facts

facts

8

What X-info Needs from us (cs)(not drawn to scale)

Science Data & Questions

Scientists

DatabaseTo store

dataExecuteQueries

Plumbers

Data Mining

Algorithms

Miners

Question & AnswerVisualizat

ion

Tools

9

Experiment Budgets ¼…½ Software

Software for• Instrument scheduling• Instrument control• Data gathering• Data reduction• Database • Analysis • Visualization

Millions of lines of code

Repeated for experiment after experiment

Not much sharing or learning

Let’s work to change this

Identify generic tools• Workflow schedulers• Databases and libraries • Analysis packages • Visualizers • …

Simulation (computational science) are > ½ software

10

Data Access Hitting a WallCurrent science practice based on data download

(FTP/GREP)Will not scale to the datasets of tomorrow

• You can GREP 1 MB in a second• You can GREP 1 GB in a minute • You can GREP 1 TB in 2 days• You can GREP 1 PB in 3 years.

• Oh!, and 1PB ~5,000 disks

• At some point you need indices to limit searchparallel data search and analysis

• This is where databases can help

• You can FTP 1 MB in 1 sec• You can FTP 1 GB / min (~1$)• … 2 days and 1K$• … 3 years and 1M$

11

Next-Generation Data Analysis

• Looking for– Needles in haystacks – the Higgs particle– Haystacks: dark matter, dark energy,

turbulence, ecosystem dynamics• Needles are easier than haystacks• Global statistics have poor scaling

– Correlation functions are N2, likelihood techniques N3

• As data and computers grow at Moore’s Law, we can only keep up with N logN

• A way out? – Relax optimal notion (data is fuzzy, answers are approximate)

– Don’t assume infinite computational resources or memory• Requires combination of statistics & computer science

12

Smart Data: Unifying DB and Analysis• There is too much data to move around

Do data manipulations at database– Build custom procedures and functions into DB– Unify data Access & Analysis– Examples

• Statistical sampling and analysis• Temporal and spatial indexing• Pixel processing

• Automatic parallelism • Auto (re)organize • Scalable to Petabyte datasets

Move Mohamed to the mountain, not the mountain to Mohamed.

13

Outline• New Science• Working cross disciplines

– How to help?– 20 questions – WWT example– Alternative: CS Process Models

• Data Demographics and Data Handling• Curation

?Experiments &

Instruments

Simulationsfacts

facts

answers

questions

Literature

Other Archives facts

facts

14

How to Help?

• Can’t learn the discipline before you start(takes 4 years.)

• Can’t go native – you are a CS person not a bio,… person

• Have to learn how to communicateHave to learn the language

• Have to form a working relationship with domain expert(s)

• Have to find problems that leverage your skills

15

Working Cross-Culture A Way to Engage With Domain Scientists

• Communicate in terms of scenarios

• Work on a problem that gives 100x benefit– Weeks/task vs hours/task

• Solve 20% of the problem– The other 80% will take decades

• Prototype

• Go from working-to-working, Always have– Something to show – Clear next steps– Clear goal

• Avoid death-by-collaboration-meetings.

16

Working Cross-Culture -- 20 Questions: A Way to Engage With Domain Scientists

• Astronomers proposed 20 questions• Typical of things they want to do• Each would require a week or more in old way

(programming in tcl / C++/ FTP)• Goal, make it easy to answer questions• This goal motivates DB and tools design

17

The 20 QueriesQ11: Find all elliptical galaxies with spectra that have an

anomalous emission line. Q12: Create a grided count of galaxies with u-g>1 and r<21.5

over 60<declination<70, and 200<right ascension<210, on a grid of 2’, and create a map of masks over the same grid.

Q13: Create a count of galaxies for each of the HTM triangles which satisfy a certain color cut, like 0.7u-0.5g-0.2i<1.25 && r<21.75, output it in a form adequate for visualization.

Q14: Find stars with multiple measurements and have magnitude variations >0.1. Scan for stars that have a secondary object (observed at a different time) and compare their magnitudes.

Q15: Provide a list of moving objects consistent with an asteroid.

Q16: Find all objects similar to the colors of a quasar at 5.5<redshift<6.5.

