complex networks enabling technologies capability …...fdny responders in fdny field units scope in...
TRANSCRIPT
© 2009 TASC, Inc.
Complex Networks Enabling Technologies Capability Development
2012 Annual IV&V Workshop
Jim Savarino, Jeff Northey, Shirley Savarino, Andy Chen, Bojan Cukic, Jesse Musgrove, John Ryan
© 2009 TASC, Inc.
IV&V Questions Addressed by CD
2
CD: Question addressed
Associated Tech Framework WBS
Develop modeling / simulation techniques capable of verifying that user needs are addressed.
2.2 System architecture meets user needs
For performance requirements, establish boundary conditions of how the system performs and assess robustness i.e.: margin against those requirements.
5.2 Ensure design provides required capability, reliably meets user needs, sufficiently stable for implementation
Develop a simulation that operates at varying levels of abstraction to address decomposition, interfaces and integration.
3.2 In-scope parent reqts. represented in appropriate child reqts., no unneeded capabilties added
3.4 Software interface reqts. adequate to meet needs of system
Define a data dictionary approach to address user scenarios that can feed into test analysis.
4.1.2 Ensure planned tests sufficient to confirm integrated system complies with system software reqts.
© 2009 TASC, Inc.
Scripting Products and Process Overview
3
Architecture (Simulink Model) - Varying levels of fidelity - Allows concatenation of Use Cases - Allows verification of interfaces
GUI Translates a use case to input data for each model.
The Simulink model reflects the system architecture at various levels of abstraction. Erlang distribution used for internal call processing parameters.
The Data Dictionary captures a Use Case and feeds the Simulink and Layered Queuing Network (LQN) models. We pulled all variables out of the model and made this aspect table driven
C# based Graphical User Interface (GUI) takes the parametric Use Case and turns it into individual calls and attributes for Simulink, call rate and call type probabilities for LQN.
Data Dictionary: Parameters and Variables for a particular instance (Use Case)
The LQN model allows rapid prototyping and is a “process model”. Inputs are parametric, characterizing the use case
Performance (LQN Model) - Performance across a time range - Evaluate Boundary Conditions
Models are validated against each other for correctness
© 2009 TASC, Inc.
A Representative System
4
911 Call
On- street
or In-
building call box
Computerized Triage Project
NYPD CAD Project
FDNY Fire/ EMS CAD Project
Brooklyn Hub Project (Redundancy to PSAC-1 Ties
to NYPD Radio Network)
Voice Alarms Project
ERS/ BARS
Project
Facilities Project
Logging and Recording
Project
Performance Monitoring
Project
Call Taker
Workforce Scheduling
Project
NYPD Radio Console/
ECS Project
FDNY Radio Console/
ECS Project
CPE/N
AC
D
Pro
ject
Network Projects (brown lines)
PSAC 2
Blue: Full Responsibility Projects Yellow: Oversight Coordination Projects
• We will perform IV&V on the integration and performance of the entire system (Pri 1)
• As required, this effort will evaluate other ECTP projects
• Additionally, there are seven projects which will receive “full” IV&V efforts
1. CPE/NACD (Pri 2)
2. Network (Pri 2)
3. NYPD Radio Console/ECS (Pri 3)
4. FDNY Radio Console/ECS (Pri 3)
5. FDNY Fire/EMS CAD (Pri 4)
6. Performance Monitoring (Pri 5)
7. Computerized Triage (Pri 6)
© 2009 TASC, Inc.
LQN of Police Department (stubbed out FD, EMS): Process Model
5
Spanish_Call_ Taker_Queue
Wait_for_Spa_
Call_Taker
[0s]
Verify_ Accidental_Call
[21s]
Hang_Up [3s]
CPE_NACD {inf}
Find_Call_ Taker [0s]
Get_ANI_ALI
[0s]
MSAG {inf}
Phone_to_ALI [2s]
English_Call_ Taker {50}
Opening [4s]
Get_Info_ Location
[28s]
Transfer_ Call [3s]
Spanish_Call_ Taker {8}
S_Get_ Location
[42s]
…
Get_Incident_Type
[13s]
Dial_ Appropriate
[12s]
Explain_ Dup [5s]
Continue_ Getting_Info_1
[20s]
Dropped_ Call_Redial
[3s]
Busy [3s]
No_Answer [15s]
Answer [5s]
P_MD P_U
Evaluate_ Language
[10s]
P_SPA
P_FOR
Ṗ_D
Ṗ_NE
F_Get _Info _Location
[84s]
F_Get_Incident_Type
[26s]
F_Dial_ Appropriate
[24s]
F_Explain_ Dup [10s]
Ṗ_D
Ṗ_NE
Contact_FLS [20s]
P_B P_A
English_Call_ Taker_Queue
Wait_for_Eng_ Call_Taker
[0s]
1-Ṗ_D
1-(P_FOR+P_SPA)
1-(P_B+P_A)
1-Ṗ_D
1-Ṗ_NE
1-Ṗ_POL
1-Ṗ_NE
1-(P_MD+P_U)
Classify_in_CAD [0s]
Continue_ Getting_Info_2
[77s]
F_Continue_ Getting_Info_1
[20s]
F_Continue_ Getting_Info_2
[174s]
English Call Takers
Spanish Call Takers
FLS and Accidental Calls
Emergency and Duplicate Calls
From Call Gen
To CAD
Dotted lines show Simulink model elements
© 2009 TASC, Inc.
