Computing the Concurrency Threshold of Sound FreeChoice Workflow Nets
Abstract
Workflow graphs extend classical flow charts with concurrent fork and join nodes. They constitute the core of business processing languages such as BPMN or UML Activity Diagrams. The activities of a workflow graph are executed by humans or machines, generically called resources. If concurrent activities cannot be executed in parallel by lack of resources, the time needed to execute the workflow increases. We study the problem of computing the minimal number of resources necessary to fully exploit the concurrency of a given workflow, and execute it as fast as possible (i.e., as fast as with unlimited resources).
We model this problem using freechoice Petri nets, which are known to be equivalent to workflow graphs. We analyze the computational complexity of two versions of the problem: computing the resource and concurrency thresholds. We use the results to design an algorithm to approximate the concurrency threshold, and evaluate it on a benchmark suite of 642 industrial examples. We show that it performs very well in practice: It always provides the exact value, and never takes more than 30 ms for any workflow, even for those with a huge number of reachable markings.
1 Introduction
A workflow graph is a classical controlflow graph (or flow chart) extended with concurrent fork and join. Workflow graphs represent the core of workflow languages such as BPMN (Business Process Model and Notation), EPC (Eventdriven Process Chain), or UML Activity Diagrams.
Following previous work, we do not directly work with workflow graphs, but with their equivalent representation as freechoice workflow Petri nets, which has been shown to be essentially the same model [10] and allows us to directly use a wealth of results of freechoice Petri nets [7]. Figure 1(a) shows a freechoice workflow net. The actual workflow activities, also called tasks, which need a resource to execute and which consume time are modeled as the places of the net: Each place p of the net is assigned a time \(\tau (p)\), depicted in blue. Intuitively, when a token arrives in p, it must execute a task that takes \(\tau (p)\) time units before it can be used to fire a transition. A free choice exists between transitions \(t_4\) and \(t_6\), which is a representation of a choice node (ifthenelse or loop condition) in the workflow.
If no choice is present or all choices are resolved, we have a deterministic workflow such as the one in Fig. 1(b). In Petri net terminology, deterministic workflows correspond to the class of marked graphs. Deterministic workflows are common in practice: in the standard suite of 642 industrial workflows that we use for experiments, \(63.7\%\) are deterministic. We show that already for this restricted class, deciding if the threshold exceeds a given bound is NPhard. Therefore, we investigate an overapproximation of the resource threshold, already introduced in [4]: the concurrency threshold. This is the maximal number of task places that can be simultaneously marked at a reachable marking. Clearly, if a workflow with concurrency threshold k is executed with k resources, then we can always start the task of a place immediately after a token arrives, and this schedule already achieves the fastest runtime achievable with unlimited resources. We show that the concurrency threshold can be computed in polynomial time for deterministic workflows.
For workflows with nondeterministic choice, corresponding to freechoice nets, we show that computing the concurrency threshold of freechoice workflow nets is NPhard, solving a problem left open in [4]. We even prove that the problem remains NPhard for sound freechoice workflows. Soundness is the dominant behavioral correctness notion for workflows, which rules out basic controlflow errors such as deadlocks. NPhardness in the sound case is remarkable, because many analysis problems that have high complexity in the unsound case can be solved in polynomial time in the sound case (see e.g. [1, 7, 8]).
After our complexity analysis, we design an algorithm to compute bounds on the concurrency threshold using a combination of linear optimization and statespace exploration. We evaluate it on a benchmark suite of 642 sound freechoice workflow nets from an industrial source (IBM) [9]. The bounds can be computed in a total of 7 s (over all 642 nets). On the contrary, the computation of the exact value by statespace exploration techniques times out for the three largest nets, and takes 7 min for the rest. (Observe that partialorder reduction techniques cannot be used, because one may then miss the interleaving realizing the concurrency threshold.)
The paper is structured as follows. Section 2 contains preliminaries. Sections 3 and 4 study the resource and concurrency thresholds, respectively. Section 5 presents our algorithms for computing the concurrency bound, and experimental results. Finally, Sect. 6 contains conclusions.
2 Preliminaries
Petri Nets. A Petri net N is a tuple (P, T, F) where P is a finite set of places, T is a finite set of transitions (\(P \cap T = \emptyset \)), and \(F \subseteq (P \times T) \cup (T \times P)\) is a set of arcs. The preset of \(x \in P \cup T\) is Open image in new window and its postset is Open image in new window . We extend the definition of presets and postsets to sets of places and transitions \(X \subseteq P \cup T\) by Open image in new window and Open image in new window . A net is acyclic if the relation \(F^*\) is a partial order, denoted by \(\preceq \) and called the causal order. A node x of an acyclic net is causally maximal if no node y satisfies \(x \prec y\).
A marking of a Petri net is a function Open image in new window , representing the number of tokens in each place. For a set of places \(S \subseteq P\), we define Open image in new window . Further, for a set of places \(S \subseteq P\), we define by \(M_S\) the marking with \(M_S(p) = 1\) for \(p \in S\) and \(M_S(p) = 0\) for \(p \notin S\).
A transition t is enabled at a marking M if for all Open image in new window , we have \(M(p) \ge 1\). If t is enabled at M, it may occur, leading to a marking \(M'\) obtained by removing one token from each place of Open image in new window and then adding one token to each place of Open image in new window . We denote this by Open image in new window . Let \(\sigma = t_1 t_2 \ldots t_n\) be a sequence of transitions. For a marking \(M_0\), \(\sigma \) is an occurrence sequence if Open image in new window for some markings \(M_1,\ldots ,M_n\). We say that \(M_n\) is reachable from \(M_0\) by \(\sigma \) and denote this by Open image in new window . The set of all markings reachable from M in N by some occurrence sequence \(\sigma \) is denoted by Open image in new window . A system is a pair (N, M) of a Petri net N and a marking M. A system (N, M) is live if for every Open image in new window and every transition t some marking Open image in new window enables t. The system is 1safe if \(M'(p) \le 1\) for every Open image in new window and every place \(p \in P\).
Convention: Throughout this paper we assume that systems are 1safe, i.e., we identify “system” and “1safe system”.
Net Classes. A net \(N=(P,T,F)\) is a marked graph if Open image in new window and Open image in new window for every place \(p \in P\), and a freechoice net if for any two places \(p_1, p_2 \in P\) either Open image in new window or Open image in new window .
Nonsequential Processes of Petri Nets. An (A, B)labeled Petri net is a tuple \(N=(P, T, F, \lambda , \mu )\), where Open image in new window and Open image in new window are labeling functions over alphabets A, B. The nonsequential processes of a 1safe system (N, M) are acyclic, (P, T)labeled marked graphs. Say that a set \(P''\) of places of a (P, T)labeled acyclic net enables \(t \in T\) if all the places of \(P''\) are causally maximal, carry pairwise distinct labels, and Open image in new window .
Definition 1

