56
Praveen Srivastava Snr Engineering Manager Java Oracle Corps, Bangalore [email protected] https://www.linkedin.com/in/praveen-srivastava-388a9012

Six Sigma Green Belt (Product Track)

Embed Size (px)

Citation preview

Page 1: Six Sigma Green Belt (Product Track)

Praveen Srivastava Snr Engineering Manager – Java Oracle Corps, Bangalore [email protected] https://www.linkedin.com/in/praveen-srivastava-388a9012

Page 2: Six Sigma Green Belt (Product Track)

GB Candidate : Praveen Srivastava

Page 3: Six Sigma Green Belt (Product Track)

Srl. Version Date Comments

1 Draft Aug 1, 2009 Project definition, SDFSS Tools used, schedule

2 1.2 Sep 24, 2009 Updated with CPM Analysis slides, P chart

3 1.3 Oct 30, 2009 Update Plan, Initial transfer function

4 1.4 Jan 08, 2010 Changed few factors for provisioning dependency

5 1.5 Mar 03, 2010 Example minitab charts

6 1.6 Mar 27, 2010 Updaintg Architectural analysis

7 1.7 Apr 04,2010 Rerun of DOE and sheet update

8 1.8 Apr 09,2010 Updated for FTR - IHLR-K10250

9 1.9 Apr 14, 2010 Updates in progress.

10 1.10 Apr 23,2010 Bulletin issued to the iDEN markets

11 1.11 June 28,2010 Value statement received from Technical Lead

12 1.12 July 22,2010 Updated with MBB review comments and success story

Page 4: Six Sigma Green Belt (Product Track)

• Success Story

• Product (s) Overview

• SDFSS Project Overview

• SDFSS Deliverables and Tool Usage

• Project Definition (Charter)

• Requirements Phase

• Architecture Phase

• Value Statement

• Lessons Learned

Page 5: Six Sigma Green Belt (Product Track)

SSPD Program Success Story

Networks (iDEN): HLR Provisioning Project

Green Belt: Praveen Srivastava, Bangalore Risk: High Provisioning Rate Impacting Service ( Lose Registration Request, Impacted HLR Functionality, Failed Provisioning Request, Increased High Severity NPRs)

Success Statement: First application of DFSS tools and methods for iDEN programs Increased maintenance window provisioning rate from 7 msg per second to 15 msg per

second for customers. Reduced RPN from 45 to 24 using DFMEA activities for identified critical risks Generated new feature ideas to reduce MOL effort with architectural findings. CPM and minitab-DOE helped to identify significant factors and their interactions affecting

HLR provisioning rate and establish a mathematical model which is used to predict HLR provisioning capacity. This allows customers to increase provisioning system utilization during maintenance window with 100 % improvement in performance.

No high provisioning rate issues from SR18 upgrade markets (Chicago, Peru) after revised recommendations.

HLR – Home Location Register, NPR – Number of Problems Reported, DOE – Design of Experiments, RPN – Risk Priority Number MOL – Maintenance of Line

Melody HLR

Page 6: Six Sigma Green Belt (Product Track)

The iDEN system is an integration of traditional Push-To-Talk (PTT), half-duplex, analog radio technology and feature-rich, full-duplex digital cellular communications. This integration of mobile communication technologies provides state-of-the-art functions and benefits to mobile users while optimizing the available infrastructure resources.

The iDEN system provides both full and half-duplex operations. This melding of communications methods allows much of the voice traffic to be run in half-duplex mode, while providing full-duplex functionality when required.

Page 7: Six Sigma Green Belt (Product Track)

- The iHLR Provisioning System is built on the framework of a Web Server. Transactions sent to the provisioning server are handled by the instantiation of provisioning servlets. These servlets are responsible for updating the subscriber database and propagating the provisioning changes to the iDEN network.

-The provisioning interface to iHLR (iPP), is based upon HTTP/TCP/IP over ethernet.

Page 8: Six Sigma Green Belt (Product Track)

• HLR Provisioning interface, allows end users to open multiple provisioning sessions ( max 20) using http requests over tcp/ip connection. Provisioning requests are initiated by provisioning clients based on Motorola provided framework termed as iPP ( iDEN Provisioning Protocol).

• HLR handles the provisioning requests using a provisioning server and in turn update database, send message to DAP through MCP tasks and responds back to provisioning client with error or success code.

• HLR’s capability to process these requests is governed by several factors such as mobility message requests, cpu utilization, ethernet utilization, DB response etc. Provisioning craft person ( operator) is suppose to keep the provisioning rate under a certain limit to allow HLR to successfully process all the requests.

Page 9: Six Sigma Green Belt (Product Track)

• HLR lacks the ability to dynamically determine the provisioning requests it can handle at any given time, which may vary with box performance and other conditions such as load shedding. This results in under utilization of HLR resources under non-busy hour traffic.

• This project aims to analyze the critical parameters affecting the provisioning performance. Identified critical parameter will be analyzed to optimize provisioning performance using modeling simulation and prototyping.

Page 10: Six Sigma Green Belt (Product Track)

Phase Green Belt Deliverables DFSS Tools

Project Definition SDFSS Charter Charter

Requirements VOC, Requirements Development

and Analysis

QFD, UML Use Case,

Cognition Cockpit

Requirements Perform and document Critical

Parameter flow down and analysis,

Critical Parameter Tree,

Cognition Cockpit.

Requirements Documented risk analysis FMEA

Requirements Perform Initial transfer function. Critical Parameter Tree,

Cognition Cockpit.

Page 11: Six Sigma Green Belt (Product Track)

Architecture Evaluate Architecture Evaluate Architecture

using Modeling &

Simulation Prototyping.

DOE

Architecture Update risk analysis FMEA

Architecture Measure non-functional Critical

Parameter

Measured Quality

Attributes.

