Presentation is loading. Please wait.

Presentation is loading. Please wait.

High KBC Jan Tielemans. Agenda  Back in time  How the HA KBC looked like in 2006  Pro - Cons  Discuss the different steps(projects)

Similar presentations


Presentation on theme: "High KBC Jan Tielemans. Agenda  Back in time  How the HA KBC looked like in 2006  Pro - Cons  Discuss the different steps(projects)"— Presentation transcript:

1 High KBC Jan Tielemans

2 Agenda  Back in time  How the HA KBC looked like in 2006  Pro - Cons  Discuss the different steps(projects) to move to a HA environment anno today  Future directions/plans

3 Why a ‘High Availability’ environment ?  Back in time (1996/7….)  Objective HA environment :  Keep the 24*7 applications running during ‘technical maintenance’  Be able to ‘re-direct’ 24*7 applications to ‘one’ system in case of disaster (no DRP)  Stable performance  How it was implemented technically :  ‘Headoffice & Branches’ workload runs on one lpar(M1)  Retail workload (24*7) runs on the other lpar(M2) Both ‘workloads’ share the same DB2 data  Developed a ‘Switch procedure’ to redirect the retail workload to another machine

4 anno 2006 (DSN A) (DNS B) (DNS XXXX) SIPC SDPC SMPC IC1IC2 DNS A IPC1 DPC1 MPC1 IC5IC6 DNS B IPC2 DPC2 MPC2 IC3IC4 DNS XXXX M1M2 WASWAS TCP / IP Retail (DNS B) Retail (IPC1  DNS B) Headoffice & Branches (DNS A) Headoffice & Branches (DNS A) Headoffice & Branches (DNS A) Headoffice & Branches (SIPC  DNS A) Retail (IPC2  DNS XXXX) AGF frame work (KCF => connection cleanup after 20 min) 3270‘ Headoffice & Branches’

5  Technical environment :  DB2  Datasharing, 2 active members – 1 sleeping member  IMS  No Shared Queues  Subsystems/Regions for Retail not the same as ‘Headoffice & Branches’  MQ  No Shared Queues  Subsystems for Retail not the same as ‘Headoffice & Branches’  For some queues we use the MQ Clustering technique

6  Switch Procedure  Very complex and error prone  Often resulted in unavailability 24*7 applications  Lot of interaction with WAS servers  Redirecting workload is done on each WAS server  Controlled via a mainframe application  Knowledge and maintenance by one person  Mainframe application designed in NetView ($)  Static Workload distribution  Under utilization of CPU capacity at certain time periods  Availability was high, +99%

7  Things changed after initial setup in 1996  More CPU/Memory available per Machine  New (critical) applications for the ‘Headoffice & Branches’ and Retail workload  Other workload implemented  More and more a mix of concurrent Online & Batch workload  Regulations of BASEL – CBFA - …..  …….

8  Why change ?  Complex – error prone Switch procedure  Better utilize CPU capacity  Better use and exploit sysplex technology  Prepare for scalability (Cloned subsystems)  A 3 step approach  Make Open Systems independent from Mainframe  Dynamic Transaction routing to the Mainframe  Workload Balancing Retail  Workload Balancing ‘Headoffice & Branches’

9 Implementation DTM Switch Procedure (06/2007) TCP/IP Sysplex Distributor (DNS B  M2) SIPC SDPC SMPC IC1+xitIC2+xit DNS A ‘Headoffice & Branches’ (SIPC  DNS A) Retail (GIPC  DNS B) AGF frame work (KCF => time to leave 5 Minutes) BankSys (GIPC  DNS B) IPC1 DPC1 MPC1 IC5+xitIC6+xit DNS B IPC2 DPC2 MPC2 IC3+xitIC4+xit DNS B KBC Phone (GIPC  DNS B) M1M2 TCP/IP Sysplex Distributor (DNS B  M1) 3270‘ Headoffice & Branches’

10 Implementation DTM Switch Procedure (06/2007)  Time managed connections to IMS  DTM Switch Procedure :  Written in automation  Retail Workload redirected in less then 10 minutes  No interface/communication with WAS Servers  Switch is a Mainframe ONLY operation  Technical maintenance for M2 starts now on 13:00h vs 22:00h  Cloned IMS and MQ subsystems  Exploit TCP/IP sysplex Distributor  Implemented IMS exit to resolve :  Map the IMS group name to an active IMS subsystem  Control if IMS subsystem is active, if not redirect to other

11 To be solved for the next step(s) : Workload Balancing  Application changes :  Eliminate system affinity in applications logic  Get IMSID – If substr(IMSID,3,1) eq ‘P’ then …..  Identified application which could suffer from DB2 Datasharing  Identified serial transaction  How to serialize trx’s in a parallel environment ?  Communication – presentations for different departments  Mind Change

