Errata for The Practical Performance Analyst

Errata for The Practical Performance Analyst

In order to view any mathematical notations correctly, you may need to check here before continuing.
Where Who What
p.12 para 4 W. A. Gunther Elide "of" in "... We shall study (of) this property of ..."
p.25 W. Pelz and students Corrected [Eqn.(1-10) on p.25] appears at the end of this table.
p.26  Author, and P. Jakubik Eqn.(1-15) s.b. MTBF = exp(−lnλ) = 1/λ
p.26  C. Millsap Section: Exponential distribution. Final sentence s.b. "to construct exponential based reliability models."
p.27 after Eq.(1-19) W. A. Gunther Elide "of the " and replace "are" with "is" in "... where each of the Ak is calculated ..."
p.36 line 7 W. A. Gunther Insert "results" in "...most powerful results in queueing theory..."
p.51 D. Tran Example 2-1. Eq 2.6 s.b. Eq 2.16
p.51 W. A. Gunther System Residency. 3rd bullet should be "X be the throughput"
p.52 line 7 T. Lange s.b. Eqn. (2-18)
p.56 Line 1, D. Tran "...service time (2 s)..." s.b. "service time (1 s)"
p.58 W. A. Gunther Last line prior to Parallel centers: Elide the first "in".
p.59 W. Pelz and students Line 6th from bottom: "(1-29)" s.b. "(2-29)"
p.64 R. Lane Para 2. MM/m/∞ s.b. M/M/m//∞
p.67 W. Pelz and students line 3: "(2-6)" s.b. "(2-16)"
p.68 W. Pelz and students Example 2-6. Time unit s.b. milliseconds throughout.
p.69 line 18 K. Christensen Text: "/* Define service demand ..." is missing terminating characters:  "*/" (close of C comment)
p.72 Eq.(2-51) P. Sinclair Corrected [Eqn on p.72] appears at the end of this table.
p.72 W. Pelz and students Example 2-7: (2-6) s.b. (2-42)
p.77 L. Seshadri Line 5: Equation L = Q − m s. b. L = Q − m ρ
p.89 Para under Fig.3-6 M. Allen Subscripts misaligned. For example: exp(λ1t) exp(λ2t)exp(λ3t) = exp([λ1 + λ2 + λ3]t) and at the end of the next para: exp(λt) = exp([p1 + p2 + p3]λt).
p.95 J. A. Nolazco-Flores Service times: 20, 600, 300, and 60 seconds missing from Fig.3-12.
p.97 I.S. Hobbs Truncated C code. Corrected [C code for p.97] appears at the end of this table.
p.99 H. Schwetman, D. Molero MVA code is wrong. Corrected [MVA code in C for p.99] appears at the end of this table.
p.101 W. Pelz and students Line 5th from bottom: "(3-1)" should be "(3-10)"
p.104 First bullet W. A. Gunther "The only thing that matters ..." should read "The only things that matter are the workload intensities..."
p.110 2nd para. I. S. Hobbs "..then Da would be become the new bottleneck." Delete one of be/become.
p.111 W. Pelz and students Eqn (3-25): the k in front of the D should be elided
p.115 line 12th from last A. Cockcroft Truncated code line in printf("Q at FESC: %3.2f\n\n", qlength);
p.117 Table 3-2 I. S. Hobbs Last row. Bdsk Aggregate should be: 54.0 + 428.4 = 482.4 Sec.
p.120 para after C code Author Second reference to figures: Fig. 3-25 s.b. 3-26
p.120 para after C code C. Millsap "20 percent" s.b. "200 percent" (actually 153.4% from Table 3-4)
p.120 line 9 from bottom W. Pelz and students the second Xc in the formula should be Xo
p125 para 2, line 2 A. Cockcroft "shear" should be "sheer"
p125 2nd last line W. A. Gunther Insert "of" in "some of the more prominent architectures ..."
p129 last section A. Cockcroft Section: MIB naming tree: "enterpise" s.b. "enterprise"
p.139 W. A. Gunther Section: Workload monitoring mid para: "counter" should be plural in "... the value of certain counters and gauges."
p142 Fig 4-5 A. Cockcroft umasetthesh s.b. umasetthreshold and umareqconfig s.b. umarequestconfig
p144 A. Cockcroft Example uses older version of UMA data, not final spec.
p.151 under Table 5-2 B. Myers In 2nd line: Fig. 5-4 s.b. Fig. 5-3
p.154 para 2 W. A. Gunther "On SMPs is it ..." should read "On SMPs it is ..."
p176 first item, last line A. Cockcroft insert "bound" in "... upper bound on network latency."
p176 table 6-1 last col A. Cockcroft "SC1000" s.b. "SC2000"
p.191 V. Davis Eqn.(6-25): λ in numerator s.b. σ.
p.191 W. Pelz and students Concurrency item: s.b. "If there were no interaction"
p.199 Table 6-2 W. Pelz and students Order of the headings Hyperbolic and PDQ model s.b. reversed.
pp. 207, 208, 412 and p. xxxi A. Cockcroft Cockroft s.b. Cockcroft
p. 256 PDQ code. B. Vestermark The 'request/reply' service pair needs to be added in a single PDQ_SetDemand() call for each transaction type.
p.270 End of second para. K. Christensen "HyperText Transport Protocol" s.b. "HyperText Transfer Protocol"
p.275 Eq(9-2) Author Should read: 143.42 x 103 B/s. And the line immediately after should read "... throughput of about 1.15 Mbps."
p.277 para 3 W. A. Gunther "daemon" should be "demon" for consistency. Same for: p. 278 Fig 9-2; p. 280 para 2 and 3; p. 281 para 1; p. 282 para 1
p280 A. Cockcroft Table 9-3 should follow end of the code at /* main */
p.311 line after Eq.(11-18) W. A. Gunther Insert commas: "... of a particle, incurred due to a force F, is proportional ..."
p.312 D. Chan Section: The action. End of first para: L = T(x) + U(x) s.b. L = T(x) − U(x)
p.327 Last para. K. Christensen "CD/CSMA" s.b. "CSMA/CD"
p.330 line 1 W. A. Gunther Replace "... network's dynamics ..." with "network dynamics".
p.338 W. A. Gunther Section: Stationary points second last line: "points in the interval (−1, +1) with the inflection point coincident with at pc; ..." Also, elide second "with".
p.369 top 2 paras L. Dantzler Corrected text for [Appendix A on p.369] appears at the end of this table.
p.384 first item A. Cockcroft Updated URL for CMG is
p.384 K. Christensen IBM GUIDE has now merged with IBM SHARE at
p.407 W. A. Gunther Section: Statistical Model: Elide "few" in "Multivariate Analysis - Many few random variables ..."
p423 A. Cockcroft packer-network s.b. packet-network