Page 12: Six Sigma Green Belt (Product Track)

SDFSS Requirements through Architecture

Flow Summary

D&I to optimize Provisioning Performance

SDFSS Methods Used to Optimize HLR Provisioning Performance

Raw VOC Input

Architecture

Simulation VOC HOQ

Use

Cases CPM DOE Brainstorm

Sessions

RunDuration

(D)

DB Size

(K)

MM Rate

(msg/sec)

Roamers

(%)

PT1

Rate

(msg/sec)

PT1

RTT

(ms)

PT1

msg/sec

(R1)

PT2

Rate

(msg/sec)

PT2

RTT

(ms)

PT2

msg/sec

(R2)

Prov

Response

(R1+R2))

Transactions in

prov log

(N)

( N/D )

msg/sec

MAX CPU

(%)

1 300 100 393 20% 55 34 18.08 55 35 18.06 36.14 10912 36.37 68

2 300 100 393 0% 25 30 31.46 25 28 32.73 64.19 19390 64.63 70

3 300 100 393 0% 30 30 31.63 30 28 32.94 64.57 19486 64.95 66

4 300 100 393 0% 20 30 31.55 20 28 32.86 64.41 19442 64.81 68

5 300 100 100 20% 20 25 38.01 20 25 36.33 74.34 22428 74.76 56

6 300 100 100 20% 25 26 37.35 25 25 36.55 73.9 22294 74.31 57

7 300 100 100 0% 25 24 39.6 25 23 38.96 78.56 23686 78.95 58

8 300 100 100 0% 20 24 39.62 20 23 38.72 78.34 23650 78.83 56

9 300 100 100 0% 40 24 24.87 40 22 24.85 49.72 15004 50.01 44

11 300 800 393 20% 25 42 23.07 25 42 22.49 45.56 13748 45.83 72

12 300 800 393 20% 40 41 23.6 40 41 22.9 46.5 14038 46.79 70

13 300 800 393 20% 45 36 22.1 45 35 22.1 44.2 13338 44.46 71

14 300 800 393 20% 55 35 18.08 55 34 18.08 36.16 10739 35.80 66

15 300 800 393 0% 40 29 24.82 40 28 24.87 49.69 15002 50.01 67

16 300 800 393 0% 30 31 30.87 30 30 30.98 61.85 18680 62.27 71

17 300 800 393 0% 35 30 28.39 35 30 27.98 56.37 17022 56.74 69

18 300 800 393 0% 40 27 24.86 40 26 26.87 51.73 15999 53.33 65

19 300 800 100 20% 20 24 40.5 20 25 36.31 76.81 22840 76.13 55

20 300 800 100 20% 25 34 39.69 25 25 35.75 75.44 22766 75.89 56

21 300 800 100 0% 25 24 39.72 25 24 37.35 77.07 23262 77.54 56

22 300 800 100 0% 20 24 40.72 20 25 36.47 77.19 23287 77.62 52

23 300 800 100 0% 15 25 38.43 15 25 35.62 74.05 22351 74.50 55

24 300 800 100 0% 20 25 38.4 20 25 36.4 74.8 22586 75.29 57

Verification 100 250 0% 25 27 35.01 25 27 35.01 70.02 19696 65.65 60

1 300 100 250 20% 30 34 28.93 30 33 28.49 57.42 17328 57.76 63

2 300 800 118 20% 25 24 39.66 25 26 35.6 75.26 22717 75.72 57

3 300 800 118 20% 20 25 38.51 20 26 34.89 73.4 22148 73.83 57

4 300 800 237.5 20% 20 28 34.26 20 28 32.63 66.89 20184 67.28 61

First House Of Quality for Project HLR Provisioning Project

Rating Links Legend

H (9) = High Effect

M (3) = Medium Effect

L (1) = Low Effect

0 = No Effect

Roof Links Legend

Equal Effect: =

No Effect: 0 or Blank String

Relative Effect: +, ++, -, --

Direction of Goodness Legend

Increases Customer Satisfaction: +

Decreases Customer Satisfaction: -

On Target For Customer Satisfaction: 0 or Blank String

9 H H L

4 H M

8 M H M M

10 H M M

3 H M

6 M H M

8 M H M H

8 H M M M

153 105 240 108 108 51 42 18 96 54

15.7% 10.8% 24.6% 11.1% 11.1% 5.2% 4.3% 1.8% 9.8% 5.5%

6 4 10 4 4 2 2 1 4 2

0 0 0 0 0 0 0 0 0 0

System Requirements

Direction Of Goodness

Voice of Customer, VOC's Imp

ort

an

ce

HLR

will

send I

DB

C w

arn

ing a

nd O

MC

ala

rm if

rate

is e

xceeded

HLR

will

contr

ol in

com

img p

rovis

ionin

g

flow

HLR

will

optim

ize M

M t

ask a

nd c

pu

utiliz

ation f

or

pro

vis

ionin

g r

equests

HLR

will

c

om

e o

ut

of

load c

onditio

n

as

soon a

s p

ossib

le

Faile

d p

rovis

ionin

g t

ransation w

ill b

e r

e-

trie

d.

Deta

iled p

rovis

ionin

g t

ransaction

info

rmation w

ill b

e c

olle

cte

d

and s

ore

d

IDB

C e

rror

and s

uccess r

esponses w

ill

have m

ore

info

rmation

HLR

will

support

few

er

pro

vis

ionin

g

transactions u

nder

load c

onditio

ns.

HLR

will

add a

dditio

nal lo

cal ala

rms f

or

inte

rmid

iate

conditio

ns.

Faile

d M

M m

essages w

ill b

e r

etr

ied

Minimize number of HLR provisioning related field

cases

Maximize HLR supported provisioning rate

Minimize load shedding duration in case of exceeded

provisioning rate

Minimize service impact if provisioning operator

stresses the system.

