pavan poluri pavan poluri siddharth deokar varun sudhakar
play

Pavan Poluri Pavan Poluri Siddharth Deokar Varun Sudhakar Generic - PowerPoint PPT Presentation

Pavan Poluri Pavan Poluri Siddharth Deokar Varun Sudhakar Generic View of System Get the article I/O Processing Workload count and word & File size Distribution count management Create one Create one Redistribute files


  1. Pavan Poluri Pavan Poluri Siddharth Deokar Varun Sudhakar

  2. Generic View of System Get the article I/O Processing Workload count and word & File size Distribution count management Create one Create one Redistribute files Redistribute files Perform a Perform a final Suffix for calculating Binomial Array per P suffix arrays Reduction Retrieve the top LEGEND: Done!!! R interesting Siddharth ngrams Varun Pavan All

  3. FOSTER’S DESIGN IN OUR PROJECT � Partitioning: Domain Decomposition � Communication : Broadcasting, Point to Point Communication and Customized Communication � Agglomeration: Gathering of suffix arrays � Agglomeration: Gathering of suffix arrays � Mapping: Cyclic Mapping Strategy

  4. Data Structure � Customized suffix array1 to hold the following data � Position of ngram in the file � File index to identify the file � Term Frequency � Term Frequency � Document Frequency � Customized suffix array2 to hold the following data � Position of ngram in the file � File index to identify the file � Term Frequency � TF*IDF value

  5. Algorithm � I/O processing � Reading directory and storing file information � File size Management � Partitioning files Partitioning files � Communication � Workload Distribution � Interleaved Allocation

  6. Contd….. � Alpha Requirement: � Calculating the number of words and articles � Reduction � MPI_Reduce() � MPI_Reduce()

  7. Contd… � Suffix Array Calculation � Every word has a suffix array associated with it � Allocating memory to suffix array based on the alpha output output � Filling the details of suffix arrays of all words � Getting the position of the word in the file � Getting the file index of the file the word is in � Assigning term frequency � Assigning document frequency

  8. Contd… � Sorting the Suffix arrays � Based on Quick sort algorithm � Timing Complexity of quick sort : O(NlogN) (average case) � Memory Requirement : O(NlogN) � Memory Requirement : O(NlogN)

  9. Contd… � Finding Distinct terms in same article Cat(TF=1, DF=1) Cat(TF=1, DF=1) House(TF=6,DF=1) House(TF=1,DF=1) House(TF=2,DF=1) House(TF=3,DF=1)

  10. Contd… � Finding Distinct terms in different articles Cat(TF=1, DF=1) Cat(TF=1, DF=1) House(TF=3,DF=2) House(TF=1,DF=1) House(TF=2,DF=1)

  11. Contd… � Merging Suffix Arrays � Input: Two sorted suffix arrays � Reading ngrams from file � Output: One sorted suffix array � Output: One sorted suffix array

  12. MERGE EXAMPLE F C C F C C F C C F F I I I H H H H R R R Z Z Z S S S F C C F C C F C C F C C F F F F I I I I H H H H H H H H R R R R Z Z Z Z I I I I S S S S R R R S S Z

  13. Contd… � Communication Strategies � Reading and Writing files (Strategy 1 - deprecated) � Binomial Tree Reduction and Nomenclature � Use of MPI_Barrier � Use of MPI_Barrier � Single file corresponding to suffix array � Communicating Structures (Strategy 2) � Binomial Tree Reduction � Use of MPI_Pack, MPI_Unpack()

  14. Contd… � Binomial Tree Reduction 3 �� �� 1 1 �� �� �� �� 1 1 �� ��

  15. Contd… � Finding top R interesting terms � Calculation and Storage � New suffix array structure with IDFTF measure � Sorting � Sorting � Merging

  16. Analysis Alpha � Alpha 20 15 �� �� �� �� ���� ���� ��� ���� ���� ��� ������� ��� ���� ���� ���� Time in seconds ds �� ���� 10 Alpha �� ���� 5 �� ��� ��� ��� 0 ��� ��� 16 32 64 128 256 Number of Processors

  17. �� �� �� �� ���� ���� ��� ���� ���� ��� ��� ������� ���� ���� ���� ���� ���� ���� ���� � ��� ����� 10000 � ��� ����� 9000 8000 7000 seconds 6000 �� �� �� �� ���� ���� ��� ���� ���� ��� ��� ������� ���� ���� ���� ���� ���� ���� ���� 0.6mb 5000 � � ���� ���� ���� ���� Time in sec 80mb 80mb 4000 � ���� ���� 120mb 3000 2000 1000 0 �� �� �� �� ������� ���� ������� ������� ������� ���� ���� ���� ���� ���� ���� ���� 2 4 8 16 32 64 � ���� ����� Number of Processors �� ���� �����

  18. ngram = 1 1200 �� �� �� �� �������� �������� �������� �������� ���������� ���������� ���������� ���������� �� �� �� �� ����������� ����������� ����������� ����������� 1000 � ��� ��� ata in mb 800 � �� ���� 600 600 Data i � � �� �� ���� ���� ngram = 1 �� ��� ���� 400 �� ��� ����� 200 �� ���� ����� 0 2 4 8 16 32 64 Number of Processors

  19. Formula � Amdahl’s Law � Ψ <= 1/f+(1-f)/p � where f is the serial component and p is the number of processors processors � Ψ is the speedup � Gustafson’s Law � Ψ <= p+(1-p)s � Ψ is the scaled speed up � s is the serial component and p is the number of processors

  20. Contd… � Using our results for data of size 120 MB � Speed up = 7680/3156=2.4 � Considering the case where 4 processors as serial and 16 processors as parallel processors as parallel � Using the formula for Amdahl’s Law and substituting Ψ as 2.4 we get f = 0.22 � According to Gustafson’s Law using s = 0.22, Ψ (scaled speed up) = 3.34

  21. Contact Info � Project web page: giga word corpus � Email � Pavan Poluri: polur007@d.umn.edu � Siddharth Deokar: deoka001@d.umn.edu Siddharth Deokar: deoka001@d.umn.edu � Varun Sudhakar: sudha002@d.umn.edu

Recommend


More recommend