Hierarchical Software Size Metrics
Petri Vesterinen
Nokia Research Center
P.O.Box 422. FIN-00045 Nokia Group. Finland
petri.vesterinenOresearch.nokia.com
Nokia Corporation
p
::!
-.
.-
n e t sales 52.6 billion FIM (nearly 10 billion USD)
t o t a l personnel 36 600
s a l e s in 130 countries
.main products:
mobile phones, communicators
digital exchanges and
telecommunications networks
monitors. satellite and cable receivers.
multimedia equipment
-7
Introduction
Software size metric is the most important software metric
.used in several planning and tracking activities
@usedin derived metrics like productivity and quality
Software size can be measured in several ways
'physical' size
Lines of Code (LOC, NLOC, ...)
Number of modules
Amount of memory used, etc.
'functional' size
FPA and variants
'logical' size
Complexity metrics
I
-.zT7:L-d<
ONfUU.
,
C ~ l U ~ Wl
~ t oa m
l 1 9 e I P . C 1 V s r r a n n p.g1 3
NaKIA
Some Problems
LOC is not suitable for
estimation where LOC is not known accurately enough
early estimation
new technology, new functionalities
.measuring productivity or quality
different languages & different coding styles
FP
i s not suitable for very early estimation where function types
can not be identified
counting may be too laborious in early phases
.may not be accurate enough in late project phases
FP & LOC: Primary Functional Areas
TIME-
* This explains why:
F P is often used in estimation (availability)
LOC based effort estimates are often more accurate
. ---.- -.A . . - - ..
,4zz
P-NOKlA
,
.
.A,--
ON(xw
C(t(Uba.nl
~
T 11 no
I ~wm~PmtVsarnp
m w5
Available Solutions (112)
Use of only one SW size metric despite of problems, e.g.
*only LOC because of huge amount of history data
automatic code counting tools available for LOCs
PROBLEMS:
selected metric is good only for some purposes
-KIA
.-
-.-.
d
.
Available Solutions (212)
Different metrics for different purposes, e.g.
O b j e c t Points, FP and LOC in COCOMO 2.0
F P : agree project size and early estimation; LOC: quality
measurements and predictor of memory consumption
PROBLEMS:
several metrics has to be taken -> increased measuring cost
d e p i c t different views and not fully comparable
hard to find reliable conversion factors (Jones's backfiring
error may be >Sooh)
What We Need (112)
We need a HIERARCHICAL software size metric that:
uses same metric units in all project phases
a v o i d s conversion between very different units
is cost-effective to measure in all phases
is implementation independet in early phases (like FP)
is implementation dependent in late phases (like LOC)
takes use of the increased understanding of the nature of the
work as the specification and design proceeds
lcan be used in derived metrics like productivity and quality
is easy to understand
is based on solid theory
What We Need (212)
Amount of
information
Information
available
/~
> TIME
4
Nokia Solution (113)
b
!
p
Software
NSM: "Nokia Sizing Method'
,
-
.-...----
I
Nokia Solution (213)
Subjective
Level
Name of Feature
Size
Architecture
independent
Size
X S ~ SMIL I X L
Architecture
dependent
X S ~ S I M I1x1
I
"Feature" Name of Task
Level
"Details" Architectural item
Level
Ned
or
Every level uses same metric units
(called here as NSP, "Nokia Sizing Point")
,
,
7
ONCXU
Objective
NOKIA
,
,
--
C O X M I I H _ O I W T / '28 R) ISME I P l b l V e # a n s l prp I 1
-
.
. - -
"
Nokia Solution (313)
I
~ ' k
XS
RELEASE
FEATURE
CALIBRATED
item weights
item weights
CALIBRATED
item weights
Available Solutions for FP
Integration
0
TIME
Early Function Points is hierarchical in the context as it uses
same FP units as IFPUG FPA
3D-FP and FFP are not hierarchical as they use their own units
P
-:.- .-.:"
NOKIA
crrcun,s-o~
a~ws~Pa~vs*m-,
ONCXU
WTlm
p l p 13
Summary
Current size metrics are suitable only for restricted use
It is hard to cope with different size units
The metric should 'grow up' or 'mature' along with SW projects.
There is really a need for a hierarchical software size metric
The Nokia solution
Preliminary results suggest that the idea really works
Calibration is the critical activity in forming the metric
A research interest in the future to form a true hierarchical size
measurement framework in the FP context
Another interesting context would be 00