12 LET OP! Wij draaien in HIGH AVAILABILITY Transacties kunnen op beide “online” productiesystemen in uitvoering gaan!

13 HA for Retail (11/2008) KBC Phone (DNS B  GIPC) Sysplex Distributor WLM MANAGED (DNS B  M1, M2) Headoffice & Branches (SIPC  DNS A) ONL (elb & ipa) / AUT / KID / IIP (DNS B  GIPC) AGF frame work (KCF => time to leave 5 Minutes) BankSys (DNS B==> GIPC) SIPC SDPC SMPC IC1+xitIC2+xit DNS A IC3+xitIC4+xit DNS B IPC1 DPC1 MPC1 IC3+xitIC4+xit DNS B M1 M2 WLMWLM WLMWLM 3270‘ Headoffice & Branches

14 HA for Retail (12/2008) KBC Phone (DNS B  GIPC) Sysplex Distributor WEIGHTEDACTIVE (DNS B  M1, M2) ‘Headoffice & Branches’ (SIPC  DNS A) ONL (elb & ipa) / AUT / KID / IIP (DNS B  GIPC) AGF frame work (KCF => time to leave 5 Minutes) BankSys (DNS B==> GIPC) SIPC SDPC SMPC IC1+xitIC2+xit DNS A IC3+xitIC4+xit DNS B IPC1 DPC1 MPC1 IC3+xitIC4+xit DNS B M1 M2 3270‘ Headoffice & Branches’

15  IMS Exit :  Build in logic to not distribute ‘some’ serial transactions from ‘Headoffice & Branches’  Run these serial transactions only on ONE ims subsystem(SIPC)  Switch from SERVERWLM to WEIGHTEDACTIVE distribution method  To much important work defined in WLM  Heterogeneous workload on M1 – M2  What information sends WLM to the Sysplex Distributor ?  Weightedactive weights can easily be modified with ‘simple’ commands.

16 HA for Retail (12/2008) KBC Phone (DNS B  GIPC) Sysplex Distributor WEIGHTEDACTIVE (DNS B  M1, M2) ‘Headoffice & Branches’ (SIPC  DNS A) ONL (elb & ipa) / AUT / KID / IIP (DNS B  GIPC) AGF frame work (KCF => time to leave 5 Minutes) BankSys (DNS B==> GIPC) SIPC SDPC SMPC IC1+xitIC2+xit DNS A IC3+xitIC4+xit DNS B 10 M1 IPC1 DPC1 MPC1 IC3+xitIC4+xit DNS B M2 Sysplex Distributor WEIGHTEDACTIVE (DNS B  M1) ‘ Headoffice & Branches’

17 HA for Retail (12/2008)  Benefits  One environment (IPC2-DPC2-MPC2) less to manage/maintain  ‘Retail’ Workload balancing is dynamic adjustable…..  Better utilization of resources (Lpars, cpu, memory..)  Pre z10 40% to M1, 60% to M2  z10 10% to M1, 90% to M2  (Cons)  WLM management of the workload not possible due to the heterogeneous workload on the system(s)

18 HA for ‘Headoffice & Branches’ (7/2009) KBC Phone (DNS B  GIPC) Sysplex Distributor WEIGHTEDACTIVE (DSNA  M1,M2) (DNS B  M1, M2) ‘Headoffice & Branches’ (SIPC  DNS A) ONL (elb & ipa) / AUT / KID / IIP (DNS B  GIPC) AGF frame work (KCF => time to leave 5 Minutes) BankSys (DNS B==> GIPC) SIPC SDPC SMPC IC1+xitIC2+xit DNS A IC3+xitIC4+xit DNS B 10 M1 IPC1 DPC1 MPC1 IC1+xitIC2+xit DNS A M2 Sysplex Distributor WEIGHTEDACTIVE (DSNA  M1) (DNS B  M1) IC3+xitIC4+xit DNS B MSC - IP connection IP link ‘ Headoffice & Branches’

19  IMS MSC (Multi Systems Coupling)  Defined SIPC as base system for serial transactions  Define serial transactions as Local - Remote  Still have (and will have) 21 serial transactions  All serial transactions are now managed by MSC, removed logic in the IMS Exit

20 HA in Performance figures SSPC LP 18 (2094 –z9) Memory 40gb IBS I IBS II FF SSQC LP 10 (2097 z10) Memory 40gb Io <2ms 25K io/sec <20µsec Housekeeping Personeelsnet Service center Asset center Webseal …… ‘Headoffice Branches’ Retail ‘Headoffice Branches’ Ldap Sysplex Distributor WEIGHTEDACTIVE ‘Headoffice & Branches’ ms (2/3 DB2 -1/3 PGM) milj trx /day – 200 trx/sec 9:00 – 17:00 ONL (elb & ipa) / AUT / KID / IIP ms ((2/3 DB2 -1/3 PGM) milj trx /day – 250 trx/sec 00:00 – 24:00 AGF frame work (KCF) 90 Ldap Batch Mirrored (GDPS managed)

21 Future Plans  Standardization on Subsystems names  Done  Implement “subsystem” failure management  Design for a SERVERWLM distribution  Design for CBU implementation vs Priorities  Retail & ‘Headoffice & Branches’ can not run on peek times (normal business hours) on one Lpar (system)

22 QUESTIONS ?


Download ppt "High KBC Jan Tielemans. Agenda  Back in time  How the HA KBC looked like in 2006  Pro - Cons  Discuss the different steps(projects)"

Similar presentations


Ads by Google