[Eqn.(1-10) on p.25]

λ(t) =
dt → 0 

 R(t) − R(t + dt)


 =   f(t)

[Eqn on p.72]

R(N) =   N

 − Z
[C code for p.97]
nodes = PDQ_CreateNode("Window1", CEN, FCFS);
nodes = PDQ_CreateNode("Window2", CEN, FCFS);
nodes = PDQ_CreateNode("Window3", CEN, FCFS);
nodes = PDQ_CreateNode("Window4", CEN, FCFS);

[MVA code in C for p.99]
NOTE: Code was always correct in PDQ.
/* MVA iteration on every node for each value of N users */

for(n = 0; n < N; n++) {

    /* response time at node k */
    for(k = 0; k < K; k++) {
        R[k] = D[k] * (1.0 + Q[k]);
    /* system throughput */
    for(k = 0; k < K; k++) {
        Z += R[k];
    Xsys = N / Z;
    /* queue-length at node k */
    for(k = 0; k < K; k++) {
        Q[k] = Xsys * R[k];
} /* loop over n */

[Appendix A on p.369]
Uses the canonical solution technique. See Chapter 2 (and the online PDQ User Manual) for more details. Only consistent in the context of a open queueing circuit. Used as an argument for the PDQ_Solve() function.
This solution technique uses the iterative MVA (Mean Value Analysis) method for up to three workload classes in versions of PDQ later than the original release 1.0. Only consistent in the context of a closed queueing circuit. Used as an argument for the PDQ_Solve() function.

File translated from TEX by TTH, version 3.38.
On 4 Jun 2012, 12:49.