Page 229 - DCAP402_DCAO204_DATABASE MANAGEMENT SYSTEM_MANAGING DATABASE
P. 229

Database Management Systems/Managing Database




                    Notes
                                                        Figure  13.3:  Sales-monthly  Segmentation
                                                                                      Disk 1

                                                                                     Jan PSU 1
                                                           January sales             Feb PSU 1
                                                             Segment
                                                                                    Mar PSU 1
                                                                                      Disk 2

                                                           February sales            Jan PSU 2
                                                             Segment                 Feb PSU 2
                                                                                    Mar PSU 2

                                                                                      Disk 3
                                                            March sales
                                                             Segment                 Jan PSU 3
                                                                                     Feb PSU 3
                                                                                    Mar PSU 3


                                   13.2.1 Horizontal Partitioning

                                   Horizontal partitioning a fact table speed up queries without indexing, by minimizing the set of
                                   data to be scanned. It is not feasible to partition a fact table into segments (virtual object like
                                   table) as we could not assure of each segment to be of same size as others, because the number
                                   of transaction within the business at given point will vary it may between year to year.
                                   Since  business transaction  volumes will  be higher  in seasons  like festivals and occasions  as
                                   compared to the rest of year.
                                   So we have to think all possible ways while partitioning the fact data at the same time we also
                                   have to consider the requirements for manageability of the data warehouse.

                                   Partitioning by Time into Equal Segments

                                   This is a standard form of partitioning. Here partition of fact table is done on a time period basis,
                                   where each time period represents a significance retention period within the business. Like if
                                   maximum queries are on month-to-date values, then the partition is done into monthly segments
                                   or if maximum query period is fortnight-to-date, consider partition into fortnightly segments
                                   but we should consider the number of tables should not exceed the limit.

                                   We can reuse table partitions by removing all the data from the partitions, by round robin by
                                   the warehouse manager. It is better to consider that number of partitions will store transactions
                                   over a busy period of the business.

                                   Partitioning by Time into Different-sized Segments

                                   This type of partitions is done where previous data is accessed infrequently. This will be as set
                                   of small partitions for current data, large partition for active data and even larger partition for
                                   inactive data.







          222                               LOVELY PROFESSIONAL UNIVERSITY
   224   225   226   227   228   229   230   231   232   233   234