Example 4.15 Resource Allocation

In the previous example, a summary of the resource utilization is obtained. Suppose that you want to schedule the project subject to constraints on the availability of ENGINEERS. The activity data, as in Example 4.14, are assumed to be in a data set named WIDGRES. The resource variable, engineer, specifies the number of engineers needed per day for each activity in the project. In addition to the resource engineer, a consumable resource engcost is computed at a daily rate of 200 for each unit of resource engineer used per day. The following DATA step uses the Activity data set from Example 4.14 to create a new Activity data set that includes the resource engcost.

data widgres;
set widgres;
if engineer ^= . then engcost = engineer * 200;
run;

Now suppose that the availability of the resource engineer and the total outlay for engcost is saved in a data set named WIDGRIN, displayed in Output 4.15.1.

In the data set WIDGRIN, the first observation indicates that engineer is a replenishable resource, while engcost is a consumable resource. The second observation indicates that an extra engineer is available, if necessary. The remaining observations indicate the availability profile starting from December 1, 2003. PROC CPM is then used to schedule the project to start on December 1, 2003, subject to the availability as specified.

proc cpm date='01dec03'd interval=weekday
         data=widgres holidata=holdata resin=widgrin
         out=widgschd resout=widgrout;
   tailnode tail;
   duration days;
   headnode head;
   holiday hol;
   resource engineer engcost / period=per obstype=otype
                               schedrule=shortdur
                               delayanalysis;
   id task;
   run;

In the first invocation of PROC CPM, the scheduling rule used for ordering the activities to be scheduled at a given time is specified to be SHORTDUR. The data set WIDGSCHD, displayed in Output 4.15.2, contains the resource constrained start and finish times in the variables S_START and S_FINISH. On December 8, three activities can be scheduled, all of which require the resource engineer. Using the scheduling rule specified, PROC CPM schedules the activities with the shortest durations first; thus, the activity 'Drawings' is delayed by five working days, until December 15, 2003.

The DELAYANALYSIS option in the RESOURCE statement helps analyze the cause of the delay by adding three new variables to the Schedule data set, R_DELAY, DELAY_R, and SUPPL_R. In this example, the R_DELAY and DELAY_R variables indicate that there is a delay of five days in the activity 'Drawings' due to the resource engineer. Such information helps to pinpoint the source of resource insufficiency, if any.

Other activities that follow 'Drawings' also have S_START>E_START, but the slippage in these activities is not caused by resource insufficiency, it is due to their predecessors being delayed. The entire project is delayed by five working days due to resource constraints (the maximum value of S_FINISH is 17MAR04, while the maximum value of E_FINISH is 10MAR04).

In this invocation, the DELAY= option is not specified; therefore, the supplementary level of resource is not used, since the primary levels of resources are found to be sufficient to schedule the project by delaying some of the activities.

The data set WIDGROUT, displayed in Output 4.15.3, contains variables Rengineer and Aengineer in addition to the variables Eengineer and Lengineer. The variable Rengineer denotes the usage of the resource engineer corresponding to the resource-constrained schedule, and Aengineer denotes the remaining level of the resource after resource allocation. For the consumable resource engcost, the variables Eengcost, Lengcost, and Rengcost indicate the rate of usage per unit routinterval (which defaults to INTERVAL=WEEKDAY, in this case) at the start of the time interval specified in the variable _TIME_. The variable Aengcost denotes the amount of money available at the beginning of the time specified in the _TIME_ variable.

proc cpm date='01dec03'd
         interval=weekday
         data=widgres
         resin=widgrin
         holidata=holdata
         out=widgsch2
         resout=widgrou2;
   tailnode tail;
   duration days;
   headnode head;
   holiday hol;
   resource engineer engcost / period=per
                               obstype=otype
                               schedrule=lst
                               delayanalysis;
   id task;
   run;