Presentation is loading. Please wait.

Presentation is loading. Please wait.

Anindya Datta Debra VanderMeer Krithi Ramamritham Presented by –

Similar presentations


Presentation on theme: "Anindya Datta Debra VanderMeer Krithi Ramamritham Presented by –"— Presentation transcript:

1 Parallel Star Join + DataIndexes : Efficient Query Processing in Data Warehousing and OLAP
Anindya Datta Debra VanderMeer Krithi Ramamritham Presented by – Ashutosh Joshi

2 Motivation OLAP involves efficient retrieval of data from data warehouses for decision-support purposes Data Warehouses are extremely large and queries are highly computationally expensive DataIndex is a storage structure serving as both index and data Parallel Star Join (PSJ) is an efficient algorithm for performing star join in parallel

3 The Road Map A physical design principle for exploiting parallelism
Parallel Star Join algorithm Experiment results

4 The Star Schema Dimension Table Fact Table PART CUSTOMER PartKey 4
Name 55 Mfgr 25 Brand Type Size Others... 41 164 CustKey Name Address Nation Region Phone AcctBal MktSegment 10 Comment 269 SALES PartKey 4 SuppKey 4 CustKey Quantity ExtPrice Discount Tax RetFlag Status ShipDate CommitDate 2 ReceiptDate 2 ShipInstruct 25 ShipMode Comment 137 200,000 SUPPLIER 150,000 SuppKey 4 Name 25 Address 40 Nation 25 Region 25 Phone AcctBal 8 Comment 101 243 TIME TimeKey 2 Alpha 10 Year 4 Month 4 Week 4 Day 28 6,000,000 2,557 10,000

5 A Physical Design Principle
DataIndexes Serve as both index as well as data Based on vertical partitioning of tables Two types Projection Index (PI) Join Index (JI)

6 Projection Index Base Table PI PI PI CustKey Qty ExtPrice Discount CK1

7 Join Index Base Dimension Table Base Fact Table PI PI PI JI PI PI Name
Address CustKey CustKey Tax ExtPrice Discount N1 A1 CK1 CK1 T1 E1 D1 N2 A2 CK2 CK2 T2 E2 D2 N3 A3 CK3 CK3 T3 E3 D3 CK3 T4 E4 D4 PI PI PI JI PI PI Name Address CustKey RIDs RID1 RID2 RID3 Tax ExtPrice Discount N1 A1 CK1 T1 E1 D1 N2 A2 CK2 T2 E2 D2 N3 A3 CK3 T3 E3 D3 T4 E4 D4

8 The Principle Each foreign key column in the fact table is stored as Join Index (JI) Rest of the columns (for both dimension as well as fact table) are stored as Projection Index (PI)

9 Parallel Star Join Data placement strategy
Based on shared nothing architecture with N processors Assume a d dimensional data warehouse Partition N processors into d+1 groups Assign to each group j, dimension table Dj and Jj , the fact table join index Assign metric PIs to the group d+1

10 Processor Group Partitioning
Number of processors is governed by the size of dimension table Dj Size of jth processor group Size of metric group

11 Physical Data Placement
Horizontally partition JI’s across all processors Replicate PI’s on all processors Use round-robin strategy for partitioning JI’s

12 The Parallel Star Join Algorithm
A general k- dimensional star join query Select AdP, AmP from F, D1, … , Dk where Pjoin and Pselect The algorithm has three phases Local rowset generation Global rowset synthesis Output preparation

13 Local Rowset generation
Load PI fragment Pc P1 P2 PI fragment PI fragment PI fragment 25 5 7 15 1 Qty > 10 PI fragment Rowset fragment

14 Local Rowset Generation (contd)
Merge dimension rowset fragments Distribute dimension rowset Rowset fragment P1 P2 P3 P4 OR Rdim,i

15 Local Rowset Generation (contd)
Load JI fragment Merge partial fact rowsets 1 1 RIDs RID1 RID2 RID3 Rfact,i Rdim,i JIi

16 Global Rowset Synthesis
Merge local fact rowsets Distribute global rowset to groups participating in the output phase Rfact,1 G1 G2 Rfact,2 G3 G4 AND Rglobal

17 Output Preparation Distribute global rowset to individual processors
Load PI columns necessary for output Merge output CustKey CK1 CK2 CK3 CK4 1 Output CK1 CK2 RIDs RID1 RID2 RID3 PIi JIi Rglobal

18 Performance Comparison
The PSJ algorithm was compared with Bitmapped Join Index algorithm and the Pipelined Hash join algorithm Two performance metrics used Response time in block access (RTBA) Aggregate Data Transmission (ADT)

19 Scalability Experiments
The curves rise as the scale factor and number of processors increase PSJ cost is much lower than BJI and HASH costs At large memory sizes, PSJ approaches “near-perfect” scalability

20 Scalability Experiments(contd)
Transmission costs for PSJ and BJI are the same Both curves exhibit imperfect scalability HASH has substantially higher transmission costs than PSJ

21 Conclusion DataIndex is a physical design strategy which provides efficient partitioning of the schema Parallel Star Join algorithm provides a means to perform star join in parallel PSJ algorithm performs better than BJI and HASH algorithms in terms of I/O and transmission costs


Download ppt "Anindya Datta Debra VanderMeer Krithi Ramamritham Presented by –"

Similar presentations


Ads by Google