Tracing Lineage Beyond Relational Operators

Size: px

Start display at page:

Download "Tracing Lineage Beyond Relational Operators"

Edward Nathaniel Byrd
6 years ago
Views:

1 Tracing Lineage Beyond Relational Operators Mingwu Zhang 1 Xiangyu Zhang 1 Xiang Zhang 2 Sunil Prabhakar 1 1 Computer Science 2 Bindley Bioscience Center Purdue University

2 Introduction Lineage (Data Provenance) is defined as description of the origin of the data and the process by which the data is derived. Lineage is critical for determining data quality and reliability (e.g. biological data, data cleansing) mandated by law (e.g. audit trails for FDA) essential for data dissemination and reproduction Informative (e.g. querying lineage) Database support for tracing lineage is urgent

3 Lineage tracing Input File 1 File 2 Program X Coarse-grained Output File 2 File 4 e.g Workflow level lineage Input File 1 T1,1,2 T2,3,4 Program X Fine-grained Output File 2 R1, 14 R2, 12

4 Tracing fine-grained lineage Input T1,1,2 T2,3,4 Relational Operators State-of-the-art is limited to three Internal special cases Reversible Mathematical Functions Array Manipulation Language (AML) External Output R1,12 R2, 34 Currently no known techniques for general fine-grained lineage tracing??? Cui et al Woodruff & stonebraker A.P.Marathe

5 Contributions Enable fine-grained external lineage tracing for any arbitrary program without requiring Domain expertise Understanding the semantics of the operation Source code Computed lineage is accurate (no false positives) Lineage is derived directly from program execution.

6 Outline Introduction Lineage tracing Case study Conclusion

7 Our approach Automatically trace lineage using only binary executables. Monitor the data flow during program execution. As the program is executed, the lineage of each variable is traced. Each binary instruction is modified to keep track of the data/control dependencies generated by the instruction.

8 Tracing Lineage A statement S data depends on another statement t if and only if a variable is defined at t and then used at S. A statement S control depends on a predicate statement t if and only if the execution of S is the result of the branch outcome of t Definition: Given a program execution, the data lineage of variable v at an execution point of S i, denoted as DL(v@S i ), is the set of input items that are directly or indirectly involved in the computation of the value v at S i

9 Tracing example 1: y=3; 2: a1=3; 3: b1=4; 4: if(y>2) 5: x = a1 + b1; DL( x@5) = ( DL( a1@5) Υ DL( b1@5)) Υ DL(4) = DL( a1@ 2) Υ DL( b1@3) Υ DL( 4) = DL( a1@ 2) Υ DL( b1@3) Υ DL( At Statement 5, x depends upon a1,b1 (data dependency) and y (control dependence) Thus, the lineage of x is the union of the lineages of a1, b1 and y

10 Deriving lineage Let s i : dest =? t j : f( use, use 1,, use n ) be the executed statement instance s i, which assigns a value to variable dest by using variables use, use 1,, use n and s i control depends on t j. Let DEF(x) be the latest statement instance that defines x. U DL(dest@s i ) = ( DL(use i ))U DL(t j ) x = DL(t j )U( U DL(use x )) x.def(use x ) φ U( U {use x })) x.def(use x )=φ

11 Instrumenting the code We use an open-source instrumentation kernel called Valgrind The lineage is typically set data and is stored using a structure called robdd which is optimized for set operations. Shadow memory (SM) stores the lineage sets associated with variables in stack/heap and shadow register file (SRF) stores lineage sets for variables in registers

12 Instrumenting the code Example instrumentation A=(int*) malloc(1) SM(A) = malloc_in_shadow(1) add (x884dc), eax mov SM[x884dc] U SRF(eax), SRF(eax)

13 Architecture We have developed a prototype system based on Valgrind engine to trace finegrained lineage code x86 binary input Valgrind Kernel instrumented code event Instrumenter Runtime bdd bdd robdd output Lineage Repository lineage

14 Outline Introduction Lineage tracing Case study Conclusion

15 Case study: Cancer biomarker discovery Cancer Digestion Isotope Labeling Label1 Normal Digestion Isotope Labeling Label2 1:1 mix LC-MS Doublet Doublet De-Isotope

16 De-isotoping Seq1: ATLNELVEYVSTNR Figure to show challenge of de-isotoping

17 De-isotoping intenisty

18 De-isotoping Seq1: ATLNELVEYVSTNR Seq2: ITCAELR intenisty

19 De-isotope algorithm Complex, mostly heuristics Numerous parameters picked by experts Validity of results can be affected by choice of parameters Identifying a reverse function is impossible, even for experts. Using state-of-the-art algorithm.

20 De-isotope result 25 Pep 1 (H) (ν) Pep 1 (L) (µ) α β γ δ ε Pep 2 (L) (ξ) ζ η θ Pep 2 (H) (ο) ι κ λ

21 Fine-grained lineage Pep 1 (H) (ν) α β γ δ ε ζ η θ ι κ λ Pep 1 (L) (µ) Pep 2 (L) (ξ) Pep 2 (H) (ο)

22 Case study External fine-grained lineage is crucial for our biomarker discovery application Our technique enabled experts to detect errors assess the reliability of data identify false positives identify program limitations

23 Detect error β α γ δ ε ζ η θ ι κ λ µ ν ξ ο π ρ ς σ +4 τ +4 υ +4 φ α β γ δ ε ζ η θ ι ν µ λ κ ξ ο π ρ ς σ +3 τ +3 υ +3 φ

24 Identifying false positives δ 15 α β ε 1 5 γ ζ η θ ι κ λ

25 Program limitations ω ο ι η θ ρ λ α β γ κ δ ε ζ µ ν ξ π ς σ τ υ +2 φ +1 χ +1 ψ ω χ +2 ω υ +1 Ω +1 φ +1 χ +1 ψ υ +2 ψ

26 Performance cluto deisotope apriori rainbow lemur Image processing Auto-class Tracing/ valgrind Tracing (sec) Valgrind (sec) Original (sec) benchmark

27 Memory consumption Benchmark Orig(MB) BDD(MB) Tracing(MB) Auto-class Image processing lemur rainbow apriori deisotope cluto

28 Outline Introduction Tracing lineage Case study Conclusion

29 Conclusion This is the first work that can trace external general fine-grained lineage Advantages Fully automated Does not require user input or domain knowledge Does not need source code The lineage is accurate

30 Conclusion Disadvantage Performance Tracing lineage incurs a slowdown but acceptable for applications in need of lineage Part of the overhead is caused by Valgrind, other industrial instrumentation engines such as dbt (Intel) and valcun (Mirosoft) incur less overhead.

Tracing Lineage Beyond Relational Operators

Purdue University Purdue e-pubs Department of Computer Science Technical Reports Department of Computer Science 2007 Tracing Lineage Beyond Relational Operators Mingwu Zhang Xiangyu Zhang Purdue University,