Maximize HLR provisioning transaction information

Minimize failed provisioning transactions

Minimize DVLR and HLR not in synch issues due to

high provisioning rate

Minimize time taken to root cause the issue to high

provisioning rate

Units

Scoring Totals

Relative Scores

Normalized Scores

Target Nominal Values

Start ProvisioningProvisioning

Client

Close Prov Session

Open Prov Session

Authenticate new session<<include>>

Update DB

Send MM Message

Send IDBC response back to

client

<<include>>

<<include>>

<<include>>

Check load condition<<include>>

<<include>>

Experiments run HA HLR to measure impact of factors on provisioning rate

Transfer function found by statistics analysis indicates which factor impacts the provisioning rate and on this basis identifes scenarios where max prov can be supported.

200

72.5

70.0

67.5

65.0

62.5

60.0

57.5

55.0

Roamers

Me

an

100

800

of subs

Number

Number of subs and Roamers

393100

80

70

60

50

40

MM Rate

Me

an

100

800

of subs

Number

Number of subs and MM Rate

200

80

70

60

50

40

30

Roamers

Me

an

100

393

Rate

MM

MM Rate and Roamers

393100 200

80

60

40

80

60

40

Number of subs

MM Rate

Roamers

100

800

of subs

Number

100

393

MM Rate

Number of subs, Roamers and MM Rate

House of Quality is build to identify most critical customer need.

decompose requirement into its functional pieces

Page 13: Six Sigma Green Belt (Product Track)

Business Case •There are close to 112 HLR deployment across the globe which comprises of iDEN, Harmony and

Melody releases. Existing implementation do not have any mechanism to control or monitor the

provisioning flow. HLR provisioning performance remains same even when box resources are under

utilized or box experiences load conditions. This project addresses this issue and provides a solution

for optimizing provisioning performance to be implemented in new releases.

•Field issues due to high provisioning rate ( than recommended) are escalated to development for

analysis. The solution will monitor the rate and send alarm for high provisioning, thus reducing the

number of such field cases escalated to DART or development. In 2008, 4 field cases out of total 13,

were investigated for high provisioning rate by development

•Working on this project will reduce the field support and improve customer satisfaction.

Opportunity Statement Potential gains from this solution :

•Key focus of the project will be to identify critical parameters affecting customer’s provisioning

performance.

•Results will enable development team to implement improved provisioning performance in new

releases.

•Implemented monitoring solution in future releases will reduce number of field cases. Each of these

cases are complex in nature and consume lots of effort, this will approximately reduce support effort

by 25%.

•In some cases customer stresses the system beyond recommended rate. By completing this project

we can make the system more robust and better handle customer business.

What pain are you currently experiencing :

•Under utilized provisioning performance.

•No mechanism to control provisioning flow.

•Many HLR field cases (NPR) are root caused to higher provisioning rate resulting in other functional

issues. In 2008, 8 out of 16 and in 2009 7 out of 20 were investigated for HPR. Each of these cases are

complex in nature and consume lots of effort.

•Sometimes customer stresses the system beyond prescribed limit, it causes issues which impacts

system functioning.

Goal Statement Use DFSS tools/techniques :

•To identify critical parameters based on customer's provisioning performance needs.

•To analyze architecture, create a performance model that predicts provisioning performance.

•To identify and maximize highest provisioning rate which can be supported hence better meeting with

customer need.

•Performance Model should predict the following metrics :

1. System Capacity : Typically indicated by provisioning capacity

Project Scope The scope of the project :

•Analyze provisioning performance parameters from customer need perspective.

•Identify product risk and mitigate it.

•Enhance product robustness.

•Specify DOEs, perform analysis and make provisioning performance predictions

Project Plan Plan Start Plan End Actual End

------------------------------------------------------------------------------------

•Requirement Development

and CP Identification 07/06/2009 08/31/2009 08/31/09

•Documented Risk Analysis

and Initial transfer function 09/01/2009 10/30/2009 11/31/10

•DOE Plan, Execute

and analyze) 03/01/2010 03/31/2010 03/31/10

•Lesson Learned and

Project closure 03/31/2010 04/23/2010 04/23/10

Team Role Name Expertise

-------------------------------------------------------------------------------------------------------------------

Sponsor HLR Box Manager

Champion Engineering Manager

Team Leader Srivastava Praveen GB Candidate

Team Member HLR Architecture

Team Member Bangalore DSS support

DSS Consultants Black belt

DSS Consultants Master Black belt

Technical Lead HLR/DAP/SSC Box Manager

Page 14: Six Sigma Green Belt (Product Track)

Phase Green Belt Deliverables DFSS Tools

Project Definition SDFSS Charter Charter

Requirements VOC, Requirements Development

and Analysis

QFD, UML Use Case,

Cognition Cockpit

Requirements Perform and document Critical

Parameter flow down and analysis,

Critical Parameter Tree,

Cognition Cockpit.

Requirements Documented risk analysis FMEA

Requirements Perform Initial transfer function. Critical Parameter Tree,

Cognition Cockpit.

Page 15: Six Sigma Green Belt (Product Track)

HLR Provisioning requirements IR

Minimize number of HLR provisioning related field cases 9

Maximize HLR supported provisioning rate 4

Minimize load shedding duration in case of exceeded provisioning rate 8

Minimize service impact if provisioning operator stresses the system. 10

Maximize HLR provisioning transaction information 3

Minimize failed provisioning transactions 6

Minimize DVLR and HLR not in synch issues due to high provisioning rate 8

Minimize time taken to root cause the issue to high provisioning rate 8

Had multiple interaction with technical team and DART (Support Team) to arrive at the requirements and importance rating.