Q17: Find binary stars where at least one of them has the colors of a white dwarf.

Q18: Find all objects within 30 arcseconds of one another that have very similar colors: that is where the color ratios u-g, g-r, r-I are less than 0.05m.

Q19: Find quasars with a broad absorption line in their spectra and at least one galaxy within 10 arcseconds. Return both the quasars and the galaxies.

Q20: For each galaxy in the BCG data set (brightest color galaxy), in 160<right ascension<170, -25<declination<35 count of galaxies within 30"of it that have a photoz within 0.05 of that galaxy.

Q1: Find all galaxies without unsaturated pixels within 1' of a given point of ra=75.327, dec=21.023

Q2: Find all galaxies with blue surface brightness between and 23 and 25 mag per square arcseconds, and -10<super galactic latitude (sgb) <10, and declination less than zero.

Q3: Find all galaxies brighter than magnitude 22, where the local extinction is >0.75.

Q4: Find galaxies with an isophotal surface brightness (SB) larger than 24 in the red band, with an ellipticity>0.5, and with the major axis of the ellipse having a declination of between 30” and 60”arc seconds.

Q5: Find all galaxies with a deVaucouleours profile (r¼ falloff of intensity on disk) and the photometric colors consistent with an elliptical galaxy. The deVaucouleours profile

Q6: Find galaxies that are blended with a star, output the deblended galaxy magnitudes.

Q7: Provide a list of star-like objects that are 1% rare.Q8: Find all objects with unclassified spectra. Q9: Find quasars with a line width >2000 km/s and

2.5<redshift<2.7. Q10: Find galaxies with spectra that have an equivalent width

in Ha >40Å (Ha is the main hydrogen spectral line.)

Also some good queries at: http://www.sdss.jhu.edu/ScienceArchive/sxqt/sxQT/Example_Queries.html

18

Two kinds of SDSS data in an SQL DB(objects and images all in DB)

300M Photo Objects ~ 400 attributes10B rows overall

400K Spectra

with ~30 lines/ Spectrum

100 M rows

19

An easy one: Q7: Provide a list of star-like objects that are 1% rare.

• Found 14,681 buckets, first 140 buckets have 99% time 104 seconds

• Disk bound, reads 3 disks at 68 MBps.

Select cast((u-g) as int) as ug, cast((g-r) as int) as gr, cast((r-i) as int) as ri, cast((i-z) as int) as iz,count(*) as Population

from starsgroup by cast((u-g) as int), cast((g-r) as int), cast((r-i) as int), cast((i-z) as int) order by count(*)

20

Then What?

1999. 20 Queries were a way to engage– Needed spatial data library – Needed DB design

2000. Built website to publish the data

2001. Data Loading (workflow scheduler).

2002. Pixel web service evolved to

2003. SkyQuery federation evolved to

2004. Now focused on spatial data library. Conversion to OR DB (put analysis in DB).

21

Alternate Model

• Many sciences are becoming information sciences

• Modeling systems needs new and better languages.

• CS modeling tools can help – Bio, Eco, Linguistic, …

• This is the process/program centric view rather than my info-centric view.

22

Outline• New Science• Working cross disciplines• Data Demographics and Data Handling

– Exponential growth– Data Lifecycle– Versions– Data inflation – Year 5– Overprovision by 6x– Data Loading – Regression Tests– Statistical subset

• Curation?

Experiments &Instruments

Simulationsfacts

facts

answers

questions

Literature

Other Archives facts

facts

23

Information Avalanche• In science, industry, government,….

– better observational instruments and – and, better simulations producing a data avalanche

• Examples– BaBar: Grows 1TB/day

2/3 simulation Information 1/3 observational Information

– CERN: LHC will generate 1GB/s .~10 PB/y– VLBA (NRAO) generates 1GB/s today– Pixar: 100 TB/Movie

• New emphasis on informatics:– Capturing, Organizing,

Summarizing, Analyzing, Visualizing

Image courtesy C. Meneveau & A. Szalay @ JHU

BaBar, Stanford

Space Telescope

P&E Gene Sequencer Fromhttp://www.genome.uci.edu/

24

Q: Where will the Data Come From?A: Sensor Applications

• Earth Observation – 15 PB by 2007

• Medical Images & Information + Health Monitoring– Potential 1 GB/patient/y 1 EB/y

• Video Monitoring– ~1E8 video cameras @ 1E5 MBps

10TB/s 100 EB/y filtered???

