CT2 Development Process v2

From UG

Revision as of 17:15, 26 July 2009 by Alex (Talk | contribs)
Jump to: navigation, search


! This is UNDER CONSTRUCTION (NOT COMPLETED YET). Old version of this article here: CT2 Development Process


Contents

Intro

This article defines what methodology, roles, workflow, etc we use in CT2 Web Software Development.

CT2 Supercomponents

CT2 System is large and can be divided into the following supercomponents (supertasks):

  • Acc
  • Client App
  • OpsReps
  • Ops
  • OpsPdfs
  • Admin
  • Misc

Each supercomponent has its own Development team, project plan, etc.

Some coordination between teams is needed because supercomponents are interrelated to a certain degree.

Business Team (Client/Sponsor)

Acceptance Board (The Board)(Committee)(Demo Meeting)

  • group of people who makes final decision about acceptance, tweaks, course of the project.

Chairman of Acceptance Board

  • person who makes final decision in case there is a conflict

Business Analyst

  • logistics/transportation expert/manager
  • invents fetures, components, formulates business needs and requirements
  • communicates/transfers these requirements to Systems Analyst

Systems Analyst

  • IT expert, core knowledge of logistics/transportation required
  • interview Business Analyst
  • creates wiki spec doc that describes detailed business requirements and specifications
  • this wiki spec is all developer and QA need to fully implement feature/component
  • transfers specs to development team

Project manager

  • time/scope management expert
  • accountable for delivering component on time and on budget
  • managing/coordinating business phases of CT2 Software Development Worflow (see below)

Software Development (Vendor) team

Development Manager (Lead) - leader of the development team (Slava)

  • coordinates development effort
  • manages resources
  • manages low level of proj plan: detailed development schedule that includes phases of development, bugfix, QA, graphics (for every sub-component/dev task)
  • gets daily updates from developer, QA, graphic designer regarding deadlines
  • if there is a chance for ETA to be postponed then he immediately notifies Project Manager and they both decide what to do: move tasks around, add resources or push ETA forward or do something else

QA engineer (or manager) (Andrei)

  • makes sure that specs are complete
  • creates test plans
  • runs test plans
  • forwards bugs to developer
  • provides ETA for his work to Development Lead
  • must immediately notify Development Lead if his ETA can slip

Developer (Sasha)

  • makes sure specs are complete
  • implements new sub-components, changes, tweaks, bug fixes
  • deploy, forwards his work to QA
  • provides ETA for his work to Development Lead
  • must immediately notify Development Lead if his ETA can slip

Graphic designer (Slava)

  • responsible for creating overall Look and Feel concept
  • creates Style Guide
  • creates all graphics for application
  • provides ETA for his work to Development Lead
  • must immediately notify Development Lead if his ETA can slip

Project Workflow and Mantis

Note: All below is specific to Mantis soft we use.

It is explained using Client Application CT2 Super Component as an example.

Decomposing Super Component into Components and Tasks

At some point it will be clear how to split specific super component (and its development) into a number of large subcomponents.

At this point Sys Analyst creates "mantis parent records" for each super component under Mantis project=[SuperCompName]. For example under Mantis Project=>>(Client) for Client Application you will see:

595	[Client.*Misc] =================== Client			
924	[Client.CTdetails]			
919	[Client.HomePage]			
921	[Client.ListAll]			
1015	[Client.LogInPanel]			
685	[Client.LookAndFeel]			
939	[Client.MAWBtracking]			
922	[Client.MyProfile]			
591	[Client.Reports]			
920	[Client.WatchList]			
925	[Client.WhereIs]

Creating specific tasks for each component

Sys Analyst creates tasks for each component and links as child to parent record so it is easy to see any time what tasks are there for each component and in what state.

Main task would be "Implement ver x.xx". But some of these tasks are bugs, tweaks, etc. For example under parent component 0001015: [Client.LogInPanel] we see:

parent of 	0001016	started	andrei 	4.QAbasic 	[Client.LogInPanel] Implement ver1.0  
parent of 	0001017	new	slava 	00.Graphics 	[Client.LogInPanel] Create Look and Feel for ver# 1.0  
parent of 	0001024	new	dima 	3b.Dev 	        [Client.LogInPanel] BUG: Login from another application hyperlink (MS Exel - hyperlink - Client)

Tasks are classified into:

  • changes (new features, new versions of existing features/components, tweaks/changes)
  • bugs

Mantis Projects

Mantis Project is a "top filter" (see top right corner). We use it to separate different phases/states of development: specing (sys. analysis), development, QA, demo, etc.