Page 16: Six Sigma Green Belt (Product Track)

First House Of Quality for Project HLR Provisioning Project

Rating Links Legend

H (9) = High Effect

M (3) = Medium Effect

L (1) = Low Effect

0 = No Effect

Roof Links Legend

Equal Effect: =

No Effect: 0 or Blank String

Relative Effect: +, ++, -, --

Direction of Goodness Legend

Increases Customer Satisfaction: +

Decreases Customer Satisfaction: -

On Target For Customer Satisfaction: 0 or Blank String

9 H H L

4 H M

8 M H M M

10 H M M

3 H M

6 M H M

8 M H M H

8 H M M M

153 105 240 108 108 51 42 18 96 54

15.7% 10.8% 24.6% 11.1% 11.1% 5.2% 4.3% 1.8% 9.8% 5.5%

6 4 10 4 4 2 2 1 4 2

0 0 0 0 0 0 0 0 0 0

System Requirements

Direction Of Goodness

Voice of Customer, VOC's Impo

rtanc

e

HLR

will s

end I

DBC

warni

ng an

d OMC

alarm

if rat

e is e

xcee

ded

HLR

will c

ontro

l inco

mimg

prov

ision

ing

flow

HLR

will o

ptimi

ze M

M tas

k and

cpu

utiliz

ation

for p

rovisi

oning

requ

ests

HLR

will

come

out o

f load

cond

ition

as

soon

as po

ssibl

e

Faile

d prov

ision

ing tr

ansa

tion w

ill be

re-

tried.

Detai

led pr

ovisi

oning

tran

sacti

on

inform

ation

will

be co

llecte

d an

d sore

d

IDBC

error

and s

ucce

ss re

spon

ses w

ill

have

more

infor

matio

n

HLR

will s

uppo

rt few

er pro

vision

ing

trans

actio

ns un

der lo

ad co

nditio

ns.

HLR

will a

dd ad

dition

al loc

al ala

rms f

or

interm

idiate

cond

itions

.

Faile

d MM

mess

ages

will

be re

tried

Minimize number of HLR provisioning related field

cases

Maximize HLR supported provisioning rate

Minimize load shedding duration in case of exceeded

provisioning rate

Minimize service impact if provisioning operator

stresses the system.

Maximize HLR provisioning transaction information

Minimize failed provisioning transactions

Minimize DVLR and HLR not in synch issues due to

high provisioning rate

Minimize time taken to root cause the issue to high

provisioning rate

Units

Scoring Totals

Relative Scores

Normalized Scores

Target Nominal Values

Page 17: Six Sigma Green Belt (Product Track)

Critical Parameter

Provisioning Rate of HLR Application is identified as critical parameter.

Provisioning operator controls the flow of provisioning rate.

Incoming provisioning messages impacts HLR functional capabilities.

Page 18: Six Sigma Green Belt (Product Track)

• Preconditions: – HLR is up and running – Provisioning client is connected and ready to send provisioning transactions. – DAP is sending MM messages to HLR for call processing.

• Actor(s): – Provisioning client – Dispatch application processor ( DAP) – Operations and Maintenance control (OMC)

• Flow of Events: 1. Provisioning user starts a sessions by sending a login requests to HLR. 2. Provisioning server on HLR authenticates the session and allow user to login. 3. Provisioning client sends a provisioning requests over tcp/ip. 4. Provisioning server validates the request and parse it. 5. Provisioning server checks the load conditions on the box and send a IDBC messages for database

updates. Underlying database APIS are called to update the database.. 6. Provisioning server determines if requests requires an unsolicited messages to be send to DAP. Mobility

management (MM) messages is send to MM tasks for dap communication. MM task generates the unsolicited request to DAP depending upon it retry and pending queue.

7. Provisioning server receives ack from DB and MM task and send a success response to provisioning client. 8. End of use case

• Post Conditions: 1. HLR database is updated with new provisioning information. 2. DAP’S VLR is updated and synched with HLR database. 3. Provisioning client receives a success response with success code.

Page 19: Six Sigma Green Belt (Product Track)

• Alternate Flow #1: HLR rejects the new provisioning session requests 1. HLR is already under load conditions and new provisioning session service is not supported.

2. Provisioning client is send an error response and provisioning requests can not be initiated.

3. End of Alternate Flow # 1

◦ Post Conditions Alternate Flow #1: 1. An IDBC error response is send to client.

• Alternate Flow #2: The Database updates fail.

1. A critical alarm is generated and send to HLR alarm logs.

2. Provisioning server is send an error response which in turns generate an error response and send to provisioning client.

3. No unsolicited MM message is send to DAP.

◦ Post Conditions Alternate Flow #2: 1. An IDBC error response is send to client.

2. A critical local alarm is send to HLR application alarm logs.

• Alternate Flow #3: The Unsolicited MM messages are blocked by other messages in retry/pending queue. 1. Success ack is sent to provisioning server which in turns send success IDBC ciode to provisioning client.

2. MM messages in kept in retry/pending and re tried at regular intervals through a separate procedure. – Notes. Some times unsolicited MM messages are stuck in retry/pending queue due to other system

conditions such as few blocked messages for an IGW in the market. These messages remain in retry/pending queue until the blocker messages are cleared. This some times may take even few days. This condition creates a situation where DAP’s VLR and HLR are not in synch and may results in failed subscriber registration.

◦ Post Conditions Alternate Flow #3 1. Unsolicited MM messages are tried at regular interval and till thes messages are cleared HLR and DAP’s VLR will not

be in synch.

Page 20: Six Sigma Green Belt (Product Track)

Start ProvisioningProvisioning

Client

Close Prov Session

Open Prov Session

Authenticate new session<<include>>

Update DB

Send MM Message

Send IDBC response back to

client

<<include>>

<<include>>

<<include>>

Check load condition<<include>>

<<include>>

Page 21: Six Sigma Green Belt (Product Track)

• HLR Rejecting subscriber registration request – If HLR is already under load condition due to busy hour

traffic, continuous flow of provisioning requests may force HCPT to lose incoming provisioning requests.

– Provisioning at a rate higher than recommended rate can fill retry/pending queue due to which subscribers of that fleet have no Service for some time.

• HLR goes to load shedding – If customer stresses the provisioning system, it can

cause HLR to load shed impacting other functionalities.