LQN Performance Demonstration – Waiting Time
• For the first test, all values with the exception of Call_Rate were held constant for steady state values as listed in DataDictionary5, and the system was modeled with only the first section of the ECTP model as described in Design of the ECTP Layered Queueing Network10
• The Call_Rate value was tested at different values from 1250 to 12845 calls/hour in increasing increments of five
0
2
4
6
8
10
12
14
16
18
20
12115
12150
12185
12220
12255
12290
12325
12360
12395
12430
12465
12500
12535
12570
12605
12640
12675
12710
12745
12780
12815
Wait
Tim
e (
seco
nd
s)
Call Rate (Calls/Hour)
Wait Time in Queue for
English Call Taker
0
10
20
30
40
50
60
70
80
90
100
1255
1835
2415
2995
3575
4155
4735
5315
5895
6475
7055
7635
8215
8795
9375
9955
10535
11115
11695
12275
Call Rate (Calls/Hour)
Percent Utilization of
English Call Takers
© 2009 TASC, Inc.
Simulink Top Level Blocks (architecture model)
7
Conn1
Conn3
Conn5
Conn2
Conn4
Conn6
VESTA
PD Radio In
PD Report Back
FDNY Radio In
FDNY Report Back
EMD Radio In
EMD Report Back
PD Responders
FDNY Responders
EMD Responders
Radios
In Out
PD Responders
IN
NYPD Field Units Scope
In Out
FDNY Responders
IN
FDNY Field Units Scope
In Out
EMD Responders
IN
EMD Field Units Scope
Conn3
Conn1
Conn2
Conn4
E9-1-1
Calls Out
Call Generator
Conn1
Conn3
Conn5
Conn2
Conn4
Conn6
CAD
Call Generator
E911 VESTA CAD
Radios Reponders/ Field Units
© 2009 TASC, Inc.
Model Comparison – Simulink L2 Complex Network
8
2
Emergency Non-Duplicated
1
Call In
IN
Unintentional Calls
util
IN
OUT
Spanish Call Takers
in
Spa CT Util ization
#n
wIN
OUT
Spa CT FIFO Queue
in
Spa Avg Wait Time
IN1
IN2
IN3
OUT
Path Combiner
IN
Non-Emergency Calls
IN OUT
NYPD-SPN Calls Scope
IN OUT
NYPD-FLS Calls
IN OUT
NYPD-English Calls Scope
IN
Mis-Dialed Calls
IN
OUT1
OUT2
OUT3
OUT4
OUT5
Language and Mistake Calls
IN OUT
Foreign Language Service
util
IN
OUT
Eng Call Takers
in
Eng CT Util ization
#n
wIN
OUT
Eng CT FIFO Queue
in
Eng Avg Wait Time
IN
OUT1
OUT2
Emergency Switch
IN
Emergency Duplicated CallsIN
OUT1
OUT2
Duplicate Switch
IN OUT
Delayed Calls
in
# of Calls in Spa CT Queue
in
# of Calls in Egn CT Queue
English Call Takers
Spanish Call Takers
FLS and Accidental Calls
Emergency and Duplicate Calls
From Call Gen
To CAD
Dotted lines show LQN model elements
© 2009 TASC, Inc.
Simulink and LQN Results at 13k Calls/hour for 24 hrs (we wanted to make sure models were in sync)
9
CallRate AvgECTQueueTime ECT_Utilization EngProbOfDelay EngAcceptablePercent AvgSCTQueueTime SCT_Utilization SpaProbOfDelay SpaAcceptablePercent
13300 7.1544 0.97782 0.612330436 0.885415365 0.0001 0.465530556 6.55E-05 0.99999948
Simulink Results
LQN Results
This data avail in Simulink, not LQN
This data avail in Simulink, not LQN
Note: some variation due to translation (e.g. LQN had 13300 calls and Simulink had ~13500 calls/hour in the 24 hour period)
© 2009 TASC, Inc.
Status, and Method Assessment
10
CD: Question addressed
CD Status and Method Assessment Associated Tech Framework WBS
Develop modeling and simulation techniques capable of verifying that user needs are addressed.
LQN and Simulink models were implemented by creating a custom data dictionary with a GUI to automatically draw from the dictionary and create input files for the two models. This is a new, demonstrated technique for IV&V.
2.2 System architecture meets user needs
For performance requirements, establish boundary conditions for how the system performs and assess robustness i.e.: margin against those requirements.
The LQN model was used to establish boundary conditions which adequately simulated performance to provide an assessment of system capability against requirements and provided an assessment of robustness. In addition the model revealed there was a high sensitivity to changes in the boundary conditions. This is a new demonstrated technique for IV&V
5.2 Ensure design provides required capability, reliably meets user needs, sufficiently stable for implementation
Develop a simulation that operates at varying levels of abstraction to address decomposition, interfaces and integration.
Significant progress has been demonstrated towards achieving this objective. The Simulink level 1 model has been completed which represents system architecture at the highest level, derived from the driving artifacts. The model also includes a lower level of abstraction which includes additional details of the system. The level 2 model elaborates architecture, with an increased focus on processes. The level 2 model has been defined and potential Simulink tool limitations evaluated.
3.2 In-scope parent reqts. represented in appropriate child reqts., no unneeded capabilties added
3.4 Software interface reqts. adequate to meet needs of system
Define a data dictionary approach to address user scenarios that can feed into test analysis.
This has been accomplished by both the LQN and Simulink models. The LQN model can automatically vary the range of a parameter to produce a performance profile. The Simulink model can concatenate use cases to evaluate a complete scenario. This is a new demonstrated technique for IV&V.
4.1.2 Ensure planned tests sufficient to confirm integrated system complies with system software reqts.
We have drafted a method to cover the new techniques developed during this CD.