ricardo valerdi [email protected]/documents/presentations/ieee_ct_may07... ·...
TRANSCRIPT
1
IEE
E C
on
necti
cu
t S
ecti
on
May 2
4,
2007
Ricardo Valerdi
rvalerdi@
mit.edu
2
Ro
ad
map
(1)Description of the m
odel;
(2)Explanation of its size and cost
drivers;
(3)Lim
itations;
(4)Recent developments;
3
•USC CSSE Corporate Affiliate Program
•LAI Consortium Members
•Boeing, LMCO, NGC, Raytheon, General
Dynamics, SAIC, BAE Systems
•Practical Software & Systems Measurement
•INCOSE
–Measurement Working Group
–Corporate Advisory Board
COSYSMO Development Support
All m
odels are wrong…
…but some of them are useful.
5
Esti
mati
on
Accu
racy
Feasibility
Plans/Rqts.
Design
Develop
and Test
Phases and M
ilestones
Relative
Size
Range
Operational
Concept
Life Cycle
Objectives
Life Cycle
Architecture
Initial
Operating
Capability
x
0.5x
0.25x
4x
2x
6
Key D
efi
nit
ion
s &
Co
ncep
tsC
ali
bra
tio
n:
the t
un
ing
of
para
mete
rs b
ased
on
pro
ject
data
CE
R:
a m
od
el th
at
rep
resen
ts t
he c
ost
esti
mati
ng
rela
tio
nsh
ips o
f fa
cto
rs
Co
st
Esti
mati
on
: p
red
icti
on
of
bo
th t
he
pers
on
-eff
ort
an
d e
lap
sed
tim
e o
f a p
roje
ct
Dri
ver:
A f
acto
r th
at
is h
igh
ly c
orr
ela
ted
to
th
e a
mo
un
t o
f
Syste
ms E
ng
ineeri
ng
eff
ort
Para
metr
ic:
an
eq
uati
on
or
mo
del th
at
is a
pp
roxim
ate
d b
y
a s
et
of
para
mete
rs
Rati
ng
Scale
: a r
an
ge o
f valu
es a
nd
defi
nit
ion
s f
or
a
part
icu
lar
dri
ver
Un
ders
tan
din
g:
an
in
div
idu
al’s s
ub
jecti
ve ju
dg
men
t o
f
their
level o
f co
mp
reh
en
sio
n
7
CO
CO
MO
II
•C
OC
OM
O i
s t
he
mo
st
wid
ely
us
ed
, th
oro
ug
hly
do
cu
me
nte
d a
nd
ca
lib
rate
d s
oft
wa
re c
os
t m
od
el
•C
OC
OM
O -
the
“C
On
str
uc
tiv
e C
Os
tM
Od
el”
–C
OC
OM
O I
I is
th
e u
pd
ate
to
CO
CO
MO
19
81
–o
ng
oin
g r
es
ea
rch
wit
h a
nn
ua
l c
ali
bra
tio
ns
ma
de
av
ail
ab
le
•O
rig
ina
lly
de
ve
lop
ed
by
Dr.
Ba
rry
Bo
eh
m a
nd
pu
bli
sh
ed
in
19
81
bo
ok
So
ftw
are
En
gin
ee
rin
g
Ec
on
om
ics
•C
OC
OM
O I
I d
es
cri
be
d i
n S
oft
wa
re C
os
t E
sti
ma
tio
n
wit
h C
OC
OM
O I
I (P
ren
tic
e H
all
20
00
)
8
Historical Overview of COCOMO Suite of Models
COQUALMO
1998
COCOMO 81
1981
COPROMO
1998
COSoSIM
O
2004
Legend:
Model has been calibrated with historical project data and expert (Delphi) data
Model is derived from COCOMO II
Model has been calibrated with expert (Delphi) data
COCOTS
2000
COSYSMO
2002
CORADMO
1999
iDAVE
2003
COPLIM
O
2003
COPSEMO
1998
COCOMO II
2000
DBA COCOMO
2004
COINCOMO
2004
Security
Extension 2004
Costing Secure
System 2004
Software Cost Models
Software Extensions
Other Independent
Estimation Models
Dates indicate the time that the first paper was published for the model
9
Sta
te o
f th
e P
racti
ce
•Capability to m
easure systems engineering is