Page 22: Six Sigma Green Belt (Product Track)
Page 23: Six Sigma Green Belt (Product Track)
Page 24: Six Sigma Green Belt (Product Track)

P- Diagram

CONTROL FACTORS

- Netowrk Bandwidth

- Connectivity

- MAP dialogue limitation

between DAP-HLR

External Comunication

- Database response time

- DB Application capability to

handle transactions

- Database centrice intensive

operations such as backup,

update stattistics, subscriber

reports etc.

Database Related

HLR is up and running

MM and CP messages

Provisioning Requests

SIGNAL

NOISE FACTORS

Higher Supported Provisioning Rate

Dynamic Provisioning threshold

Determination

RESPONSE

Hardware

----------------------

Number of CPUs

Available Memory

Processor

Ethernet capability

Software and Others

------------------------

MM Application

DB Application

Load levels

Number of subscribers

Webserver performance

Database performance

MM Rate

Load levels

HLR PROVISIONING PROJECT

HLR Supported Provisioning Rate

Unit - Message/Second

HLR Goes to load shedding

Can't Support higher rate

Failed Provisioning transactions

ERROR STATES

Those parameters that change with

deviations caused by process or

design.Unexpected Variation due to external

environment,internal, piece-to-piece, customer

usage, wear out)

What Errors , or failure modes are we likely to

see impact the ideal function performed as

a result of pottential "noise factor" interactions.

Here we conside the result of noise factors on

the ideal function.

The desired input signals

necessary to provide the

desired output. Causes the

system to deliver the user

intent

These are the customer intended

ideal functions. List all the positive

functions or the attributes from the

boundary diagram.System output that

determines the perceived result)

These are the design parameters whose

nominal values can be adjusted by the user

or programming modes in the software)

.

Page 25: Six Sigma Green Belt (Product Track)

VOC related to CP

Initial Transfer Function :

Max Prov rate = f(MM_traffic, # of subscribers,roamers)

Page 26: Six Sigma Green Belt (Product Track)
Page 27: Six Sigma Green Belt (Product Track)

Architecture Evaluate Architecture Evaluate Architecture

using Modeling &

Simulation Prototyping.

DOE

Architecture Update risk analysis FMEA

Architecture Measure non-functional Critical

Parameter

Measured Quality

Attributes.

Page 28: Six Sigma Green Belt (Product Track)

Provisioning Rate: 1. Evaluate Architecture : 7 Step DOE 2. Measure Performance for Critical

Parameter

Page 29: Six Sigma Green Belt (Product Track)

• Design of Experiments conducted in Bangalore, India Lab

• Experiments carried out on Melody HLR software

• Hardware and Software – Dual Chassis ATCA (7880) blades with HLR Melody

version - SR18.00.17

– Sun based Provisioning Loader version – R13.06.00

– Sun Based Map loader – R12.00.04

– Results observed on Melody hardware are applicable to iDEN (TS40) HLR.

Page 30: Six Sigma Green Belt (Product Track)

• Critical Parameter: – For the critical parameter understand factors affecting:

• Provisioning Rate

• To determine the factors that will be the main effects on satisfying this requirement

• To determine the relationship between the factors

• To identify the transfer function between the factors and this requirement.

Page 31: Six Sigma Green Belt (Product Track)

• Tests were done on a Melody HLR (ATCA 7880), bi nodal setup in Bangalore iDEN lab. Provisioning messages were sent to HLR using 2 provisioning loaders. Provisioning rate is configurable from loader. Mobility and Call Processing messages were generated using map loader with standard load profile for each scenario.

• Maximum supported provisioning rate without sending HLR to loadshedding was determined using this.

Page 32: Six Sigma Green Belt (Product Track)

Final RTT - RoundTripTime

Run MM - Mobility Message

Verification PT1 - Prov Tool 1/2

RunDuration

(D)

DB Size

(K)

MM Rate

(msg/sec)

Roamers

(%)

PT1

Rate

(msg/sec)

PT1

RTT

(ms)

PT1

msg/sec

(R1)

PT2

Rate

(msg/sec)

PT2

RTT

(ms)

PT2

msg/sec

(R2)

Prov

Response

(R1+R2))

Transactions in

prov log

(N)

( N/D )

msg/sec

MAX CPU

(%)

