Corrigenda for Perl::PDQ book

Last updated Feb 12, 2011

NOTES

1. Page 79
   Equation (2.62) should read:
   

   φ(m, ρ)     =      1 − ρ 1 − ρm     =      1 1 + ρ+ ρ2 + ... ρm−1

   
   
   
2. Page 102
   In items just prior to Section 2.10.1, the Figure and Equation callouts are messed up. They should read:
   1. An M/M/1 queue ...(Fig. 2.24) ... (2.35) ...
      
      
   2. An M/M/2 queue ...(Fig. 2.16) ... (2.63) ...
      
      
   3. A multiqueue (Fig. 2.14) ... (2.44) ...
      
      
   
   
   
3. Page 115
   Cite M/M/1//N →  machine repairman model on pp. 88, 238, 412, as forward reference to Amdahl's law in Chap. 6 of GCaP book.
   Similarly, add and cite M/G/1//N →  load-dependent repairman as forward reference to USL in Chap. 6 of GCaP book.
   
   
4. Page 140
   MVA Approximation. The independent Perl subroutine on p. 140 is hosed. The following code works correctly and will be included with the next PDQ release, in the Perl subdirectory (not the PDQ subdirectory). [NOTE: Since these subroutines are not used by PDQ, the APPROX solution method for pdq::Solve() works as advertized.]

   sub approx
   {
      # Schweitzer algorithm
      for ($k = 1; $k <= $K; $k++) {
         $Q[$k] = $N / $K; # equi-partition queues initially
      }
   
      do {
         for ($k = 1; $k <= $K; $k++) {
            $Qprev[$k] = $Q[$k];
            $R[$k] = $D[$k] * (1.0 + $Q[$k] * ($N - 1) / $N);
         }
         
         $rtt = $Z;
         for ($k = 1; $k <= $K; $k++) {
            $rtt += $R[$k];
         }
         $X = $N / $rtt;
   
         for ($k = 1; $k <= $K; $k++) {
            $Q[$k] = $X * $R[$k];
         }
      }  while (converging())
   }
   
   sub converging {
       my $k, $threshold = 0.001; # stopping criterion
      
       # Equivalent of Max_i[|(Qnew_i - Qold_i)/Qnew_i|] pseudo-code 
       # seen in various textbooks
       for ($k = 1; $k <= $K; $k++) {
           $delta = abs($Q[$k] - $Qprev[$k]) / $Q[$k];
           return(1) if($delta > $threshold);
       }
       
       return(0); # must've converged below threshold
   }
   
   

   
   
   
5. Page 345
   Recalculated numbers using R:

   From equation (10.2):
      BW = 512/3.57e-3                   =  143417.37 B/s
   From Table 10.1:
   1. Tsend  (536 * 2 + 74)/BW           =       7.99 ms
   2. RTT from eqn.(10.1)                =      73.99 ms
   3. Tserv from eqn.(10.3)              =      26.13 ms
   4. Trecv (512 * 3 + 316 + 1 + 1)/BW   =      12.93 ms
   Performance improvement:
      Old total response time            =     599.04 ms
      New total response time            =     121.04 ms
      Percentage change                  =      79.79 %
   
   

   
   
   
6. Page 365
   Dummy Queues. Chapter 10, code fragment at bottom of p. 365 should read:

       use constant MAXDUMMIES => 12;
       ...
       $think    = 0.0 * 1e-3;  # free parameter
       $dtime    = 2.2 * 1e-3;  # dummy service time 
       ...
       # Create dummy nodes with their service times ... 
       for ($i = 0; $i < MAXDUMMIES; $i++) {
           $dnode = "Dummy" . ($i < 10 ? "0$i" : "$i"); # 2-digit number
           $pdq::nodes = pdq::CreateNode($dnode, $pdq::CEN, $pdq::FCFS);
           pdq::SetDemand($dnode, $work, $dtime);
       }
   
   

   
   
   
7. Page 386
   Presper Eckert was interviewed in 1989 about the development of the ENIAC computer. The interview was published in ComputerWorld on the occassion of the ENIAC's 60-th birthday. He has the following to say about von Neumann's contribution:

   He (von Neumann) came and looked at our stuff and went back to Princeton and wrote a long document about the principles. He gets a lot of credit but the inventions were ours. Someday I'll write a book on who really invented the computer. It wasn't Atanasoff or Von Neumann. We did it.

   It looks to me like Eckert and Mauchly (being engineers) did not fully appreciate what von Neumann did, and vice versa.

   > Comments: On page 386 you state:
   > 
   > "The IBM 701 was designed exclusively for business data processing. It was a
   > vacuum tube computer programmed with punch cards."
   > 
   > My recollection is that there were two IBM computer lines, the 701 - 704 - 709
   > - 7090 - 7094 line of *scientific* computers, and the 702 - 705 - 7080 line of
   > *commercial* computers. The scientific line had a 36 bit binary word-addressed
   > architecture and the commercial line had a character-based architecture. I
   > don't know a whole lot about the commercial line, but I did earn a living
   > programming 7094s early in my career.
   > 
   > All were true Von Neumann machines, with programs and data stored in the same
   > random access memory. They were not "programmed with punch cards" but did have
   > card readers. The 701 and 702 had Williams tube (CRT-based) RAM, IIRC, and
   > succeeding members of the lines had magnetic core RAM. The 7080 and 7090 and
   > later had solid-state logic; predecessors had vacuum tube logic.
   
   

   
   
   
8. Page 403 Tables. Incorrect values for F_n caused by the old Excel gotcha of cell coordinates auto-incrementing in a copy/paste operation.
   Table 1: Corrected Table D.2

   n	Data	Ranked	Emp	Thy	Fn	Fn−1	D+	D−
   1	11.72	1.43	0.1220	0.1797	0.1667	0.0000	-0.0130	0.1797
   2	10.43	4.12	0.3515	0.4348	0.3333	0.1667	-0.1015	0.2682
   3	8.02	7.58	0.6468	0.6500	0.5000	0.3333	-0.1500	0.3167
   4	7.58	8.02	0.6843	0.6707	0.6667	0.5000	-0.0040	0.1707
   5	1.43	10.43	0.8899	0.7642	0.8333	0.6667	0.0692	0.0975
   6	4.12	11.72	1.0000	0.8027	1.0000	0.8333	0.1973	-0.0306

   Table 2: Corrected Table D.3

   N	6
   Dmax	0.3167
   Kcrit	0.7757
   p-value	0.7632
   α	0.2368