• Airplane Engines– 1 GB sensor data/flight, – 100,000 engine hours/day– 30PB/y

• Smart Dust: ?? EB/y

http://robotics.eecs.berkeley.edu/~pister/SmartDust/http://www-bsac.eecs.berkeley.edu/~shollar/macro_motes/macromotes.html

25

CERN Tier 0

Instruments: CERN – LHCPeta Bytes per Year

Looking for the Higgs Particle

• Sensors: ~1 GB/s (~ 20 PB/y)

• Events 100 MB/s

• Filtered 10 MB/s

• Reduced 1 MB/s

Data pyramid: 100GB : 1TB : 100TB : 1PB : 10PB

26

Like all sciences, Astronomy Faces an Information Avalanche

• Astronomers have a few hundred TB now– 1 pixel (byte) / sq arc second ~ 4TB– Multi-spectral, temporal, … → 1PB

• They mine it looking for new (kinds of) objects or more of interesting ones (quasars), density variations in 400-D space correlations in 400-D space

• Data doubles every year• Data is public after 1 year• So, 50% of the data is public• Same access for everyone

271998 1999 2000 2001 2002 2003

Year

20000

50000

100000.

200000.

500000.

Count of Spectra Proteomics MS Data Generation

Moore’s Law in Proteomics

Roche Center for Medical Genomics (RCMG): number of mass-spectra acquired for proteomics doubled every year since first mass spectrometer deployed.

R2=0.96

Courtesy of Peter Berndt, Roche Center for Medical Genomics (RCMG)

28

Data Lifecycle

• Raw data → primary data → derived data• Data has bugs:

– Instrument bugs– Pipeline bugs

• Data comes in versions – later versions fix known bugs– Just like software (indeed data is software)

• Can’t “un-publish” bad data.

instrumentor

simulatorpipeline

otherdata

otherdata

pipeline

Level 0raw

Level 1calibrated

Level 2derived

29

Data Inflation – Data Pyramid

Level 1AGrows X TB/year ~ .4X TB/y compressed (level 1A in NASA terms)

Level 2Derived data products ~10x smaller But there are many. L2≈L1

• Publish new edition each year – Fixes bugs in data.– Must preserve old editions– Creates data pyramid

• Store each edition – 1, 2, 3, 4… N ~ N2 bytes

• Net: Data Inflation: L2 ≥ L1

E1

E2

E3E4

4 editions oflevel 1A data(source data)

4 editions of level 2 derived data products. Note that each derived product is small, but they are numerous. This proliferation combined with the data pyramid implies that level2 data more than doubles the total storage volume.

time

Level 1A 4 editions of 4 Level 2 products

30

The Year 5 Problem Yearly Demand

0

20

40

60

80

100

120

140

160

180

0 2 4 6 8 10Year

Yea

rly

Dem

and

( R

)

Depreciated Inflated Demand

Inflated Demand

Naive Demand

Yearly Capital Cost

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

0 2 4 6 8 10Year

Mar

gin

al C

apit

al C

ost

• Data arrives at R bytes/year• New Storage & Processing

– Need to buy R units in year N

• Data inflation means ~N2R– Need to buy NR units

• Depreciate over 3 years– After year 3

need to buy N2R + (N-3)2R

• Moore’s law: 60%/year price decline

• Capital expense peaks at year 5

• See 6x Over-Power slide next

31

6x Over-Power Ratio

• If you think you need X raw capacity, then you probably need 6X

• Reprocessing

• Backup copies

• Versions

• …

• Hardware is cheap, Your time is precious.

