Estimating Labour Distribution EEE493 2001 References:HvV 7'4

1
Estimating Labour Distribution EEE493
2001ReferencesHvV 7.4
Royal Military College of Canada Electrical and
Computer Engineering

• Major Greg Phillips
• greg.phillips_at_rmc.ca
• 1-613-541-6000 ext. 6190

Major Ron Smith smith-r_at_rmc.ca 1-613-541-6000
ext. 6030
2
Teaching Points

• COCMO II Development Time (TDEV)
• Other Models
• Typical Staffing Profiles
• Brooks Law
• COCOMO II Exercise (continued)

3
COCOMO II - Development Time

• recall, to determine actual calendar development
  time requires a translation from effort in
  person-months to development time in calendar
  months
• TDEV c x (PMNS)d
• where c 3.67
• d 0.28 0.2 x b - 1.01
• example if PMNS 100 person-months
• and b 1.15,
• then
• TDEV 3.67 x (100)0.308 15 calendar months

4
Schedule Realities

• in the earlier example a 100 person-month effort
  translates into 15 calendar months
• why not
• 100 people working for 1 month each, or
• 20 people working 5 months, or
• 1 person working for 100 months ?
• are these valid alternatives?

5
Other Development Time Models

• while parametric models may vary wildly on
  equation form and estimates of effort, note how
  consistent they are with respect to basic form
  and translation of effort to development time
• Watson-Felix - T 2.5 E 0.35
• Putnam - T 2.4 E 1/3
• COCOMO.81 - T 2.5 E 0.38
• COCOMO II - T 3.0 E 0.330.2(b-1.01)

6
Staffing Profiles

• effort and development time provide estimates in
  support of total costs and total schedule
  duration, but they say nothing of the actual
  labour distribution
• early concept phases may require small
  percentages of total effort yet may require a
  significant percentage of overall schedule
• certain tasks are difficult to achieve with high
  degrees of concurrent activities
• implementation activities may often represent two
  thirds of the total effort, but are ongoing for
  less than half the overall schedule time
• concurrent activities are the norm

7
Typical Staffing Profiles (1)

• profile of a conventional (Waterfall) project

8
Typical Staffing Profiles (2)

• default distributions of effort and schedule for
  a Unified (iterative) project

65
20
Effort
10
5
50
30
Schedule
10
10
9
Brooks Law

• adding manpower to a late project only makes it
  later
• in order to decrease the development time
  (schedule) of a software project you must
  increase the number of people
• the total manpower increases and you need more
  people sooner
• increasing the number of people on a project
  comes with an additional cost
• communication overhead increases
• existing team members productivity decreases
  initially to assist new team members

10
Optimum Labour Distribution (1)

• more formally, productivity can be derived as
  follows
• average productivity , P? P - L(N-1)? (0 lt
  ? lt 1)
• and total productivity, Ptot N x P?
• where
• P - maximum individual productivity
• N - number of members on the team
• L - the loss associated with each
  communication link
• ? - a measure of the number of communication
  links

11
Optimum Labour Distribution (2)

• example
• given a team of 11 people with average
  productivity of 10 task points per iteration
• assume a productivity loss of 10 per
  communication link and 80 interaction among team
  members
• average productivity , P? P - L(N-1)? (0 lt
  ? lt 1)
• 10 - 1(11-1)0.80
• 3.69
• and total productivity, Ptot N x P?
• 11 (3.69)
• 40 task points per iteration

12
COCOMO II Exercise - Conclusion

• recall that the total effort for PIT was 65
  person-months (based upon a nominal schedule)
• now determine the nominal development time
• TDEV c x (PMNS)d
• where c 3.67
• d 0.28 0.2 x b - 1.01
• and recall b 1.01 0.01(11343) 1.13
• therefore
• TDEV 3.67 (66.4)0.304
• 13.1 calendar months

13
exercise continued (1)

• but your boss wants it in 6 months!
• no problem just add people, right?
• apply maximum schedule compression
• will you now satisfy yours bosses demand?
• how much more (roughly) will this cost?
• assuming max schedule compression (75 of
  nominal)
• TDEV c x (PMNS)d x SCED/100
• 13.1 (.75)
• 9.8 calendar months
• you are part way there! but at what cost?

14
exercise continued (2)

• go back to the original effort equation, but this
  time take into account the fact that it is not a
  nominal schedule
• PMNS 2.94 x 14.9(1.13) x ? EMi
• but now ? EMi ACAP x APEX x PCON x SITE x
  SCED
• 1.067 (1.29) 1.376
• therefore PM 2.94 x 14.9(1.13) x 1.376
• 85.6 person-months
• a cost increase in total effort of almost 30

15
exercise continued (3)

• if we investigate relaxing the schedule to the
  maximum of 160 of nominal, we get
• TDEV c x (PMNS)d x SCED/100
• 13.1 (1.60)
• 21.0 calendar months
• PMNS 2.94 x 14.9(1.13) x ? EMi
• with ? EMi ACAP x APEX x PCON x SITE x SCED
• 1.067 (1.00) 1.067
• finally PM 66.4 person-months
• relaxing the schedule did not save any effort??

no change
16
Supplemental References

• Boehm, Barry, et al., Software Cost Estimation
  with COCOMO II , Prentice-Hall, 2000. ISBN
  0-13-026692-2.
• Royce, Walker, Software Project Management - A
  Unified Framework, Addison-Wesley, 1998. ISBN
  0-201-30958-0

17
Next ClassPeople Management