unavailable in current cost models
•Possible approaches
–Heuristics/rules of thumb (Honour)
–Analogy
–% of SW or HW effort (COCOMOII, PRICE-H)
–% of total effort (Honour)
–A function of complexity (Ernstoff)
•Systems Engineering is evolving
–INCOSE (est. 1992)
–Standards (EIA/ANSI632, EIA/ANSI731, ISO/IEC15288)
–Academic degrees
•We can start where COCOMO left off…
10
CO
SY
SM
O S
co
pe
•A
dd
res
se
s f
irs
t fo
ur
ph
as
es
of
the
sy
ste
m
en
gin
ee
rin
g l
ife
cy
cle
(p
er
ISO
/IE
C 1
52
88
)
•C
on
sid
ers
sta
nd
ard
Sy
ste
ms
En
gin
ee
rin
g
Wo
rk B
rea
kd
ow
n S
tru
ctu
re t
as
ks
(p
er
EIA
/AN
SI
63
2)
Conceptualize
Develop
Op
er
Test
& E
val
Tra
nsit
ion
to
Op
era
tio
n
Op
era
te,
Main
tain
,
or
En
han
ce
Rep
lace
or
Dis
man
tle
11
CO
SY
SM
O
Siz
e
Dri
vers
Eff
ort
Mu
ltip
liers
Eff
ort
Cali
bra
tio
n
# R
eq
uir
em
en
ts
# In
terf
aces
# S
cen
ari
os
# A
lgo
rith
ms
+
3 A
dju
stm
en
t
Facto
rs
-A
pp
licati
on
facto
rs
-8 f
acto
rs
-T
eam
facto
rs
-6 f
acto
rs
-S
ch
ed
ule
dri
ver
(fu
ture
)W
BS
gu
ided
by
EIA
/AN
SI 632
CO
SY
SM
O O
pera
tio
nal
Co
nc
ep
t
12
Visualization of Cost Driver Im
pacts
1.00
1.00
1.15
1.41
1.41
1.63
2.31
2.66
2.39
2.10
2.10
2.29
2.50
2.25
2.25
0.50
1.00
1.50
2.00
2.50
3.00
3.50
RELY
CPLX
RUSEDOCUBAND
PVOLSCAP
DCAPPCON
APEX
PLEX
DTEX
LIFE
SITE
SCED
System
Size
Project
Cost
Nominal
Cost
13
Where
:
PM
NS
= e
ffort
in P
ers
on M
onth
s (
Nom
inal S
chedu
le)
A=
calib
ration c
onsta
nt
derive
d f
rom
his
torical pro
ject
data
k=
{R
EQ
, IF
, A
LG
, S
CN
}
wx
=
weig
ht
for
“easy”,
“nom
inal”,
or
“difficult”
siz
e d
river
= q
uantity
of
“k”
siz
e d
river
E= represents d
isecon
om
y o
f scale
EM
= e
ffort
multip
lier
for
the j
thcost
driver.
T
he g
eom
etr
ic p
roduct
results in a
n
overa
ll eff
ort
adju
stm
ent
facto
r to
the n
om
ina
l eff
ort
.
xΦ
Model Form
∏∑
=
⋅
Φ
+Φ
+Φ
⋅=
14 1
,,
,,
,,
)(
j
j
E
k
kd
kd
kn
kn
ke
ke
NS
EM
ww
wA
PM
COSYSMO Data Sources
Army Transformation (Orlando, FL)
Integrated Data Solutions & Analysis (McLean, VA)
SAIC
National Security Solutions/ISS (San Diego, CA)
Information & Electronic Warfare Systems (Nashua, NH)
BAE Systems
Maritime Digital Systems/AIS (Pittsfield, MA)
Surveillance & Reconnaissance Systems/AIS
(Bloomington, MN)
General Dynamics
Transportation & Security Solutions (Rockville,MD)
Integrated Systems & Solutions (Valley Forge, PA)
Systems Integration (Owego, NY)
Aeronautics (Marietta, GA)
Maritime Systems & Sensors (Manassas, VA;
Baltimore, MD; Syracuse, NY)
Lockheed Martin
Mission Systems (Redondo Beach, CA)
Northrop Grumman
Intelligence & Information Systems (Garland, TX)
Raytheon
15
4 S
ize D
rive
rs
1.
Nu
mb
er
of
Sy
ste
m R
eq
uir
em
en
ts
2.
Nu
mb
er
of
Sy
ste
m I
nte
rfa
ce
s
3.
Nu
mb
er
of
Sy
ste
m S
pe
cif
ic A
lgo
rith
ms
4.
