Digital Signal Processing Reference
In-Depth Information
3.2
Characteristics of the Development Process
It may be tempting to focus exclusively on the technical specifications of HEP
applications as the primary source of complexity in HEP system design. However,
it is important not to overlook the challenges in the development process itself,
which may also be a significant factor in the development time, cost, quality,
and maintainability of a system. For example, during the initial design of the
CMS experiment, the development and testing of firmware for programmable logic
devices proved to be a major bottleneck. These development challenges stem from
several distinguishing features of the HEP system design process.
Wide-Scale Collaboration and Disparate Design Styles
Most HEP systems are too large to be created by a single institution. For example,
the design of the CMS Trigger system was the product of the CMS Collaboration,
made up of over 3,000 scientists and engineers from nearly 200 universities spread
across 38 countries. Although this collaboration supplies the great breadth and
diversity of knowledge needed for HEP engineering projects, managing large groups
across multiple languages and time zones is a formidable task. The existence of
numerous, separate design teams may also give rise to multiple design approaches,
some of which may not be compatible. One of the consequences of the diversity of
the collaborative development teams and the large scale of the system design is that
the CMS Trigger subsystems have been designed by many groups from different
organizations and different countries. Because the firmware designers come from a
variety of backgrounds, it is natural that they produce a variety of disparate design
styles using their favored hardware design languages and methodologies. A famous
example of the problems that can arise from these differences is the loss of the
measurements would be made in Imperial units whereas another group assumed
they would be made in metric units. Because of these differing assumptions, the
probe's thrusters fired at the wrong time, and it crashed through the atmosphere.
Frequent Adjustments and Alterations
Despite significant simulation and theoretical modeling, the experimental nature of
HEP and the unprecedented energy of the LHC mean that it is inevitable that some
aspects of the system and the algorithms it implements will need to be adjusted
after the system comes online and starts collecting real collision data. It is not
cost-effective to replace hundreds of devices each time an algorithm is changed,
so designers must anticipate which portions of the system may require adjustment
after deployment and design those sections with reconfigurable logic to provide the
needed flexibility.
