landsend teched2016 (1)
TRANSCRIPT
Customer Case StoryLands’ End
Dodgeville, WI
Who we areLands’ End – retailer with sailing roots – 1.4BFY15• 4 UAC environments TEC, TDQ, PRD, CC• Production Environment
– 24 agents, windows, unix, z/OS– >600 workflows, 16,350 tasks– Over 600,000 task runs each month
How we became UAC customers• The mainframe is going away!
The mainframe is going away!• We used CA’s ESP for more than 20 years
– Mainframe engine – IPLs affected all workload• Among the first to move off mainframe• LE Core Team: Ann as PM, Brian, Mary
Jo, Lisa
The OdysseyGetting to our final go-live with Opswise was a multi-year effort 2013-2015• Demos and Request for Proposal (5
vendors/products)• POC process
– Recreate existing test ESP schedules as Opswise workflows
• Lost funding• Regained funding• Lost time; deadline looms
Conversion• Initial conversion team arrival January 2014• Decisions galore throughout the conversion
process• OOPS. Due to internal lack of communication:
Re-converted a sub-set of workload• It takes a Stonebranch village
– Thanks to Tech Support (Laurie!), Bob ***, Jackie ***, Tracy ***, Colin ***, and especially Denis ***
Activity During Conversion
• Training for developers • 9 weeks (2x week) training delivery for
System Operators, also recorded & on Sharepoint
More Activity During ConversionLE custom dashboard
Verification of Conversion
Sub-set of converted code verified line-by-line
• Core team: 10% of converted workload • Developers: at their discretion
• Reinforced training in tool• Raised confidence
Cut-Over
• Two cut-overs– First was small-subset of simple and mostly
independent workflows, Sept 2014• Interface between scheduler environments
– Second & final cutover Feb 2015• Around the clock onsite support for a week• Started Feb 3; contractually required to shut down
old tool Feb 14
Lessons about a conversion:• Involve some developers in your POC – no matter how
busy they are• Missed accounting for “NOINHERIT” & “NODE” in ESP• Find a way to keep tabs on changes in old system during
conversion• (Opinion) Put jobdoc info in a central repository outside
of UAC
Life after conversion• Keep development/QA workload completely
separate from production workload development• Don’t create anything new* directly in the
production system• Follow references when promoting• Don’t allow predecessors to task monitors or
(time-of-day) timer tasks
(Some of) What we did right:Standard Dashboard Series of training sessions
Resource for each agent Recorded training sessions
Resources for online systems Concatenated JCL libraries
Standards for workflows• Names end in LDATE ${TRDATE}• ACTION email using variables for
certain status
Standard naming conventions• Task monitors end in TM• Resources end in abbrv of type• Resources representing an agent
named for agent• File monitors end in FM• Triggers end in TR
Project manager arranged for food 24x7 during go-live week (big hit!)
The BIG Wins• First night’s processing only 0.03% failures
attributed to Opswise • Very few problems in converted code since• Reduced CPU usage on z/OS by ~ 19hrs of
CPU time/month• Of course $$ savings
Some Additional Advantages
Template courtesy of: www.presentationmagazine.com
Can change commands or parameters in task instance waiting to run
When z/OS LPAR is IPL-ed, distributed workload keeps running
Fail-over to passive controller within minutes
Automatic restarts are easier to code and easier to identify
More flexibility in how permissions are defined for users
Not dependent on messages to z/OS master console
Easier to see # of task monitors running (rough estimate of remaining workload)
Due to custom dashboard the following is easy:• Identify Tasks stuck in submitted state• Notice Late starts / Late ends very visible• Identify Tasks waiting for resources• See what is using a resource and what is defined to use it• See what is currently processing
ContactsLisa.***@landsend.com
Brian.***@landsend.com