Presentation is loading. Please wait.

Presentation is loading. Please wait.

DB2 Memory management DB2 GSE 01/12/2011

Similar presentations


Presentation on theme: "DB2 Memory management DB2 GSE 01/12/2011"— Presentation transcript:

1 DB2 Memory management DB2 GSE 01/12/2011
14/11/2018 DB2 Memory management DB2 GSE 01/12/2011

2 Agenda Overview Statistics EDMPOOL DBDPOOL
14/11/2018 Agenda Overview Statistics EDMPOOL DBDPOOL Dynamic Statement Cache (DSC) Bufferpools TS / Mainframe Systems | 14/11/ |

3 Overview CF CF 14/11/2018 PLXA PLXB PLXB PLXC PLXC PLXT SYA1 SYA2 SYA3
MQP1 IMP1 DBP1 SYA2 CF DBPI MQP2 IMP2 DBP2 CF SYA3 MQP3 IMP3 DBP3 SYA4 MQP4 IMP4 DBP4 SYA5 SYA6 SYA7 MQP7 IMP7 DBP7 SYA8 MQP8 IMP8 DBP8 PLXB PLXB SYB2 CF MQC2 IMC2 DBC2 DBCF SYB1 MQC3 IMC3 DBC3 MQC1 IMC1 DBC1 SYB7 MQC7 IMC7 DBC7 SYB8 MQC8 IMC8 DBC8 PLXC PLXC SYB2 SYC2 CF MQC2 DBD2 IMC2 DBT2 DBC2 DBQ2 DBCF DBI1 SYC1 MQQ1 IMQ1 DBQ1 MQC3 MQT1 IMC3 IMT1 DBC3 DBT1 MQC1 MQD1 IMC1 IMD1 DBC1 DBD1 SYC7 MQQ7 IMQ7 DBQ7 MQT7 IMT7 DBT7 MQD7 IMD7 DBD7 PLXT CF SYT1 MQN1 IMN1 DBN1 DBZ1 SYT2 MQN2 IMN2 DBN2 DBN3 TS / Mainframe Systems | 14/11/ |

4 Statistics SMF 100 Statistics trace records
14/11/2018 Statistics SMF 100 Statistics trace records Format and Load in DB2 tabel with BMC Mainview Detail information (10 minute interval) Summary by hour System statistics Buffer statistics 3 months history Table field descriptions in User Guide Additional descriptions in SDSNIVPD member DSNWMSGS TS / Mainframe Systems | 14/11/ |

5 EDMPOOL EDM Pools – V9 14/11/2018 TS / Mainframe Systems
| 14/11/ |

6 14/11/2018 EDMPOOL TS / Mainframe Systems | 14/11/ |

7 14/11/2018 EDMPOOL TS / Mainframe Systems | 14/11/ |

8 14/11/2018 EDMPOOL TS / Mainframe Systems | 14/11/ |

9 14/11/2018 EDMPOOL TS / Mainframe Systems | 14/11/ |

10 14/11/2018 EDMPOOL TS / Mainframe Systems | 14/11/ |

11 14/11/2018 EDMPOOL TS / Mainframe Systems | 14/11/ |

12 14/11/2018 EDMPOOL TS / Mainframe Systems | 14/11/ |

13 Dynamic Statement Cache
14/11/2018 Dynamic Statement Cache TS / Mainframe Systems | 14/11/ |

14 Dynamic Statement Cache
14/11/2018 Dynamic Statement Cache TS / Mainframe Systems | 14/11/ |

15 14/11/2018 DBDPOOL TS / Mainframe Systems | 14/11/ |

16 14/11/2018 DBDPOOL TS / Mainframe Systems | 14/11/ |

17 EDMPOOL - DBDPOOL - DSC Triggers for critical problems:
14/11/2018 EDMPOOL - DBDPOOL - DSC Triggers for critical problems: EDMFAILS (QISEFAIL) : EDM pool full failures SKELPOOLFULL (QISEKFAL) : SKELETON POOL FULL failures STMTPOOLFULL (QISECFAL) : below the 2GB bar EDM pool statement cache full failures STMTPOOLABFULL (QISESFAL) : above the 2GB bar EDM pool statement cache full failures DBDPOOLFULL (QISEDFAL) : DBD pool full failures TS / Mainframe Systems | 14/11/ |

18 Bufferpools Main separation
14/11/2018 Bufferpools Main separation The first bufferpool by page size is for Catalog, Directory and DB2 Tools The last bufferpool by page size is for Sort and Work By Type AG BNP INFOCENTER Application Tablespaces are in bufferpools with an even number Application indexes are in bufferpools with an odd number By Size By access TS / Mainframe Systems | 14/11/ |

