[90] in 6.033 discussion
project cost
Saltzer@ATHENA.MIT.EDU (Saltzer@ATHENA.MIT.EDU)
Wed Mar 20 11:42:02 1996
At 10:37 AM 3/20/96, Jonathan Pfautz wrote:
>Cost of the system is not explicitly mentioned in the design project
>handout (but it is implicitly).
>
>Cost ==> dollars or man-hours
>
>Should we consider it seriously? There's always the solution of
>buying a whole lot of memory and disk space and storing everything
>ever accessed there (ignoring staleness).
>
>I guess I'm just a little nervous about significance of the cost
>tradeoff issue: how much money are we saving by not using NEARnet
>vs. how much money will it cost to implement a caching scheme.
Jonathan,
We have intentionally formulated the project in a way that is typical of
many real-world projects.
Since the motivation for adding a cache is to keep a proposed but
unspecified fee from busting the budget, the cost of the solution must be
important. The decision to proceed with implementation of your design will
be made later, when and if the interconnection supplier actually imposes
the fee and the cost and potential savings from your design are known. At
this point all that you as designer can really do is parameterize the
tradeoff by making a list of the things whose cost will have to be included
in that later analysis.
The ideal approach (admittedly quite hard to do without more information
than is easily available) would be to present the reader with a tradeoff
curve showing how your design reduces traffic for any proposed level of
investment. The decision maker could then compare that tradeoff curve with
the new service fee schedule and make a sensible decision.
Jerry
