tmcclory@emdeng.Dayton.NCR.COM (Thomas.J.Tom.McClory) (10/05/88)
As a newly minted manager, I've suddenly developed an interest in effort estimating and project sizing techniqes. What I need are techiques that can be used to produce "accurate" size and error estimates early in the software life-cycle (usually with only a vague idea of what the requirements are), and for predicting the risk of estimate error (i.e. Mr. General Manager, there is a 50% likelyhood the optimistic assumptions made in this estimate will actually occur). "Accurate" means making delivery within 20% either side of the estimated date. Some of the methods I've read about, but not used are: COCOMO Barry Boehm SLIM Larry Putnam SPQR-20 Capers Jones I'm sure there are others. How well do any of these techniques work? My shop primarily produces systems software to be sold to (hopefully) many commercial customers. Development cycles are always "aggressive". Quality is always a concern. A model that shows the trade-offs between schedule/effort/quality/complexity would really be nice. Since I'm experimenting, I don't want to spend gobs of money buying some black box model, unless I see rave reviews from netland. The COCOMO model is public domain, has anyone got a public domain implementation I could get a hold of? Are there any other public domain models I can try? Any good references on sizing and project estimating I should read? There has to be a better way than finger to the wind estimates followed by prayer! ---------------------------------------------------------------------- Tom McClory | UUCP: tmcclory@emdeng.dayton.ncr.com NCR Corporation EMD-4 | Engineering & Manufacturing Dayton | Dayton, OH. 45479 | -- Tom McClory | UUCP: tmcclory@emdeng.dayton.ncr.com NCR Corporation EMD-4 | or ...gatech!hubcap!ncrcae!ncrlnk!emdeng Engineering & Manufacturing Dayton | Dayton, OH. 45479 |
spaf@cs.purdue.edu (Gene Spafford) (10/06/88)
Funny you should ask that (about size/cost estimates). It turns out
that many (most?) published methods don't fare too well when attempts
are made to validate them against real data, that is, programs not
written as student projects.
One recent exception is the COPSTAR/COPMO estimator developed by Sam
Conte & M. Rathi as part of a SERC project on metrics. It has been
validated against a few million lines of code of real projects from 6
different companies and in 3 or 4 different languages (including ADA).
You can get copies of the public tech reports by writing to
northern@cs.purdue.edu and asking for them. Employees of SERC
affiliate companies can get copies of all the literature, public or
not, plus a copy of the program that implements the metrics...as well
as info on the other 15 or so software engineering projects currently
underway in the Center.
(SERC is the Software Engineering Research Center, an
NSF-sponsored, University-Industry Cooperative Research
Center, jointly located at Purdue University and the
University of Florida.)
(If your company isn't an affiliate or you don't know if
it is an affiliate, write to me -- I don't want to make this
sound more like a commercial than it may already.)
--
Gene Spafford
NSF/Purdue/U of Florida Software Engineering Research Center,
Dept. of Computer Sciences, Purdue University, W. Lafayette IN 47907-2004
Internet: spaf@cs.purdue.edu uucp: ...!{decwrl,gatech,ucbvax}!purdue!spaf