1 300 100 393 20% 55 34 18.08 55 35 18.06 36.14 10912 36.37 68

2 300 100 393 0% 25 30 31.46 25 28 32.73 64.19 19390 64.63 70

3 300 100 393 0% 30 30 31.63 30 28 32.94 64.57 19486 64.95 66

4 300 100 393 0% 20 30 31.55 20 28 32.86 64.41 19442 64.81 68

5 300 100 100 20% 20 25 38.01 20 25 36.33 74.34 22428 74.76 56

6 300 100 100 20% 25 26 37.35 25 25 36.55 73.9 22294 74.31 57

7 300 100 100 0% 25 24 39.6 25 23 38.96 78.56 23686 78.95 58

8 300 100 100 0% 20 24 39.62 20 23 38.72 78.34 23650 78.83 56

9 300 100 100 0% 40 24 24.87 40 22 24.85 49.72 15004 50.01 44

11 300 800 393 20% 25 42 23.07 25 42 22.49 45.56 13748 45.83 72

12 300 800 393 20% 40 41 23.6 40 41 22.9 46.5 14038 46.79 70

13 300 800 393 20% 45 36 22.1 45 35 22.1 44.2 13338 44.46 71

14 300 800 393 20% 55 35 18.08 55 34 18.08 36.16 10739 35.80 66

15 300 800 393 0% 40 29 24.82 40 28 24.87 49.69 15002 50.01 67

16 300 800 393 0% 30 31 30.87 30 30 30.98 61.85 18680 62.27 71

17 300 800 393 0% 35 30 28.39 35 30 27.98 56.37 17022 56.74 69

18 300 800 393 0% 40 27 24.86 40 26 26.87 51.73 15999 53.33 65

19 300 800 100 20% 20 24 40.5 20 25 36.31 76.81 22840 76.13 55

20 300 800 100 20% 25 34 39.69 25 25 35.75 75.44 22766 75.89 56

21 300 800 100 0% 25 24 39.72 25 24 37.35 77.07 23262 77.54 56

22 300 800 100 0% 20 24 40.72 20 25 36.47 77.19 23287 77.62 52

23 300 800 100 0% 15 25 38.43 15 25 35.62 74.05 22351 74.50 55

24 300 800 100 0% 20 25 38.4 20 25 36.4 74.8 22586 75.29 57

Verification 100 250 0% 25 27 35.01 25 27 35.01 70.02 19696 65.65 60

1 300 100 250 20% 30 34 28.93 30 33 28.49 57.42 17328 57.76 63

2 300 800 118 20% 25 24 39.66 25 26 35.6 75.26 22717 75.72 57

3 300 800 118 20% 20 25 38.51 20 26 34.89 73.4 22148 73.83 57

4 300 800 237.5 20% 20 28 34.26 20 28 32.63 66.89 20184 67.28 61

Page 33: Six Sigma Green Belt (Product Track)

Full Factorial Design created in Minitab for 3 factors Factors: 3 Base Design: 3, 8 Runs: 8 Replicates: 2 Blocks: none Center pts (total): 3

StdOrder RunOrder CenterPt Blocks Number of subs MM Rate Roamers Provisioning Rate

1 1 1 1 100 100 0 78.34

2 14 1 1 800 100 0 77.07

3 8 1 1 100 393 0 64.57

4 2 1 1 800 393 0 51.73

5 5 1 1 100 100 20 73.9

6 15 1 1 800 100 20 75.44

7 12 1 1 100 393 20 36.14

8 3 1 1 800 393 20 36.16

9 16 1 1 100 100 0 78.08

10 10 1 1 800 100 0 77.19

11 6 1 1 100 393 0 64.81

12 13 1 1 800 393 0 49.69

13 9 1 1 100 100 20 74.34

14 11 1 1 800 100 20 76.81

15 7 1 1 100 393 20 36.14

16 4 1 1 800 393 20 36.16

Page 34: Six Sigma Green Belt (Product Track)

Number of subs

Roamers

MM Rate

800100

200200

393100393100393100393100

80

70

60

50

40

30

Pro

vis

ion

ing

Ra

te

Box Plot of Provisioning Rate

Page 35: Six Sigma Green Belt (Product Track)

DOE Step 2 & 3: Create Model Estimated Effects and Coefficients for Provisioning Rate (coded units)

Term Effect Coef SE Coef T P

Constant 61.66 0.1578 390.84 0

Number of subs -3.26 -1.63 0.1578 -10.33 0

MM Rate -29.47 -14.74 0.1578 -93.4 0

Roamers -12.05 -6.02 0.1578 -38.19 0

Number of subs*MM Rate -3.72 -1.86 0.1578 -11.79 0

Number of subs*Roamers 4.27 2.14 0.1578 13.54 0

MM Rate*Roamers -9.5 -4.75 0.1578 -30.11 0

Number of subs*MMRate*Roamers 2.73 1.36 0.1578 8.65 0

S = 0.631056 PRESS = 12.7434 R-Sq = 99.93% R-Sq(pred) = 99.72% R-Sq(adj) = 99.87

1.Significant factors have p-value < .05 2. Model is good because R-Sq and R-Sq (adjusted) has difference of less than 10%

Page 36: Six Sigma Green Belt (Product Track)

ABC

A

AB

AC

BC

C

B

9080706050403020100

Te

rm

Standardized Effect

2.31

A Number of subs

B MM Rate

C Roamers

Factor Name

Pareto Chart of the Standardized Effects(response is Provisioning Rate, Alpha = .05)

MM Rate is most significant factor followed by Roamers and then followed by MM Rate and Roamer interactions.

Page 37: Six Sigma Green Belt (Product Track)

0-25-50-75-100

99

95

90

80

70

60

50

40

30

20

10

5

1

Standardized Effect

Pe

rce

nt

A Number of subs

B MM Rate

C Roamers

Factor Name

Not Significant

Significant

Effect Type

ABC

BC

AC

AB

C

B

A

Normal Plot of the Standardized Effects(response is Provisioning Rate, Alpha = .05)

Significant effects are further away from the line in Normal Probability Plot and marked in red

Page 38: Six Sigma Green Belt (Product Track)

Source DF Seq SS Adj SS Adj MS F P

Main Effects 3 4097.39 4097.39 1365.80 3429.65 0.000

2-Way Interactions 3 489.46 489.46 163.15 409.70 0.000

3-Way Interactions 1 29.78 29.78 29.78 74.79 0.000

Residual Error 8 3.19 3.19 0.40

Pure Error 8 3.19 3.19 0.40

Total 15 4619.82

Since p value for main effect and 2 way interaction and 3 way interaction is < 0.05. All are significant.

Page 39: Six Sigma Green Belt (Product Track)

1.00.50.0-0.5-1.0

99

90

50

10

1

Residual

Pe

rce

nt

8070605040

1.0

0.5

0.0

-0.5

-1.0

Fitted Value

Re

sid

ua

l

1.00.50.0-0.5-1.0

8

6

4

2

0

Residual

Fre

qu

en

cy

16151413121110987654321

1.0

0.5

0.0

-0.5

-1.0

Observation Order

Re

sid

ua

l

Normal Probability Plot Versus Fits

Histogram Versus Order

Residual Plots for Provisioning Rate

Residual vs. Fitted (Predicted) value and residual vs run order plots do not show any patterns. Residuals are normally distributed (See Normal Probability Plot)

Page 40: Six Sigma Green Belt (Product Track)

Term Coef Constant 82.3497 Number of subs 0.00474676 MM Rate -0.0398537 Roamers 0.217479 Number of subs*MM Rate -6.28961E-05 Number of subs*Roamers -