PubDB 3.6TB

DR3C 2.4TB

DR2C 1.8TB

DR2M 1.8TB

DR2P 1.8TB

DR3M 2.4TB

DR3P 2.4TB

32

Data Loading• Data from outside

– Is full of bugs– Is not in your format

• Advice– Get it in a “Universal Format”

(e.g. Unicode CSV)– Create Blood-Brain barrier

Quarantine in a “load database”– Scrub the data

• Cross check everything you can• Check data statistics for sanity• Reject or repair bad data• Generate detailed bug reports

(needed to send rejection upstream)

– Expect to reload many times Automate everything!

Test UniquenessOf Primary KeysTest UniquenessOf Primary Keys

TestForeign Keys

TestForeign Keys

TestCardinalities

TestCardinalities

TestHTM IDs

TestHTM IDs

Test parent- childconsistency

Test parent- childconsistency

Test the uniqueKey in each table

Test for consistencyof keys that link tables

Test consistency of numbers of various quantities

Test the HierarchicalTriamgular Mesh IDsused for spatial indexing

Ensure that all parentsand children and linked

Test UniquenessOf Primary KeysTest UniquenessOf Primary Keys

TestForeign Keys

TestForeign Keys

TestCardinalities

TestCardinalities

TestHTM IDs

TestHTM IDs

Test parent- childconsistency

Test parent- childconsistency

Test the uniqueKey in each table

Test for consistencyof keys that link tables

Test consistency of numbers of various quantities

Test the HierarchicalTriamgular Mesh IDsused for spatial indexing

Ensure that all parentsand children and linked

LOADLOAD

PUBLISHPUBLISH

FINISHFINISH

EXPEXP

CHKCHK

BLDBLD

SQLSQL

VALVAL

BCKBCK

DTCDTC

Export

Check CSV

Build Task DBs

Build SQL Schema

Validate

Backup

Detach

PUBPUB

CLNCLN

Publish

Cleanup

FINFIN

LOADLOAD

PUBLISHPUBLISH

FINISHFINISH

EXPEXP

CHKCHK

BLDBLD

SQLSQL

VALVAL

BCKBCK

DTCDTC

Export

Check CSV

Build Task DBs

Build SQL Schema

Validate

Backup

Detach

PUBPUB

CLNCLN

Publish

Cleanup

FINFIN

33

Performance Prediction & Regression

• Database grows exponentially

• Set up response-time requirements – For load– For access

• Define a workload to measure each

• Run it regularly to detect anomalies

• SDSS uses – one-week to reload– 20 queries with response of 10 sec to 10 min.

34

Data Subsets For Science and Development

• Offer 1GB, 10GB, …, Full subsets

• Wonderful tool for you– Debug algorithms

• Good tool for scientists– Experiment on subset– Not for needle in haystack,

but good for global stats• Challenge: How make

statistically valid subsets?– Seems domain specific– Seems problem specific– But, must be some general

concepts.

35

Outline• New Science• Working cross disciplines• Data Demographics and Data Handling• Curation

– Problem statement– Economics – Astro as a case in point

?Experiments &

Instruments

Simulationsfacts

facts

answers

questions

Literature

Other Archives facts

facts

36

Problem Statement• Once published,

scientific data needs to be available forever,so that the science can be reproduced/extended.

• What does that mean?– Data can be characterized as

• Primary Data: could not be reproduced • Derived data: could be derived from primary data.

– Meta-data: how the data was collected/derivedis primary

• Must be preserved • Includes design docs, software, email, pubs, personal

notes, teleconferences,

NASA “level 0”

37

The Core Problem: No Economic Model

• The archive user is not yet born. How can he pay you to curate the data?

• The Scientist gathered data for his own purposeWhy should he pay (invest time) for your needs?

• Answer to both: that’s the scientific method• Curating data

(documenting the design, the acquisition and the processing)

Is difficult and there is little reward for doing it.Results are rewarded, not the process of getting them.

