Scientific Workflows: Even More eScience Mileage from Cyberinfrastructure

1
Scientific Workflows (Even) More e-Science
Mileage from Cyberinfrastructure

• Bertram Ludäscher
• Dept. of Computer Science
• Genome Center
• University of California, Davis
• ludaesch_at_ucdavis.edu

2
SUMMARY (first things first, but
not necessarily in that order)

• Scientific workflows are CI upper-ware for
  eScience
• Scientific workflows are universal (every-ware)
  and there are many interesting technical
  challenges
• execution models, semantic extensions,
  provenance,
• modeling and design!
• Workflow Thinking!
• Collection-Oriented Modeling Design (COMAD) is
  good for you! (YMMV)
• Consilience The Holy eScience Grail

3
Science has been changing lately

• All science is either physics or stamp
  collecting.
• Ernest Rutherford, British chemist
  physicist (1871 - 1937)
• J. B. Birks "Rutherford at Manchester (1962)
• From few data, lots of thinking
• to lots of data, lots of analysis
• ... and hopefully still lots of thinking!

4
The Diversity Unity of Science
Natural Sciences

Earth Sciences
Life Sciences
Physical Sciences
Observations, Measurements, Models, Simulations,
Analyses, Hypotheses Understanding, Prediction
in vivo, in vitro, in situ, in silico,
compute-intensive
structurally semantics -intensive
data-intensive
metadata-intensive
5
Scientific workflows are CI upper-ware, i.e.
the scientists way to harness
cyberinfrastructure

• Domain Scientists View
• Q When is CI (middle-ware, under-ware) good?
• A When I cant see it!
• Q When is a scientific workflow tool (CI
  upper-ware) good?
• A When I can get more, new, faster, better
  science done!
• Workflow Engineers View
• How can I (help the scientist) design implement
  the desired wfs?
• How does wf make my life easier? Beyond Perl,
  Python?
• Choice of platforms, standards reuse of existing
  tools, semantic extensions, scheduling on the
  Grid?
• How do I make all of this robust, fault-tolerant,
  etc.
• Computer Scientists View
• workflow language design, static analysis,
  optimization, theoretical limits what can /
  cant be done
• The quest for the right language(s)
• The holy grail of eScience Join the Quest!

6
Science Environment for Ecological Knowledge
(SEEK)
7
Scientific Workflow

• Capture how a scientist works with data and
  analytical tools
• data access, transformation, analysis,
  visualization
• possible worldview dataflow-oriented (cf.
  signal-processing)
• Scientific workflow (wf) benefits (compare w/
  script-based approaches)
• wf automation
• wf component reuse
• wf design, documentation
• wf archival, sharing
• built-in concurrency
• (task-, pipeline-parallelism)
• built-in provenance support
• distributed execution
• (Grid) support

8
Ex SEEK Ecological Niche Modeling Pipeline

• Scientific Workflow paradigm
• Reusable components (actors) a scientists
  verbs/actions
• Top-level workflows conceptual representation
  of the science process, sentences in the
  scientists language
• Sub-workflows increasing levels of detail
• Separation of concerns
• actors what to do
• parameters configurable behavior
• channels dataflow, pipeline composition
• directors fix execution model, scheduling
• semantic types smart discovery, linking

D Pennington, D Higgins, AT Peterson, M Jones, B
Ludaescher, S Bowers. Ecological Niche Modeling
using the Kepler Workflow System. Workflows for
e-Science, Springer.
9
Simple Kepler workflow using R (a statistics
package)
10
Plumbing with Style (Norbert Podhorszki UC
Davis, Scott Klasky ORNL)
Monitor

• Plasma physics simulation on 2048 processors on
  Seaborg_at_NERSC (LBL)
• Gyrokinetic Toroidal Code (GTC) to study energy
  transport in fusion devices (plasma
  microturbulence)
• Generating 800GB of data (3000 files, 6000
  timesteps, 267MB/timestep), 30 hour simulation
  run
• Under workflow control
• Monitor (watch) simulation progress (via remote
  scripts)
• Transfer from NERSC to ORNL concurrently with the
  simulation run
• Convert each file to HDF5 file
• Archive files to 4GB chunks into HPSS