4.57326E-05 MM Rate*Roamers -0.00444015 Number of subs*MM Rate*Roamers 2.66090E-

06

Estimated Coefficients for Provisioning Rate using data in uncoded units

Y= Predicted Provisioning Rate, Number of subs = s, MM rate = m and Roamer (%) = r

Y = 82.3497 +0.00474*s -0.03895*m +0.2174*r - 0.0000628*s*m- .0000457*s*r- 0.0044*m*r + .0000266*s*m*r

Transfer Function :

Page 41: Six Sigma Green Belt (Product Track)

800100

80

70

60

50

393100

200

80

70

60

50

Number of subs

Me

an

MM Rate

Roamers

Main Effects Plot for Provisioning RateData Means

Page 42: Six Sigma Green Belt (Product Track)

393100 200

80

60

40

80

60

40

Number of subs

MM Rate

Roamers

100

800

of subs

Number

100

393

MM Rate

Interaction Plot for Provisioning RateData Means

Page 43: Six Sigma Green Belt (Product Track)

• Pareto chart shows that MM Rate ( mobility & call processing message) has most significant effect on provisioning rate supported by HLR.

• The next significant factor is percentage of roamers. More the roamer percentage, less the provisioning capability.

• Interaction of MM rate rate and roamer is next is list and has significant effect on provisioning rate. A minimum of both MM rate and roamer percentage will yield maximum supported rate at any given point of time.

• Number of subscribers and in 3 way Interactions between factors do not impact the HLR provisioning capability significantly.

• However, number of mm & cp messages will be generally driven by registration requests which in turn is directly related to number of active subscribers in the HLR db, number of active subs in database does impact the provisioning capability indirectly. Inactive subscribers do not impact provisioning capability.

Page 44: Six Sigma Green Belt (Product Track)

• Total number of mm & cp messages are driven by number of registrations and number of SRI requests. We know that at times mm and cp messages are less when compared to busy hour rate, for example during : – Maintenance window ( 41.00 registration per sec) – 7-9 AM Registration Period (83.33 registration per sec )

• During such times, HLR can support more provisioning transactions. Model predicts peak provisioning factor for each such scenario depending on the mm & cp messages and roamers present in the market.

Page 45: Six Sigma Green Belt (Product Track)

• Provisioning Model

• Y = 82.3497 +0.00474*s -0.03895*m +0.2174*r - 0.0000628*s*m- .0000457*s*r- 0.0044*m*r + .0000266*s*m*r

• Where

– Y = Supported Provisioning Rate per second.

– m = mm and cp messages per sec

– r = percentage of roamers

– s = number of subscribers in K

• According to provisioning model, Maintenance window represented by 41.67 registration per second with 20 % roaming, will have provisioning peak factor as 2.0 and can support upto 15 msg per second. This is verified in Box test lab using maintenance window load profile.

• HA HLR supported busy hour provisioning profile is benchmarked for market conditions as 7.3 msg per sec where as lab throughput for busy hour is 36.16 mgs per sec. To get supported provisioning rate in market condition a factor of 36.14/7.3 = 4.94 is used.

• The iDEN markets are informed through a revised bulletin issued to the marked based on DOE results. This will help them to better plan their high provisioning needs during maintenance window or at a time when mm traffic is minimum

Page 46: Six Sigma Green Belt (Product Track)

Failure Modes and Effects Analysis (FMEA):

Risk Category /

Item

Potential Failure Mode / Effects Potential Effects of Failure (What is

the impact on the customer?)

Risk

Priority

Number

(RPN)

Recommended Action

Ris

k P

rio

rity

Customer impact

due to high

provisioning rate.

High provisoning rate might cause

large number of un-solicited

messages initiated for the DAP.

For example modify subs or modify

fleet will generate ISD and IFD

HLR may "lose" in incoming

registration requests impacting

subscriber service. 45

1)Inform operator about the

HLRs capacity to support

maximum provisioning rate.

So that they do not exceed

the recommended rate.

Also intimate markets of

the window when maximum

proviosning rate can be

exceeded. Based on this

information market can

defer theit high provisioning

needs to such window. 24

(Revised)

Page: 1 of 1

Results

Software Failure Mode and Effects Analysis (FMEA)

Product:iHLR Provisioning Project FMEA Date: Dec 17, 2010

FMEA Team Members: Praveen,HLR Development Team

Risk Before RPN After RPN

Service Impact 45 24

Functionalty Impact 36 20

Double click on the sheet to see the entire sheet

Page 47: Six Sigma Green Belt (Product Track)

• DOE results prove that HLR can supports higher provisioning rate during maintenance window and other times when mm rate is low.

• Defer High proviosning rate activities to maintenance window or other time when HLR is handling lower MM rate.

• Based on DOE analysis, future HLR releases can implement provisioning flow control and alarm mechanism to warn operator , if maximum supported rate is exceeded.

Page 48: Six Sigma Green Belt (Product Track)

• In 2008, 16 field cases were investigated by HLR development with 12.89 SM effort, out of these, 8 were investigated from high provisioning rate angle with 5.97 SM effort.

• In 2009, 20 field cases were investigated by HLR development with 15.08 SM effort , out of these, 7 were investigated from high provisioning rate angle with 7.80 SM effort .

• So, in 2008 and 2009 alone at least 13.87 SM effort was spent on understanding high provisioning rate for field cases. Rough estimate for implementing provisioning model and recommendations like high provisioning rate warning alarm and throttling mechanism is close to 3.5 SM.

• We could have saved at least 8-10 SM in 2 years by 1) Preventing high provisioning rate field cases from occurring 2) Straightaway ruling out high provisioning rate involvement in field case analysis

• Note – MOL effort used in the analysis does NOT include CNRC/DART effort for field case investigation. These teams also investigate issue along with development and log almost similar effort. If we add dart/cnrc efforts, overall saving will be much more than this.