Nu
mb
er
of
Op
era
tio
na
l S
ce
na
rio
s
Weighted by complexity, volatility, and degree of reuse
16
Counting Rules Example
COSYSMO example for sky, kite, sea,
and underw
ater levels where:
Sky level: Build an SE cost model
Kite level: Adopt EIA 632 as the WBS and ISO 15288 as the life
cycle standard
Sea level: Utilize size and cost drivers, definitions, and counting
rules
Underwater level: Perform statistical analysis of data with
software tools and implement model in Excel
Source: Cockburn 2001
17
14 C
ost
Dri
vers
1.
Req
uir
em
en
ts u
nd
ers
tan
din
g
2.
Arc
hit
ectu
re u
nd
ers
tan
din
g
3.
Level o
f serv
ice r
eq
uir
em
en
ts
4.
Mig
rati
on
co
mp
lexit
y
5.
Tech
no
log
y R
isk
6.
Do
cu
men
tati
on
Matc
h t
o L
ife C
ycle
Need
s
7.
# a
nd
Div
ers
ity o
f In
sta
llati
on
s/P
latf
orm
s
8.
# o
f R
ecu
rsiv
e L
evels
in
th
e D
esig
n
Ap
plicati
on
Facto
rs (
8)
18
14 C
ost
Dri
vers
(co
nt.
)
1.
Sta
ke
ho
lde
r te
am
co
he
sio
n
2.
Pe
rso
nn
el/
tea
m c
ap
ab
ilit
y
3.
Pe
rso
nn
el
ex
pe
rie
nc
e/c
on
tin
uit
y
4.
Pro
ce
ss
ca
pa
bil
ity
5.
Mu
ltis
ite
co
ord
ina
tio
n
6.
To
ol
su
pp
ort
Team
Facto
rs (
6)
47.4
22.8
9.8
# of Operational Scenarios
18.2
6.5
3.4
# of Critical Algorithms
9.8
4.3
1.7
# of Interfaces
5.0
1.00
0.5
# of System Requirements
Difficult
Nominal
Easy
Size Driver Weights
20
Cost Driver Rating Scales
1.93
0.72
0.85
1.00
1.18
1.39
Tool support
1.93
0.72
0.80
0.90
1.00
1.18
1.39
Multisite coordination
2.16
0.68
0.77
0.88
1.00
1.21
1.47
Process capability
2.21
0.67
0.82
1.00
1.22
1.48
Personnel experience/continuity
2.31
0.65
0.81
1.00
1.22
1.50
Personnel/team capability
2.31
0.65
0.81
1.00
1.22
1.50
Stakeholder team cohesion
1.93
1.47
1.21
1.00
0.87
0.76
# of recursive levels in the design
1.87
1.87
1.52
1.23
1.00
# and diversity of installations/platform
s
1.64
1.28
1.13
1.00
0.88
0.78
Documentation
2.61
1.75
1.32
1.00
0.82
0.67
Technology Risk
1.93
1.93
1.55
1.25
1.00
Migration Complexity
2.98
1.85
1.36
1.00
0.79
0.62
Level of Service Requirements
2.52
0.65
0.81
1.00
1.28
1.64
Architecture Understanding
3.12
0.60
0.77
1.00
1.37
1.87
Requirements Understanding
EMR
Extra
High
Very
High
High
Nominal
Low
Very
Low
21
22
ISO
/IE
C 1
5288
Co
nc
ep
tua
lize
Deve
lop
Tra
ns
itio
n t
o
Op
era
tio
n
Op
era
te,
Ma
inta
in,
or
En
ha
nc
e
Re
pla
ce
or
Dis
ma
ntl
eEIA/ANSI 632
Ac
qu
isit
ion
&
Su
pp
ly
Te
ch
nic
al
Ma
na
gem
en
t
Sys
tem
Des
ign
Pro
du
ct
Rea
liza
tio
n
Te
ch
nic
al
Eva
lua
tio
n
Op
era
tio
na
l
Te
st
&
Eva
lua
tio
n
Effort Profiling
23
Lim
ita
tio
ns o
f th
e m
od
el
1.
Mo
stl
y q
ua
lita
tiv
e d
riv
ers
2.
Va
ria
nc
e o
f D
elp
hi
res
po
ns
es
3.
Sm
all
sa
mp
le s
ize
4.
Ae
ros
pa
ce
-he
av
y
5.
Ca
lib
rati
on
is
bia
se
d b
y s
uc
ce
ss
ful
pro
jec
ts b
ec
au
se
su
cc
es
sfu
l p
roje
cts
s
ha
re d
ata
, b
ad
on
es
do
n’t
6.
Mo
de
l w
ill
no
t w
ork
ou
tsid
e o
f c
ali
bra
ted
ra
ng
e
7.
A f
oo
l w
ith
a t
oo
l is
sti
ll a
fo
ol
24