4.3 EDM Plan Overview

well, everyone.

My name is Sean Reynolds and welcome back to our

discussions surrounding big data at Comcast today.

We're going to go over our Enterprise data management plan

and the ML voice reporting project.

Let's get started on the next slide with an overview

of today's discussion as we look at our past big

data initiatives and begin to gaze onwards to R E.

D M Future State, the machine learning voice team at

Comcast has developed a plan to recap.

What we've discussed in prior meetings are EDM Team will

help the ML voice group provide downstream users the means

to query more enlightening information.

In order to train our AI model, The following slides

will address our DW architecture enterprise scale, some of the

staffing requirements for the overall project as well as a

cost analysis and two different requirements necessary to attain a

more robust solution.

Moving forward.

Let's start by looking at the Kimble Lifecycle before we

discuss this further.

Let's talk about the six phases.

These six phases will help us two address some of

the pre work prior to the Kimble lifecycle.

This includes addressing urgent current state issues, defining the first

version of the ADM Future state conducting a stakeholder summit

for Mitt engagement.

Feedback. Adjusting the ADM future state is necessary.

Building the faced roadmap from the current state to the

future state and then executing the road map.

The Kimball Life cycle is going to take place during

the roadmap execution phase.

The figure here acts as an overview describing the entirety

of the project and how the project flows between the

different requirements.

Firstly being the business requirements definition where the project is

defined and then moving into three separate tracks shown in

the middle as technical architecture design, dimensional modeling, FBI application

These three tracks make up the entirety of the D

W architecture.

Our technical architecture design includes some of the processes we

will use to collect data.

How that data is stored and how that data is

sourced as mentioned previously are suggested.

Architecture follows the mic to incremental design process which stands

for method for an integrated knowledge environment.

We will continue to define data sources and target tables

and our dimensional design worksheet in the dimensional modeling track.

And lastly RB application will include the Amazon redshift e

T. O Toole connected to our s three Data Lakes.

As the project progresses, let's talk about our staffing play.

We have different roles here shown in the table below,

including business sponsors, the APP, designers, data architects and database

administrators, E T l architects and metadata coordinators.

We can't anticipate having about four business sponsors from different

cross functional groups.

These will likely be VPs from the machine Learning team

and three other downstream teams that will help to advocate

the project.

We'll also look for the APP designers that are certified

in Amazon Web services and that have previous experience with

Red Shift.

Then we'll look for data architects that will eventually phase

into Deba roles.

These individuals will help us to normalize our models.

Then we will look for retail designers that will help

us with all of the detail processes.

Although the first portion of our project is likely to

be around one year, we will keep the et al

architects on most likely for the entirety of the E.

D. M project, which is three plus years, and lastly

we'll have metadata coordinators that are going to initially tag

and manipulate the user data to help us with the

GTL process The main focus for our cost analysis is

going to be with regards to staffing, seeing that most

of the source systems and tools are already in place.

The main focus is determining our project's success in determining

our project.

Success is the stakeholders satisfaction and the financial improvements made

to our storage and maintenance of the ADM Future state

Using the staffing matrix in the previous slide, we estimate

the staffing and consulting fees in the first year to

be around $1.5 million.

Next, our project is anticipated to last around 3 to

4 years, with machine learning reporting being complete in the

first year.

Staffing, or ADM initiative, with the proper personnel over the

course of several years, takes a partnership between human resources,

C level executives and many cross functional groups that are

temporarily assigned to the project.

One of the main objectives of the D M project

is to lower development costs for machine learning and AI

models, which is going to be achieved through our increased

speed and data reliability in our production environment.

The production team often waits to iterate the project further

until downstream teams can help them to run analysis on

the current state of user input medium project costs associated

with migrating.

Our data will be offset by the savings in minimizing

the diva is required for server maintenance and ad hoc

reporting. Some of the other cost factors in how the

team plans to use Amazon Redshift is the cost of

using $4 per terabyte of scanned data, which equates to

approximately $4000 in daily analytical processing.

And moving on to the conclusion, which is the overall

benefit of the idiom project is going to be to

assist the larger enterprise and providing cross functional reporting capabilities

to downstream users.

The machine learning team is expected to save approximately $500,000

annually due to the inefficiencies of moving, storage and online

processing to an entirely cloud based environment.

We appreciate your time and thank you for giving us

the opportunity to present today

4.3 EDM Plan Overview

Chapters

Audio TranscriptAuto-generated