Page 49: Six Sigma Green Belt (Product Track)

• Tests were run in BT environment to measure supported provisioning rate during maintenance window and 7-9 AM registration period.

• Maintenance Window Profile – Number of subs = 800 k – Registration Rate = 41.67 per second – mm and CP messages = 118 per second – Roamers = 20 % – Predicted Rate using Model = 72.93 – Prorated Provisioning Rate = 14.75 – Actual = 73.44 – Actual Prorated Provisioning Rate = 14.94

• 7 – 9 AM Registration Period – Number of subs = 800 k – Registration Rate = 83.33 per second – mm and CP messages = 237.49 per second – Roamers = 20 % – Predicted Rate using Model = 57.01 – Prorated Provisioning Rate = 11.53 – Actual = 65 – Actual Prorated Provisioning Rate = 13.53

• All the Values measured during verification Testing for supported provisioning rate are close to the Upper 95% Prediction interval. This validates the predictive provisioning model

Page 50: Six Sigma Green Belt (Product Track)

• The iDEN markets are informed through a revised bulletin based on DOE results on April 23,2010.

Page 51: Six Sigma Green Belt (Product Track)

From: Tewinkle Steve-EST002

Sent: Monday, June 28, 2010 8:24 PM

To: Tchon Scott-CST013

Cc: Srivastava Praveen-a23098; Kumar Taleki Prasanna-a22694; Daniels Thomas-CTD040

Subject: FW: : Green belt Project - success criterion and value statement - minutes

Scott,

Please find my value statement below for Praveen's Green Belt project:

The SDFSS Green Belt project that was done by Praveen for iHLR provisioning has identified improvements that can be made both

short and long term that will have positive impact on HLR Provisioning in terms of performance, capacity, and reduced overall

Support. Praveen's analysis was very thorough, data driven, and Customer focused. The benefits we can get from this project

are:

- This project identified the critical parameters that impact HLR provisioning as MM Rate and Roamer Percentages and based on this finding, we are able to recommend to the Customer that they can increase their maintenance window provisioning rate

from 7 to 15 msgs per second. This increased rate has already been documented in field bulletin and delivered to market in April 2010.

- Implementation of a new provisioning throttling mechanism that would regulate provisioning flow, warn the operator when maximum supported rate is being exceeded, and help reduce # of field related cases due to high provisioning rate. This will be proposed as a new iHLR feature for an upcoming system release.

- This project also will serve as a model for other iDEN teams to follow when faced with similar type of issues, etc on their box

Overall, Praveen did an outstanding job in approaching this project and his efforts will definitely result in improvements to the

provisioning process and reductions in field reported provisioning cases in the future.

Steve

Page 52: Six Sigma Green Belt (Product Track)

– The HLR’s provisioning capacity is a function of MM rate and Roamer’s percentage.

– Predictive model’s ability to dynamically determine maximum supported provisioning rate can be used to -

• Implement throttling mechanism to regulate provisioning flow.

• Implement warning ( alarm ) mechanism, so that provisioning operator can slow down if maximum supported provisioning rate is reached.

– These two implementations alone will enhance robustness and save substantial MOL effort.

– This is proposed for SR21 as a candidate feature by the Box Manager (Steve T)

Page 53: Six Sigma Green Belt (Product Track)

• What went well: – Efficient Simulation Model development – CPM Flow-down and engagement of stakeholders in identifying

critical parameters – Applying the results of DOE and issuance of revised bulletin to

customer on provisioning rate – Engagement of product team, field support team and box

management since start of the project.

• What could have gone better: – Do brainstorms and interviews direct with Customer

• Things I would do different in the future: – Work with champion and sponsor to ensure engagement of all

stakeholders in decision making at critical milestones

Page 54: Six Sigma Green Belt (Product Track)

Backup Up Slide

Page 55: Six Sigma Green Belt (Product Track)

• Service Impact (Original Text) – If provisioning transaction rates are exceeded discrepancies may occur between DLVR and iHLR

database or iHLR pending / retry queues may contain stuck transactions. Field case occurred on a TDAP that experienced registration issues due to exhaustion of internal fleet provisioning buffers.

• Service Impact (Changed) – If provisioning system is stressed and transaction rates exceed the published rate, following may

be observed: • Discrepancies may occur between DLVR and iHLR database • iHLR pending / retry queues may contain stuck transactions. • Provisioning transactions might fail due to database lock error etc. • iHLR can go to load shedding either due to high cpu utilization or filled hcpt queue • For example a field case occurred on a TDAP that experienced registration issues due to exhaustion of internal

fleet provisioning buffers.

• Note: (Original Text) – At times a higher provisioning throughput rates may be realized while the HAiHLR is not under

critical load but to eliminate possible issues it is recommended that these published rates be followed.

• Note: (Changed) – Note: It is recommended that these published rates be followed to avoid issues. However, HLR

supported provisioning rate is a function of mobility & call processing message rate and number of roamers in the market. Again, above message rate depends on registration requests to the HLR. A higher provisioning throughput may be realized while the HA HLR is not under critical load and processing less messages than compared to busy hour traffic. For example, during maintenance window with a provisioning peaking factor 2, HLR can support up to two times of following profile.

Page 56: Six Sigma Green Belt (Product Track)

MM Rate

(per sec)

Subscriber

(K)

Roamers

(%)

Predicted rate

(per sec)

Supported Provisioning

Rate

Provisioning

Peaking

Factor

100 100 20 73.9 14.95 2.0

393 100 20 36.14 7.31 1.0

100 800 0 77.07 15.59 2.1

393 100 0 64.57 13.06 1.8

100 800 20 75.44 15.26 2.1

393 800 0 51.73 10.47 1.4

393 800 20 36.16 7.32 1.0

100 100 0 78.34 15.85 2.2

*Supported provisioning rate = Prorated provisioning rate based on benchmark under market conditions.

#This rate was benchmarked with busy hour traffic profile.

*

#