PDQ: Pretty Damn Quick

(7.0.0-beta Not ready for prime time)

PDQ is an analytic solver library for queueing-network models (QNM) of computer systems, manufacturing systems, and data networks, that can be written optionally in a variety of conventional programming languages (see below). Perl is the language used in the book Analyzing Computer System Performance with Perl::PDQ, which explains the fundamental queue-theoretic concepts with example PDQ models for computer performance analysis.

Overview: What is PDQ?

Languages: C, Perl, Python, R

Platforms: Linux, OS X, Windows

Maintainers: Neil Gunther and Paul Puglia

Contributors: Denny Chen, Phil Feller, Neil Gunther, Peter Harding, Paul Puglia, Sam Zallocco

License: PDQ is distributed as OSS under the MIT license.

Synopsis: PDQ library functions

Examples: See the examples/ directory.

Here is a simple PDQ model of Help Center serving a single waiting-line of customers, written in C:

#include <stdio.h>
#include "PDQ_Lib.h"

int main(void) {
    int    agents  = 4;    // available to take calls
    double aRate   = 0.35; // customers per minute
    double sTime   = 10.0; // minutes per customer

    PDQ_Init("Call Center ");
    PDQ_CreateMultiNode(agents, "Agents", CEN, FCFS);
    PDQ_CreateOpen("Customers", aRate);
    PDQ_SetDemand("Agents", "Customers", sTime);
    PDQ_SetWUnit("Customers");
    PDQ_SetTUnit("Minute");
    PDQ_Solve(CANON);
    PDQ_Report();
}

How long can a customer calling a Help Center expect to wait for a teller? Predicted performance metrics that result from solving a PDQ model can easily be displayed using the generic Report() function:

                        PRETTY DAMN QUICK REPORT         
               ==========================================
               ***  on   Tue May 17 13:23:24 2016     ***
               ***  for  Call Center                  ***
               ***  PDQ  Version 7.x.x Build 051116   ***
               ==========================================

               ==========================================
               ********    PDQ Model INPUTS      ********
               ==========================================

WORKLOAD Parameters:

Node Sched Resource   Workload   Class     Demand
---- ----- --------   --------   -----     ------
  4  MSQ   Agents    Customers  Open     10.0000

Queueing Circuit Totals
Streams:   1
Nodes:     1

Arrivals       per Minute     Demand 
--------       --------     -------
Customers      0.3500       10.0000


               ==========================================
               ********   PDQ Model OUTPUTS      ********
               ==========================================

Solution Method: CANON

               ********   SYSTEM Performance     ********

Metric                     Value    Unit
------                     -----    ----
Workload: "Customers"
Number in system          8.6650    Customers
Mean throughput           0.3500    Customers/Minute
Response time            24.7572    Minute
Stretch factor            2.4757

Bounds Analysis:
Max throughput            0.4000    Customers/Minute
Min response             10.0000    Minute


               ********   RESOURCE Performance   ********

Metric          Resource     Work              Value   Unit
------          --------     ----              -----   ----
Capacity        Agents      Customers             4   Servers
Throughput      Agents      Customers        0.3500   Customers/Minute
In service      Agents      Customers        3.5000   Customers
Utilization     Agents      Customers       87.5000   Percent
Queue length    Agents      Customers        8.6650   Customers
Waiting line    Agents      Customers        5.1650   Customers
Waiting time    Agents      Customers       14.7572   Minute
Residence time  Agents      Customers       24.7572   Minute

PDQ predicts the average customer waiting time will be 14.7572 minutes.

Alternatively, customized reports can be created using specific performance metrics like, GetUtilization() or GetThruput().