CMU CS 15-675 Architectures for Software Systems Spring 1998
Pipe & Filter Case Study
with answers
Garlan & Kazman Questions on Readings for Lecture
7 Due: Wed Feb 2, 1998The papers:
[SG96]:Chapter 3, Section 2.
Hints:
The purpose of the readings is to illustrate the notion of a product-line architecture, and to show how the Pipe & Filter Style can be applied in an industrial context. It also focuses on the need to specialize general, pure styles to handle special requirements of the product domain.
Questions:
1. What problem frame (if any) does the oscilloscope fit into? Explain briefly why.
Generalized JSP. The system is supposed to transform input streams to output streams, but it doesn’t satisfy the JSP conditions for structured input/output and their relation. So his method doesn’t apply.
2. Why was a layered system rejected for the architecture of the system?
Overall function of the system could not be naturally segmented into opaque layers of abstraction. For example, user interface must have access to acquisition setting.
3. The architecture adopted by the oscilloscope designers departs from the general model of Pipe & Filter ways in at least two important ways. What were these?
(a) Provided special inputs so user could configure the filter
(b) Richer vocabulary of pipes: colored pipes.
4. Why was it felt necessary to introduce a richer vocabulary of pipes (i.e., "colored pipes") than is usually associated with a Pipe & Filter system?
To get acceptable performance. For example, colored pipes allowed oscilloscope to avoid copying data, and avoided the problem of a slow filter holding back its upstream producer from supplying data to other filters
Modified: 5/18/98