• Storage/archive NOT the problem (it’s almost free)

• Curating/Publishing is expensive,MAKE IT EASIER!!! (lower the cost)

38

Publishing Data

Roles

Authors

Publishers

Curators

Archives

Consumers

Traditional

Scientists

Journals

Libraries

Archives

Scientists

Emerging

Collaborations

Project web site

Data+Doc Archives

Digital Archives

Scientists

39

Changing Roles• Exponential growth:

– Projects last at least 3-5 years– Project data online during project lifetime.– Data sent to central archive only at the end of the project– At any instant, only 1/8 of data is in central archives

• New project responsibilities– Becoming Publishers and Curators– Larger fraction of budget spent on software

• Standards are needed– Easier data interchange, fewer tools

• Templates are needed– Much development duplicated, wasted

40

What SDSS is Doing: Capture the Bits(preserve the primary data)

• Best-effort documenting data and processDocuments and data are hyperlinked.

• Publishing data: often by UPS(~ 5TB today and so ~5k$ for a copy)

• Replicating data on 3 continents.• EVERYTHING online (tape data is dead data)

• Archiving all email, discussions, ….

• Keeping all web-logs & query logs.

• Now we need to figure out how to organize/search all this metadata.

41

Schema (aka metadata)• Everyone starts with the same schema

<stuff/>Then the start arguing about semantics.

• Virtual Observatory: http://www.ivoa.net/

• Metadata based on Dublin Core:http://www.ivoa.net/Documents/latest/RM.html

• Universal Content Descriptors (UCD): http://vizier.u-strasbg.fr/doc/UCD.htxCaptures quantitative concepts and their unitsReduced from ~100,000 tables in literature to ~1,000 terms

• VOtable – a schema for answers to questionshttp://www.us-vo.org/VOTable/

• Common Queries:Cone Search and Simple Image Access Protocol, SQL

• Registry: http://www.ivoa.net/Documents/latest/RMExp.htmlstill a work in progress.

42

Summary• New Science

– X-Info for all fields X– WWT as an example– Big Picture– Puzzle– Hitting the wall– Needle in haystack– Move queries to data

• Working cross disciplines– How to help?– 20 questions – WWT example– Alt: CS Process Models

• Data Demographics– Exponential growth– Data Lifecycle– Versions– Data inflation – Year 5 is peak cost– Overprovision by 6x– Data Loading – Regression Tests– Statistical subset

• Curation– Problem statement– Economics – Astro as a case in point

43

Call to Action

• X-info is emerging.

• Computer Scientists can help in many ways.– Tools– Concepts– Provide technology consulting to the community

• There are great CS research problems here– Modeling– Analysis– Visualization– Architecture

44

References http://SkyServer.SDSS.org/http://research.microsoft.com/pubs/

http://research.microsoft.com/Gray/SDSS/ (download personal SkyServer)

Extending the SDSS Batch Query System to the National Virtual Observatory Grid,

M. A. Nieto-Santisteban, W. O'Mullane, J. Gray, N. Li, T. Budavari, A. S. Szalay, A. R. Thakar, MSR-TR-2004-12, Feb. 2004

Scientific Data Federation, J. Gray, A. S. Szalay, The Grid 2: Blueprint for a New Computing Infrastructure, I. Foster, C. Kesselman, eds, Morgan Kauffman, 2003, pp 95-108.

Data Mining the SDSS SkyServer Database, J. Gray, A.S. Szalay, A. Thakar, P. Kunszt, C. Stoughton, D. Slutz, J. vandenBerg , Distributed Data & Structures 4: Records of the 4th International Meeting, pp 189-210, W. Litwin, G. Levy (eds),, Carleton Scientific 2003, ISBN 1-894145-13-5, also MSR-TR-2002-01, Jan. 2002

Petabyte Scale Data Mining: Dream or Reality?, Alexander S. Szalay; Jim Gray; Jan vandenBerg, SIPE Astronomy Telescopes and Instruments, 22-28 August 2002, Waikoloa, Hawaii, MSR-TR-2002-84

Online Scientific Data Curation, Publication, and Archiving, J. Gray; A. S. Szalay; A.R. Thakar; C. Stoughton; J. vandenBerg, SPIE Astronomy Telescopes and Instruments, 22-28 August 2002, Waikoloa, Hawaii, MSR-TR-2002-74

The World Wide Telescope: An Archetype for Online Science, J. Gray; A. Szalay,, CACM, Vol. 45, No. 11, pp 50-54, Nov. 2002, MSR TR 2002-75,

The SDSS SkyServer: Public Access To The Sloan Digital Sky Server Data, A. S. Szalay, J. Gray, A. Thakar, P. Z. Kunszt, T. Malik, J. Raddick, C. Stoughton, J. vandenBerg:, ACM SIGMOD 2002: 570-581 MSR TR 2001 104.