11
Kepler and Sensor Networks

• These ones just in (new NSF CEOP projects)
• Management and Analysis of Environmental
  Observatory Data using the Kepler Scientific
  Workflow System, NCEAS, SDSC, UC Davis, OSU,
  CENS (UCLA), OPeNDAP
• standardize services for sensor networks, support
  multiple views, protocols
• COMET Coast-to-Mountain Environmental Transect,
  UC Davis, Bodega Marine Lab, Lake Tahoe Research
  Center
• study how environmental factors affect ecosystems
  along an elevation gradient from coastal
  California to the summit of the Sierra Nevada

CEOP/COMET
CEOP/Kepler
12
Workflow Thinking (cf. Computational
Thinking)

• How should we think about scientific workflows?
• From What scientists do to produce scientific
  papers
• to its just a program
• Is that helpful?
• Depends on who you ask! What are you trying to
  do?
• (Domain) Scientist ? Workflow Engineer ? Computer
  Scientist

13
Our Starting Point Actor-Oriented Modeling

• Ports
• each actor has a set of input and output ports
• denote the actors signature
• produce/consume data (a.k.a. tokens)
• parameters are special static ports

14
Actor-Oriented Modeling

• Dataflow Connections
• unidirectional actor communication channels
• connect output ports with input ports
• for composing analysis pipelines

15
Actor-Oriented Modeling

• Sub-workflows / Composite Actors
• composite actors wrap sub-workflows
• like actors, have signatures (i/o ports of
  sub-workflow)
• hierarchical workflows (arbitrary nesting levels)

16
Actor-Oriented Modeling

• Directors
• define the execution semantics of workflow graphs
• executes workflow graph (some schedule)
• sub-workflows may have different directors
• promotes reusability

17
Models of Computation (A Wf Engineers Issue)

• Directors separate the concerns of orchestration
  and scheduling from conceptual design
• Synchronous Dataflow (SDF)
• Statically analyzable schedule, no deadlocks,
  fixed buffer requirements executable as a single
  thread by the director.
• Process Networks (PN)
• Generalizes SDF. Actors execute as separate
  threads/processes, with queues of unbounded size
  (Kahn/MacQueen networks).
• Directed Acyclic Graph (DAG)
• Special case of SDF. No loops, no pipelining.
• Continuous Time (CT)
• Connections represent the value of a continuous
  time signal at some point in time ... Often used
  to model physical processes.
• Discrete Event (DE)
• Actors communicate through a queue of events in
  time. Used for instantaneous reactions in
  physical systems.

18
Everything is a service / actor
( yeah right)
19
Scientific Workflow Design Challenges
And thats why our scientific workflows are
much easier to develop, understand and maintain!
20
Shimology Part 1 Structure Semantics

• Components and their i/o ports typically have
• Explicit structural type
• e.g., int, float, string, array.... of
  double,
• Implicit semantic type
• Not sure whether the stream of values from a port
  represents rainfall values or body size values

21
Semantic Annotation

• Label data with semantic types
• Label inputs and outputs of analytical components
  with semantic types (and overall component
  function)

22
Semantic Type Annotation in Kepler

• Component input and output port annotation
• a port can be annotated with multiple concepts
  from multiple ontologies
• Annotations are stored with the actor metadata

23
Component Annotation and Indexing

• Component Annotations
• New components can be annotated and indexed into
  the actor library
• (specializing generic actors)
• Existing components can be revised, annotated,
  and indexed (hiding previous versions)

24
Smart Discovery

• Find a component (here an actor) in different
  locations (categories)
• based on the semantic annotation of the
  component (or its ports)

25
Smart Linking (Workflow Design)

• Statically perform semantic and structural type
  checking

• Navigate errors and warnings within the workflow
• !! Search for and insert adapters (aka shims) to
  fix (structural and semantic) mismatches

