ibm電信業(yè)商業(yè)智能解決方案-預(yù)覽頁(yè)
【正文】 on, . scan, sort ?Also enables ?Parallel Index Create ?Parallel Backup and Restore ?Allows multiple processes to read or write data to/from the database ?Parallel LOAD ?Exploitation of multiple processors during load, particularly for parsing/converting/formatting data 節(jié)點(diǎn)內(nèi)部并行 ?Parallel Edition style (sharednothing) ?Data parallelism through hash partitioning ?Partitions can be... ?Physical on MPP or cluster ?Logical on SMP Run Time Agent Prefetchers Agent Prefetchers Agent Prefetchers node 0 node 1 node n SQL Query Query Optimizer Best Query Plan Threaded Code Compile Time 節(jié)點(diǎn)間并行 (數(shù)據(jù)庫(kù)分區(qū)間并行 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Single Database View Parallel Optimizer User Query Node (CPU) Node (CPU) Node (CPU) Node (CPU) ?Sharednothing software architecture supports ?Independent physical nodes ?Separate CPU, memory, and disk ?Including SMP nodes OR ?Multiple logical database partitions on single large SMP Server ?Interpartition munication is cross memory, not cross work ?Data is partitioned across nodes automatically by hashing ?Everything operates in parallel ?Select ?Insert ?Update ?Delete ?Backup/restore ?Load ?Create index ?Re 充分利用分區(qū)數(shù)據(jù)庫(kù)的能力 Social Insurance Number Name Location 123456789 JoeBoston Toronto Partition Key value Hashed to: 8 Vector Position 0 1 2 3 4 5 6 7 8 9 10 11 12 ... Node 1 2 3 1 2 3 1 2 3 1 2 3 1 ... DB2 DB2 DB2 ?Partition Map ?Determines 39。 數(shù)據(jù)挖掘的典型例子 基于歷史數(shù)據(jù)預(yù)測(cè)行為 發(fā)現(xiàn)未知分群、規(guī)則和模式 常用數(shù)據(jù)挖掘算法分為三類 Data Mining Algorithms No Prediction Predict One Thing Time Series Matching Predict Everything Associations Sequential Patterns Decision Tree RBF Classification Value Prediction Neural Neural Clustering Demographic Neural 常用數(shù)據(jù)挖掘算法 ? Clustering (Segmentation) no dependent variable ? Demographic Segmentation ? Neural Segmentation (Kohonen Map) ? Example: Identify mon characteristics in a customer data base. ? Predictive/Classification Modeling dependent variable ? Nonlinear regression ? Decision trees ? Neural works ? Radialbasis functions ? Example: Predict IBM39。 關(guān)聯(lián)分析 (link analysis) IM4D體系結(jié)構(gòu) Client Server Standard Extraction or Replication Tool Data Analyst Results Graphical User Interface Application Program Interface Data Mining Techniques Oracle Sybase Data Processing Functions Flat Files DB2 DB2 Files 數(shù)據(jù)挖掘?yàn)闃I(yè)務(wù)方案提供了一條途徑 Clearly Define the Business Problem Understand Problem Prepare Relevant Data data mining Analyze Results Present Results Implement Solution and Measure Success IBM DM on the Web ? ? ? 謝謝
教學(xué)課件相關(guān)推薦
鄂ICP備17016276號(hào)-1