The World Wide Telescope, A.S., Szalay, J., Gray, Science, V.293 pp. 2037-2038. 14 Sept 2001. MS-TR-2001-77

Designing & Mining Multi-Terabyte Astronomy Archives: Sloan Digital Sky Survey, A. Szalay, P. Kunszt, A. Thakar, J. Gray, D. Slutz, P. Kuntz, June 1999, ACM SIGMOD 2000, MS-TR-99-30,

45

46

47

How to Publish Data: Web Services

• Web SERVER:– Given a url + parameters – Returns a web page (often dynamic)

• Web SERVICE:– Given a XML document (soap msg)– Returns an XML document (with schema)– Tools make this look like an RPC.

• F(x,y,z) returns (u, v, w)

– Distributed objects for the web.– + naming, discovery, security,..

• Internet-scale distributed computing

Yourprogram

DataIn your address

space

Web Service

soap

object

in

xml

Yourprogram Web

Server

http

Web

page

48

Federation

Global Federations• Massive datasets live near their owners:

– Near the instrument’s software pipeline– Near the applications– Near data knowledge and curation

• Each Archive publishes a (web) service– Schema: documents the data– Methods on objects (queries)

• Scientists get “personalized” extracts• Uniform access to multiple Archives

– A common global schema

49

The Sloan Digital Sky Survey• Goal

– Create the most detailed map of the Northern Sky to-date

• 2.5m telescope– 3 degree field of view

• Two surveys in one– 5-color images of ¼ of the sky– Spectroscopic survey of a million

galaxies and quasars

• Very high data volume– 40 Terabytes of raw data– 10 Terabytes processed– All data public

The University of Chicago Princeton University The Johns Hopkins University The University of Washington New Mexico State University University of Pittsburgh Fermi National Accelerator Laboratory US Naval Observatory The Japanese Participation Group The Institute for Advanced Study Max Planck Inst, Heidelberg

Sloan Foundation, NSF, DOE, NASA

The University of Chicago Princeton University The Johns Hopkins University The University of Washington New Mexico State University University of Pittsburgh Fermi National Accelerator Laboratory US Naval Observatory The Japanese Participation Group The Institute for Advanced Study Max Planck Inst, Heidelberg

Sloan Foundation, NSF, DOE, NASA

50

SkyServer• A multi-terabyte database

• An educational website– More than 50 hours of educational exercises– Background on astronomy– Tutorials and documentation– Searchable web pages

• Easy astronomer access to SDSS data.

• Prototype eScience lab

• Interactive visual tools for data exploration

http://skyserver.sdss.org/

51

Demo SkyServer

• atlas• education project• Mouse in pixel space• Explore an object

(record space)• Explore literature• Explore a set• Pose a new question

52

SkyQuery (http://skyquery.net/)

• Distributed Query tool using a set of web services• Many astronomy archives from

Pasadena, Chicago, Baltimore, Cambridge (England)• Has grown from 4 to 15 archives,

now becoming international standard

• Allows queries like:SELECT o.objId, o.r, o.type, t.objId FROM SDSS:PhotoPrimary o,

TWOMASS:PhotoPrimary t WHERE XMATCH(o,t)<3.5

AND AREA(181.3,-0.76,6.5) AND o.type=3 and (o.I - t.m_j)>2

532MASS

INT

SDSS

FIRST

SkyQueryPortal

ImageCutout

Demo SkyQuery Structure

• Each SkyNode publishes – Schema Web Service– Database Web Service

• Portal is – Plans Query (2 phase) – Integrates answers– Is itself a web service

54

MyDB: eScience Workbench

• Prototype of bringing analysis to the data

• Everybody gets a workspace (database)– Executes analysis at the data– Store intermediate results there– Long queries run in batch– Results shared within groups

• Only fetch the final results

• Extremely successful – matches work patterns

55

National Center Biotechnology Information (NCBI) A good Example

• PubMed:– Abstracts and books and..