26
Smart Linking (addressing Shimology
Type 1)
Source Bowers-Ludaescher, DILS04
27
Scientific Workflow Design More Challenges
And thats why our scientific workflows are
much easier to develop, understand and maintain!
28
Behold the Beauty of Scientific Workflow Design
Author Kristian Stevens, UC Davis
29
Shimology Part 2 the ugly truth inside
Author Kristian Stevens, UC Davis
30
And now a Real Example! (ChIP-chip workflow)

• Fear Not!
• Higher-order functions (map, fold, zipWith, )
  shall oil your gears , e.g.,
• map f x1,,xn ? f(x1), , f(xn)
• MAP f1, , fk (x) ? f1(x), , fk(x)
• Your analysis pipeline shall be automated,
  your data lineage shall be recorded, and
  your provenance explained!

31
But how do we get from messy to neat reusable
designs?
32
The Answer (YMMV)

• Collection-Oriented Modeling Design (COMAD)
• embrace the assembly line metaphor fully
• ? cf. Flow-based Programming (J. Morrison)
• data tagged nested collections
• e.g. represented as flattened, pipelined
• (XML) token streams

33
How does COMAD work?

• Some COMAD principles
• data tagged, flattened, nested collections
  (token streams)
• data tokens
• metadata tokens
• inherited downwards into (sub)collections
• define an actors read scope via an (X)Path-like
  expression
• default actor behavior
• not mine?
• ? dont do anything just pass the buck!
• stuff within my scope? ?
• add-only to it (default)
• consume scope write-out result
• (but remember the bypass!)
• iteration scope is a query involving group-by and
  further refines the granularity/subtrees that
  constitute the tokens consumed by an actor firing
  
• has aspects of implicit iteration (a la Taverna)
• default iteration level to fix signature
  mismatches
• but also
• granularity/grouping is definable
• works on anything (assuming scope is matched
  correctly)

• T McPhillips, S Bowers. Pipelining Nested Data
  Collections in Scientific Workflows. SIGMOD
  Record, 2005.
• T McPhillips, S Bowers, B Ludaescher.
  Collection-Oriented Scientific Workflows for
  Integrating and Analyzing Biological Data.
  Workshop on Data Integration in the Life Sciences
  (DILS), 2006

34
COMAD What we gained

• from fragile, messy workflow designs
• to more reusable actors
• just change the read/iteration scope parameters!
• sometimes not even that is needed (working on
  that )
• and cleaner workflow design (The A-B-C method
  of wf design!)
• Crux keep the nesting structure of data (pass
  through, add-only)
• and let it drive the (semi-)implicit iteration
  (aka structural recursion )

35
Workflow Thinking Modeling Design Paradigms

• Vanilla Process Network
• Functional Programming Dataflow Network
• XML Transformation Network
• Collection-oriented Modeling Design framework
  (COMAD)

36
A Scientific Publication (the final
provenance frontier )
Title (Statement, Theorem)
Abstract (1st-Level- Expansion)
Main Text (2nd-Level Expansion)
Nature 443, 167-172(14 September 2006)
doi10.1038/nature05113 Received 27 June 2006
Accepted 25 July 2006 Published online 16 August
2006
some metadata
37
More Evidence
data reference
type of evidence
tool reference
trust me on this one

• provenance/lineage show the history and evidence
• related to proof trees
• unlike w/ scripts, SWF system can keep track of
  what happened
• In the future deposit your data workflows in a
  repository

38
Pipelined workflow for inferring phylogenetic
trees
Author Tim McPhillips, UC Davis
39
Scientific Provenance Questions we can ask

• What DNA sequences were input to the workflow?
• What phylogenetic trees were output by the
  workflow?
• What DNA sequences input to the workflow does
  this consensus tree depend on?
• What input sequences were not used to derive any
  output consensus trees?
• What was the sequence alignment (key intermediate
  data) used in the process of inferring this tree?
• plus the usual smart-rerun, VCR replay,

