As
the enterprise projects for Hadoop have started picking up steam, we need to
come up with an effective delivery methodology for Hadoop and machine learning projects.
It is evident to any industry player that the traditional waterfall methodology
won’t work for such projects. Also, a typical hackathon development attitude
from a startup culture may not fit in the CMMi criteria of multi million enterprise
bids.
Hadoopsphere
tries to present a draft delivery methodology for Hadoop development projects.
The draft still needs to go through a few reviews and comments are welcome to
make this a robust methodology.
Some
notes on this delivery methodology:
1) Agile delivery techniques are proposed – which may include Scrum or
Kanban based delivery techniques. While Scrum may rely on time bound
iterations/sprints, Kanban relies on exploratory task/event based assignment
and may not be time bound. Kanban is well suited for initial product or
application development while Scrum may be recommended for subsequent releases.
2) The methodology has been tried to accustom to both product and services
project development – further inputs are welcome.1
3) The methodology has been tailored with Big Data Analytics Process
proposed in Gartner report. 2
4) The artifacts which play a key role in Hadoop projects include but are
not limited to Hardware & Software Plan along with Data Management Plan.
The other proposed artifacts are mentioned in figure above.
[1] Key
inputs from Sachin
Ghai.
[2] Use Kanban to Manage Big DataDevelopment, Nathan Wilson, Gartner, August 2012
I assume the arrows at top indicate the repetitive agile iterations
ReplyDelete