Each task can be only in one state at a time.

Typical Workflow for Changes

  • Sys Analyst and Biz Analyst discuss change/new ver of component
  • At some point task is created, assigned to Sys An and moved to "Specs"
  • Once specs are complited (added to wiki) they must be passed to Development Team - move to "Dev" and assign to one of the members on Dev Team
option A) pass it to QA who will analyze for completeness
option B) pass it to both developer and QA
option C) pass it Dev Team Lead who will decide how to organize spec transfer
  • Once specs are accepted by dev team, developer and QA both give ETA for Scrum and Proj High Plan gets updated with Demo date for component
option A) QA creates Test Plan before development starts (this would increase total time)
option B) QA creates Test Plan during development (best? two subtasks are done in parallel)
option C) QA creates Test Plan after development starts (this would increase total time)
  • After development is complete developer must: commit, make sure all deployed, move task into "QA" and assigns to QA
  • QA tests and if any bugs found passes back the task to developer (move to "Dev")
  • Bug free task will be moved to "Demo" and assigned to Sys Analyst
  • Sys An will run top use cases to verify before showing it to the Board on Demo meeting

Typical Workflow for Bugs

TBD

Communication

If there is a question regarding task then:

  • post qustion into Mantis notes
  • set task to feedback
  • assign to person you are asking
  • if urgent contact that person in Skype

If there is a general question - use email. CC as required.

In addition there are:

  • Weekly Status Meetings
  • Communication Hours

Sys Analyst and Dev Team Lead must be able to reach each other during extended timeframe.

Note: To be always be available in Skype - chance for constant iteruption. 
It is not clear how much of interruption to allow. 
In some cases it could be very useful, in some cases very contr productive 
(especially during tasks that require high concentration)

In case of urgency (app is down, etc):

  • we must define what can happen and how to reach someone we need to reach to solve urgency

Weekly Status Meetings

  • Why: These meetings are group chats in Skype with Agenda to update each other on progress
  • Who: Dev Team + Sys Analyst
Client App Team (summer 2009 schedule): Tuesd, Thur: 10-10:45am NY time
Reports and Accounting Team (summer 2009 schedule): Tuesd, Thur: 9:15-10am NY time

Communication Hours

  • Who: Dev Team + Sys An
  • What: All must be available in Skype
  • Why: In case somebody needs to discuss not through Mantis, but "live" through chat
  • When:
For All Super Components: Mon, Tue, Wed, Thur, Fri: 9-11am NY time
  • If somebody can not attend he must notify Dev Team Lead and give alternative time frame that he is available same day

Vacations and Days Off

  • Must notify Dev Team Lead as much in advance as possible
  • Dev Team Lead must post this info into Dev Calendar

Scrum and Sprints

Scrum is an approach we use to scheduling components/tasks into development.

Planning is based on weekly or bi weekly or any regular time intervals that is called Sprint.

Before the start of the Sprint team is looking at current tasks backlog (new components, bugs, tweaks, changes, etc) and select a number of tasks for next Sprint based on prioity, size of tasks, resources available, etc.

Note that conventional approach is to estimate each task and build a sequence (plan). One of the advantages of Scrum I guess is that it has always same distance between milestones.

So under Scrum if there is an ETA slip for given task then instead of re-scheduling milestone (demo date) we move delayed component to next demo and ideally work on and show something else instead for current milestone (demo).

If Release Date is fixed then all Sprints must be planned in advance so that they cover entire backlog.

Project Plan Document with Mantis and Scrum

  • Developer must set emh(estimated man hours) hor each task
  • QA must indicate his QAemh
  • Based on: Dev emh, QA emh, ds (priority) and other constraints (like other tasks, days off, etc) Dev Team Lead together with QA and Developer must create Task List for next Sprint.
  • If Sprints are say weekly and start on Monday then this Task List consists of all tasks that will be ready for Demo on Friday
  • For all these tasks Dev Team Lead must set due(ETA) field
  • This Task List for next Sprint must be approved by all and presented to Sys Analyst and to the last Status Update Meeting of previous week.
  • These ETA will be included into high level Proj Plan and be presented to the Board.
  • Using Mantis sorting and filtering features these tasks could be easily extracted and migrated to xls or other format.
  • I suggest incorporating into planning 1 day / week of developer time for tweaks and bugs --Alex 06:30, 23 July 2009 (UTC)

Managing Time and Scope

It is very hard but if we want to fix Release Date then we all must:

1) be VERY serious about ETA

2) use techniques to avoid and solve this problem. See below.

Timeboxing

Scoping

Personal tools