40
Provenance in the COMAD Framework
Without Provenance
With Provenance
41
So what should we focus on?

• What is the bottleneck in Scientific Workflows?
• The human resource workflow design support!
• includes
• new modeling paradigms (e.g. COMAD, FP, NRC, )
• and data-orientation!
• Business workflows
• top-down, engineered, many times the same
• Scientific workflows
• bottom-up, exploratory, each time unique
• Combine best of both
• explore, capture, evolve!
• workflow sharing and reuse

42
Workflow design when was the last time

• that we ate our own dog food?
• Do we really want to use formalism X for
  scientist-oriented workflow design?
• X in Petri-net, ?-calculus, process networks,
  Turing machines, BPEL4WS,
• What are the observables of approach/language X?
• What does language X talk about, ignore, and
  allow in terms of analysis, understanding?
• Example Data Provenance in Scientific Workflows
• T R I M
• Trace (MoP) Run (MoC) I(gnored) M(odeled)

43
The Emperors Old Clothes

• Computer Science / Thin approach
• Minimize to the max Lambda calculus, Turing
  machines, Register machines, Petri nets, Kahn
  Process Networks, Relational Algebra Calculus,
  
• Thick approach
• Algol68, PL/1, XML Schema, BPEL4WS, SQL, (bloat
  to the max?)
• Premature optimization
• is the root of all evil
• Tony Hoare, Donald Knuth
• Premature standardization
• is the root of all evil

44
The Evolution of Language
Source Phil Wadler, Peter Bunemann (?)
45
Consilience The Unity of Knowledge (E. O. Wilson)

• "Literally a jumping together of knowledge by the
  linking of facts and fact-based theory across
  disciplines to create a common groundwork for
  explanation." E.O.Wilson
• eScience, Cyberinfrastructure mechanisms to make
  progress
• Scientific Workflows crucial elements to get the
  most mileage out of CI to fuel eScience,
  accelerating knowledge discovery
• Identify the real bottlenecks in this quest!
• Need workflow engineers, computer scientists,
  bioinformaticians, hybrids!

46
The Holy Grail of eScience / Scientific Workflows

• Evolution programmed us to enjoy certain things
• We should feel lucky
• the brain is so powerful a control system that
  self-conscience emerged now we also enjoy
  thinking
• hence weve been asking provenance questions
  since the dawn of man (where from? to? why?)
• Science (and now eScience) yield answers
• aside so does religion but only science is
  strongly constrained by reality
• We are an intelligent species and the use of our
  intelligence quite properly gives us pleasure. In
  this respect the brain is like a muscle. When we
  think well, we feel good. Understanding is a kind
  of ecstasy. Carl Sagan
• Call to Arms/Ploughshares
• Join the Quest for the right language for
  eScience Workflow Thinking!

47
Acknowledgements, QA

• Data and Knowledge Systems Lab (DAKS) _at_ UC Davis
• Dr. Shawn Bowers, Dr. Timothy McPhillips, Dr.
  Norbert Podhorszki
• Dave Thau, Daniel Zinn, Alex Chen
• Many Kepler collaborators
• Ilkay Altintas (SDSC/UCSD), Matt Jones (UCSB),
  Arie Shoshani (LBL), Terence Critchlow (LLNL),
  Mladen Vouk (NCSU),