19 Bufferpools Initial setup 14/11/2018 TS / Mainframe Systems
| 14/11/ |

20 Bufferpools BP setup for specific reason 14/11/2018
TS / Mainframe Systems | 14/11/ |

21 Bufferpools Data separation BP setup 14/11/2018 TS / Mainframe Systems
| 14/11/ |

22 Bufferpools Problems How to decide on random – Sequential
14/11/2018 Bufferpools Problems How to decide on random – Sequential Growth of object Changes in behavior Difference in load Example on BP3 – BP5 BP3 – pages – Index Random BP5 – pages – Index Sequential > 30% TS / Mainframe Systems | 14/11/ |

23 Bufferpools Bufferpool Getpages 14/11/2018 TS / Mainframe Systems
| 14/11/ |

24 Bufferpools Bufferpool SYNC and ASYNC I/O BP3 –Index Random
14/11/2018 Bufferpools Bufferpool SYNC and ASYNC I/O BP3 –Index Random BP5 –Index Sequential TS / Mainframe Systems | 14/11/ |

25 Bufferpools Desired Setup
14/11/2018 Bufferpools Desired Setup Only possible to change for Bufferpools based on data separation In memory objects Small objects High getpage rate No I/0 after first read Hot spot objects Objects with a high re-read factor (i.e. > 20 getpages/read) Remaining objects Specific Bufferpools TS / Mainframe Systems | 14/11/ |

26 Bufferpools How to move a object to another Bufferpool
14/11/2018 Bufferpools How to move a object to another Bufferpool DB2 table containing list of objects New Bufferpool Current Bufferpool Rexx procedure Check objects that need to be changed Check if they are in use Issue STOP AT COMMIT ALTER object START OBJECT CHECK for STOPP Done during weekly maintenance window TS / Mainframe Systems | 14/11/ |

27 Bufferpools When moving objects verify impact Original Bufferpool
14/11/2018 Bufferpools When moving objects verify impact Original Bufferpool New Bufferpool Statistics for one index for one day: TS / Mainframe Systems | 14/11/ |

28 Bufferpools PAGEFIX Fix - Release is done when memory page is changed
14/11/2018 Bufferpools PAGEFIX Fix - Release is done when memory page is changed High I/O intensity Pages read = SYNC + ASYNC + GBP read Pages written = SYNC + DEFERRED + GBP Write Is activated or deactivated at bufferpool allocation Up to 8% reduction Disadvantage Page can never be reused by any other process Pages might be necessary for DUMP Processing TS / Mainframe Systems | 14/11/ |

29 Bufferpools Actions already done
14/11/2018 Bufferpools Actions already done Move some objects to best fitting Bufferpool In memory Bufferpool BP7/BP8 PGSTEAL FIFO Follow-up of I/O PAGEFIX BP16 with pages Merge two sets of Bufferpool BP11 + BP13 each pages to one of BP12 + BP14 each pages to one of TS / Mainframe Systems | 14/11/ |

30 Bufferpools Results of actions already done
14/11/2018 Bufferpools Results of actions already done Move objects to best fitting Bufferpool Higher Getpage rate, 2500 to over 4000 a second at peak hours for in memory bufferpool BP7 with same size In memory Bufferpool BP7/BP8 FIFO had no measurable CPU difference in applications due to divers workload PAGEFIX BP16 with pages no measurable CPU difference in applications due to divers workload Merge two sets of Bufferpool No impact seen in application Follow-up small improvements/degradation for some Transactions Improvements for batch with large workingset TS / Mainframe Systems | 14/11/ |

31 Bufferpools System page residency time
14/11/2018 Bufferpools System page residency time buffer pool size / total pages read per second Random page residency time Max Bufferpool size * (1-VPSEQT/100) / sync pages read per second More info can be found in the Redpaper DB2 9 for z/OS: Buffer Pool Monitoring and Tuning of Mike Bracey TS / Mainframe Systems | 14/11/ |

32 14/11/2018 Bufferpools Merge BP TS / Mainframe Systems | 14/11/ |

33 14/11/2018 Bufferpools Merge BP TS / Mainframe Systems | 14/11/ |

34 Bufferpools PAGEFIX BP16
14/11/2018 Bufferpools PAGEFIX BP16 High I/O during peak period is more then 200 pages/sec Not measurable due to workload TS / Mainframe Systems | 14/11/ |

35 14/11/2018 Bufferpools DATE HOUR BPGETPAGE BPREADS BPFPAGESRD BPPAGEINRD TS / Mainframe Systems | 14/11/ |

36 Questions ? TS / Mainframe Systems | 14/11/ |


Download ppt "DB2 Memory management DB2 GSE 01/12/2011"

Similar presentations


Ads by Google