Original Author/link
Originally Described by Adam Rifkin (1994). No weblink to independent description available. However, (Sivilotti2003) employed the exercise in a summer workshop for middle school girls, and provided a writeup of the activity and slides associated with the activity at his website
See papers (Rifkin1994, Sivilotti2003) for additional details.
Similar Exercises:
Sorting: oddEvenTranspositionSort, sortingNetwork, cardsorting
Other activities by (Rifki1994, Sivilotti2003)
Rifkin1994: oddEvenTranspositionSort
Details
(Rifkin1994) designed the activity to take place in the course of the hour, and was introduced to high school minority students in a workshop format. Prior to the exercise, students are given an overview of the concept of sorting, and try to come up with ways to sort a collection of six random numbers.
The instructor then splits the group into nine teams (of about the same size), with three teams assigned to a separate supervisor. The supervisor will then have one of the three teams perform bubble sort, the second perform oddeventransposition sort (described in a separate exercise ), and the last perform parallel radix sort.
Each team under the care of supervisor stands in a separate row. Each row has indices noted on the ground, indicating where students should stand, numbered from 0 … n1 where n is the number of students in each row. Each student at index i is handed a random number that corresponds to the same index in a separate row. Each row is then given its own set of instructions, and they are instructed to follow the instructions specifically (without talking) to sort the numbers.
For parallel radix sort, the exercise is formed as a “race”, where the group leader walks down the line with the student at space 0.
 The student at space 0 (or “roving student”), as they walk, reveals their number to the other students standing in place.
 Students standing in place (“stationary students”) keep track of a private counter (which starts at 0). Whenever a stationary student sees a number walk pass that is less than their number, they increment their counter by one.
 As the roving student passes by each stationary student, the stationary student “acknowledges” the fact that they’ve seen the number by raising their hand, enabling the roving student to go to the next spot.
 Once the roving student reaches the end of line, he or she returns to space 0. The student at space 1 then becomes the roving student, while all other students remain stationary. Starting at space 0, they walk down the line, revealing their numbers to each stationary student, who raises their hand in acknowledgment, and updates their private counters as needed.
 The process continues until every student had an opportunity to walk down the line.
 Once everyone had the opportunity to walk down the line, the group leader indicates that everyone should simultaneously leave the space they were standing and go to the same space indicated by their private counter. After doing this, the numbers should be now be sorted.
The smallest number will have had a private counter of 0, and will stand in space 0, as they never encountered a number that was smaller. The next smallest number will be at space 1, and so forth.
Students are asked to think why the list should be sorted at the end of Parallel Radix Sort.
Variants (Sivilotti2003)
Sivilotti performed the activity exactly as described in (Rifkin1994). However, prior to introducing the activity, used an analogy to describe computers as “chefs” and programs to “recipes”. Like programs, recipes are characterized by what their ingredients (input) and the final product (output). A software engineer is thus described as a “recipeengineer”.
CS2013 Knowledge Unit Coverage
Parallel Decomposition
2. Identify opportunities to partition a serial program into independent parallel modules. [Familiarity]
TCPP Topics Coverage
TCPP algorithms

Comprehend Asymptotics: Understand upper (bigO) and lower bounds (bigOmega,); follow elementary bigO analyses, e.g., the O(log n) treedepth argument for mergesort with unbounded parallelism.

Comprehend Time: Recognize time as a fundamental computational resource that can be influenced by parallelism.

Comprehend Speedup: Recognize the use of parallelism either to solve a given problem instance faster or to solve larger instance in the same time (strong and weak scaling).

Know Sorting: Observe several sorting algorithms for varied platforms  together with analyses. Parallel merge sort is the simplest example, but equally simple alternatives for rings and meshes might be covered also; more sophisticated algorithms might be covered in more advanced courses.
Recommended Courses

K12: (Rifkin1994) successfully introduced sorting concepts to minority high school students in a summer workshop. (Sivilotti2003) successfully introduced the activity to middle school girls in a 45minute format.

CS2/DSA: TCPP recommends that topics of sorting can be covered as early as CS2, and that both sorting and algorithmic complexity topics are appropriate for DSA.
Accessibility
The exercise may be difficult for students who are mobilityimpaired, or have trouble with visual. Cards can have Braille on them, so when a blind student plays the stationary student role, he or she can reach out and touch the card to read the value. Verbal acknowledgments may be most appropriate when the blind student acts as the roving student in the exercise. Verbal acknowledgments may also be preferable to students who have trouble raising their hands.
Assessment
(Rifkin1994) gives a general overview of assessment; 81 students completed surveys, in which they overwhelmingly responded that the workshop as a whole was a positive experience, and that respondents “learned something, had fun, and had a better impression of computer science” (Rifkin1994). However, specific assessment of parallel radix sort was not provided.
(Sivilotti2003) reported the sorting activity could be completed in a 45 minute period. Students were asked a number of questions, including rating the exercises. On a scale from 1 (low) to 5 (high) students rated the parallel sorting exercise as a 4.4. Students also made statements such as “the more ‘chefs’[processors] you use, the faster the program will finish” and “sequential are slow” (Sivilotti1999). Students were also asked to evaluate the overall quality of the computer science module on scale of 1 (low) to 5 (high), and rated the cs component of the workshop as 4.6. Note that this includes other activities.
Citations

P. A. G. Sivilotti and M. Demirbas, “Introducing middle school girls to fault tolerant computing”, in Proceedings of the 34th SIGCSE Technical Symposium on Computer Science Education (SIGCSE ’03). New York, NY, USA: ACM, 2003, pp. 327–331, [Online]: http://web.cse.ohiostate.edu/~sivilotti.1/outreach/FESC02/ Available: http://doi.acm.org/10.1145/611892.611999

A. Rifkin, “Teaching parallel programming and software engineering concepts to high school students”, In SIGCSE Bulletin, vol. 26, no. 1. and Proceedings of the Twentyfifth SIGCSE Symposium on Computer Science Education, 1994, pp. 26–30. Available: http://doi.acm.org/10.1145/191033.191044