48
Some Related Publications

• Semantic Type Annotation
• S Bowers, B Ludaescher. A Calculus for
  Propagating Semantic Annotations through
  Scientific Workflow Queries. ICDE Workshop on
  Query Languages and Query Processing (QLQP),
  LNCS, 2006.
• S Bowers, B Ludaescher. Towards Automatic
  Generation of Semantic Types in Scientific
  Workflows. International Workshop on Scalable
  Semantic Web Knowledge Base Systems (SSWS), WISE
  2005 Workshop Proceedings, LNCS, 2005.
• C Berkley, S Bowers, M Jones, B Ludaescher, M
  Schildhauer, J Tao. Incorporating Semantics in
  Scientific Workflow Authoring. SSDBM, 2005.
• B Ludaescher, K Lin, S Bowers, E Jaeger-Frank, B
  Brodaric, C Baru. Managing Scientific Data From
  Data Integration to Scientific Workflows. GSA
  Today, Special Issue on Geoinformatics, 2006.
• S Bowers, D Thau, R Williams, B Ludaescher. Data
  Procurement for Enabling Scientific Workflows On
  Exploring Inter-Ant Parasitism. VLDB Workshop on
  Semantic Web and Databases (SWDB), 2004.
• S Bowers, K Lin, B Ludaescher. On Integrating
  Scientific Resources through Semantic
  Registration. SSDBM, 2004.
• S Bowers, B Ludaescher. An Ontology-Drive
  Framework for Data Transformation in Scientific
  Workflows. International Workshop on Data
  Integration in the Life Sciences (DILS), LNCS,
  2004.
• S Bowers, B Ludaescher. Towards a Generic
  Framework for Semantic Registration of Scientific
  Data. International Semantic Web Conference
  Workshop on Semantic Web Technologies for
  Searching and Retrieving Scientific Data, 2003.
• Workflow Design and Modeling
• T McPhillips, S Bowers, B Ludaescher.
  Collection-Oriented Scientific Workflows for
  Integrating and Analyzing Biological Data.
  Workshop on Data Integration in the Life Sciences
  (DILS), LNCS, 2006.
• S Bowers, T McPhillips, B Ludaescher, S Cohen, SB
  Davidson. A Model for User-Oriented Data
  Provenance in Pipelined Scientific Workflows.
  International Provenance and Annotation Workshop
  (IPAW), LNCS, 2006.
• S Bowers, B Ludaescher, AHH Ngu, T Critchlow.
  Enabling Scientific Workflow Reuse through
  Structured Composition of Dataflow and
  Control-Flow. IEEE Workshop on Workflow and Data
  Flow for Scientific Applications (SciFlow), 2006.
• S Bowers, B Ludaescher. Actor-Oriented Design of
  Scientific Workflows. International Conference on
  Conceptual Modeling (ER), LNCS, 2005.
• T McPhillips, S Bowers. Pipelining Nested Data
  Collections in Scientific Workflows. SIGMOD
  Record, 2005.
• Kepler
• D Pennington, D Higgins, AT Peterson, M Jones, B
  Ludaescher, S Bowers. Ecological Niche Modeling
  using the Kepler Workflow System. Workflows for
  e-Science, Springer-Verlag, to appear.
• W Michener, J Beach, S Bowers, L Downey, M Jones,
  B Ludaescher, D Pennington, A Rajasekar, S
  Romanello, M Schildhauer, D Vieglais, J Zhang.
  SEEK Data Integration and Workflow Solutions for
  Ecology. Workshop on Data Integration in the Life
  Sciences (DILS), LNCS, 2005.
• S Romanello, W Michener, J Beach, M Jones, B
  Ludaescher, A Rajasekar, M Schildhauer, S Bowers,
  D Pennington. Creating and Providing Data
  Management Services for the Biological and
  Ecological Sciences Science Environment for
  Ecological Knowledge. SSDBM, 2005.

49
Kepler Collaboration

• Open-source
• Builds on Ptolemy II from UC Berkeley
• Contributors from
• SEEK
• SciDAC SDM
• Ptolemy
• GEON
• ROADNet
• Resurgence
• AToL CIPRES, POD
• Goals
• Create powerful analytical tools that are useful
  across disciplines
• Ecology, Biology, Engineering, Geology, Physics,
  Chemistry, Astronomy,

Ptolemy II
Natural Diversity Discovery Project
50
Coming up Kepler Actor Repository
51
Workflow and component repositories myExperiment
is your library, or Our
Workflow Repository!

• Taverna Repository Kepler
  Repository
• Workflow system speciation,
• looking for a new symbiotic relation --- need
  enzymes!
• (yes, thats for the rest of us )