• GenBank:– All Gene sequences deposited– BLAST and other searches– Website to explore data and literature

• Entrez: – unifies many databases

with literature (books, journals,..)– Organizes the data

56

Making Discoveries

• Where are discoveries made?– At the edges and boundaries– Going deeper, collecting more data, using more colors….

• Metcalfe’s law: quadratic benefit– Utility of computer networks grows as the

number of possible connections: O(N2)

• Data Federation: quadratic benefit– Federation of N archives has utility O(N2) – Possibilities for new discoveries grow as O(N2)

• Current sky surveys have proven this– Very early discoveries from SDSS, 2MASS, DPOSS

57Federation

Global Federations

• Massive datasets live near their owners:– Near the instrument’s software pipeline– Near the applications– Near data knowledge and curation

• Each Archive publishes a (web) service– Schema: documents the data– Methods on objects (queries)

• Scientists get “personalized” extracts

• Uniform access to multiple Archives– A common global schema

58

The OGIS model

Ingest Archive

Data Management

Administer

AccessProducer Consumer

59

Jim’s Model of Library Science

• Alexandria

• Gutenberg

• (Melvil) Dewey Decimal

• MARC (Henriette Avram)

• Dublin Core

Yes, I know there have been other things.

60

Dublin CoreElements

– Title– Creator– Subject– Description– Publisher– Contributor– Date– Type– Format– Identifier– Source– Language– Coverage– Rights

Elements+– Audience– Alternative– TableOfContents– Abstract– Created– Valid– Available– Issued– Modified– Extent– Medium– IsVersionOf– HasVersion– IsReplacedBy– Replaces– IsRequiredBy– Requires– IsPartOf– HasPart– IsReferencedBy– References– IsFormatOf– HasFormat– ConformsTo– Spatial– Temporal– Mediator– DateAccepted– DateCopyrighted– DateSubmitted– EducationalLevel– AccessRights– BibliographicCitation

Encoding– LCSH (Lb. Congress Subject Head)– MESH (Medical Subject Head)– DDC (Dewey Decimal Classification)– LCC (Lb. Congress Classification)– UDC (Universal Decimal Classification)– DCMItype (Dublin Core Meta Type)– IMT (Internet Media Type)– ISO639-2 (ISO language names)– RFC1766 (Internet Language tags)– URI (Uniform Resource Locator)– Point (DCMI spatial point)– ISO3166 (ISO country codes)– Box (DCMI rectangular area)– TGN (Getty Thesaurus of Geo Names)– Period (DCMI time interval)– W3CDTF (W3C date/time)– RFC3066 (Language dialects)

Types– Collection– Dataset– Event– Image– InteractiveResouce– Service– Software– Sound– Text– PhysicalObject– StillImage– MovingImage

61

Access Challenges

• Archived information “rusts” if it is not accessed. Access is essential.

• Access costs money – who pays?

• Access sometimes uses IP, who pays?

• There are also technical problems:– Access formats are different from the storage

formats. • migration? • emulation? • Gold Standards?)

62

Ingest Challenges

• Push vs Pull

• What are the gold standards?

• Automatic indexing, annotation, provenance.

• Auto-Migration (Format conversion)

• Version management

• How capture time varying sources

• Capture “dark matter” (encapsulated data)– Bits don’t “rust” but applications do.

63

Archive Challenges• Cost of administering storage:

– Presently 10x to 100x the hardware cost.

• Resist attack: geographic diversity • At 1GBps it takes 12 days to move a PB• Store it in two (or more) places online (on disk).

A geo-plex• Scrub it continuously (look for errors)

• On failure, – use other copy until failure repaired, – refresh lost copy from safe copy.

• Can organize the copies differently (e.g.: one by time, one by space)

64

The Midrange Paradox• Large archives are curated

– Curated by projects

• Small archives are appendices to papers– Curated by journals

• Medium-sized archives are in limbo– No place to register them– No one has mandate to preserve them

• Example: – Your website with your data files– Small scale science projects– Genbank gets the sequence

but not the software or analysis that produced it.