Life of a Performance Specialist Symposium on Software Performance - PowerPoint PPT Presentation

“Into the wild…” Life of a Performance Specialist Symposium on Software Performance (SSP 2015) Munich, 5.-6.11.2015 Torsten Hellwig, Dipl.-Inf., Managing Partner @ QMETHODS Business & IT Consulting GmbH 1

You don‘t need to walk to Alaska… 2

Who am I – how did it start  first programming in 1986  Master in computer science from TU Dresden in 1998  software engineer and architect , Java, PHP, JS, …  in 2003 established testing / QA at an insurance company  „ moved to the other side “: leading Ops team – no system monitoring!  2010 first encounter with Applicaton Performance Management: Dynatrace  Performance Consultant ever since … 3

What does an application performance guy do ?  Troubleshoot application performance, yay !  Educating and evangelising APM  Understand customers maturity, organisational structure, processes  map requirements and expectations to a solution  help with tooling decisions: demonstrate tools, prove value  set up and establish tooling – develop customers maturity 4

A little theory on APM  it‘s a lifecycle thing !  get grip and control over your environment  performance requirements (NFRs to FRs)  develop performance aware or – driven  white-box testing  holistic monitoring  automation ! 5

What you need to look at…  organisational  classic siloed IT organization  DevOps  outsurcing  skill levels  maturity  driving blind – end user experience, monitoring  reactive, problem-driven  aware and effective  optimized process, continuous improvement  technical environment  tooling 6

Collaboration 7

Troubleshooting  pro-actively  new application to be rolled out – does it perform ?  QA or users constantly complain  Dev has no idea anymore  Ops has no idea but ‚ more hardware !‘  Business needs a solution  war room – the house is on fire  outtages  KPIs bad (conversion rate, revenue)  project deadline was … yesterday 8

Example 1: Troubleshooting  larger customer wants to introduce a new HR software  big bang rollout planned and communicated to >1000 employees  large peak in load expected on go-live  vendor and hosting provider did some improvised load tests with jMeter  more hosted hardware would drive up costs, still no guarantee 9

Approach  first: wire up APM to the test environment which is used by manual testers – do End-to-End analysis over a whole day  understand the app and user behaviour  see the infrastructure  check functional health  second: repeat a properly defined load test under APM diagnosis  proper definition: user behaviour patterns from above + capabilities of the load driver  estimated virtual users: 1 VU = 25 real users  system was able to sustain 300 real users … 10

Read the signs … 11

Example 2: business-driven, strategic-thinking „ We have no problems, we just want to stay ahead of the rest …“  strong growth – want to limit growing costs  high peak to normal traffic ratio – elastic cloud  business stakeholders want insight  better collaboration wanted – software vendor and hosting provider are external  application performance management is the key „ looks like a walk in the park …“ 12

What‘s the deal then ?  understand their real situation  full outsourcing ? uhm …  a Bank, with demanding stakeholders  really high traffic  propose a solution and get it up and running  integration+automation  hotline  incident management  big data analytics  collaboration process with both external parties  skill transfer 13

Example 3: Out there in the wild… The Mainframe 14

Active APM user …  it‘s about coverage  you can‘t start everything at the same time  complexity of project  cost  limited resources  customer with classic IT organization does APM in Dev, Test and Ops  started in the online business branch  technology coverage left out the backend: Mainframe  separate tooling – common architecture  again – growth (mobile frontend)  cost limiting: Mainframe pays per CPU usage ! 15

The task …  Assess requirements  propose a solution – extend APM into the Mainframe  win the Mainframers  evaluate tooling How ?  get it up and running and FIND SOME PROOF  the trick: correlation of distributed and mainframe on transaction level 16

Thanks for your patience! Questions ? Torsten Hellwig | Management Partner QMETHODS - Business & IT Consulting GmbH +49 (178) 838 0450 Stargarder Straße 10 10437 Berlin torsten.hellwig@qmethods.com Telefon : +49 (30) 22 19 86 51 Telefax : +49 (30) 22 19 86 53 E-Mail : info@qmethods.com Internet : http://www.qmethods.com