The (P, T)labeled Petri net containing for each place \(p \in P\) marked at M one place \(\widehat{p}\) labeled by p, no other places, and no transitions, belongs to Open image in new window .
 If Open image in new window and \(P'' \subseteq P'\) enables some transition t of N, then the (P, T)labeled net \(\varPi _t =(P' \uplus \widehat{P}, T' \uplus \{\,\widehat{t}\,\}, F' \uplus \widehat{F}, \lambda \uplus \widehat{\lambda }, \mu \uplus \widehat{\mu })\), wherealso belongs to Open image in new window . We say that \(\varPi _t\) extends \(\varPi \).

Open image in new window , with \(\widehat{\lambda }(\,\widehat{p}\,)=p\), and \(\widehat{\mu }(\,\widehat{t}\,)=t\);

\(\widehat{F} = \{ (\,p'', \widehat{t}\,) \mid p'' \in P'' \} \cup \{ (\,\widehat{t}, \widehat{p}\,) \mid \widehat{p} \in \widehat{P}\}\);

We denote the minimal and maximal places of a process \(\varPi \) w.r.t. the causal order by \(\min (\varPi )\) and \(\max (\varPi )\), respectively.
As usual, we say that two processes are isomorphic if they are the same up to renaming of the places and transitions (notice that we rename only the names of the places and transitions, not their labels).

For every linearization \(\sigma = t_1' \ldots t_n'\) of \(T'\) respecting the causal order \(\preceq \), the sequence \(\mu (\sigma ) = \mu (t_1') \ldots \mu (t_n')\) is a firing sequence of (N, M). Further, all these firing sequences lead to the same marking. We call it the final marking of \(\varPi \), and say that \(\varPi \) leads from M to its final marking. For example, in Fig. 2 the sequences of the right process labeled by \(t_1 t_2 t_3 t_4\) and \(t_1 t_3 t_2 t_4\) are firing sequences leading to the marking \(M=\{p_2,p_5,p_7\}\).

For every firing sequence \(t_1 \cdots t_n\) of (N, M) there is a process \((P', T', F', \lambda , \mu )\) such that \(T'=\{t_1', \ldots , t_n'\}\), \(\mu (t_i') = t_i\) for every \(1 \le i \le n\), and \(\mu (t_i') \preceq \mu (t_j')\) implies \(i \le j\).

Open image in new window , and

Open image in new window is enabled at M.
It is wellknown that every sound freechoice workflow net is a 1safe system with the initial marking \(M_I\) [2, 7]. Given a workflow net according to this definition one can construct another one with one single input place i and output place o and two transitions \(t_i, t_o\) with Open image in new window and Open image in new window . For all purposes of this paper these two workflow nets are equivalent.
Given a workflow net N, we say that a process \(\varPi \) of \((N, M_I)\) is a run if it leads to \(M_O\). For example, the net in Fig. 1(b) is a run of the net in Fig. 1(a).
Petri Nets with Task Durations. We consider Petri nets in which, intuitively, when a token arrives in a place p it has to execute a task taking \(\tau (p)\) time units before the token can be used to fire any transition. Formally, we consider tuples \(N=(P,T,F,\tau )\) where (P, T, F) is a net and Open image in new window .
Definition 2
 (1)
for every \(p_1', p_2' \in P'\): if \(p_1' \prec p_2'\) then \(f(p_1') + \tau (\lambda (p_1')) \le f(p_2')\);
 (2)
for every \(p' \in P'\): \(f(p') + \tau (\lambda (p')) \le t\); and
 (3)
for every \(0 \le u < t\) there are at most k places \(p' \in P'\) such that \(f(p') \le u < f(p') + \tau (p')\).
We call a function f satisfying (1) a schedule, a function satisfying (1) and (2) a t schedule, and a function satsifying (1)–(3) a (k, t)schedule of \(\varPi \).
Intuitively, \(f(p')\) describes the starting time of the task executed at \(p'\). Condition (1) states that if \(p_1' \preceq p_2'\), then the task associated to \(p_2'\) can only start after the task for \(p_1'\) has ended; condition (2) states that all tasks are done by time t, and condition (3) that at any moment in time at most k tasks are being executed. As an example, the process in Fig. 1(b) can be executed with two resources in time 6 with the schedule \(i, p_1, p_2 \mapsto 0\); \(p_3, p_4 \mapsto 1\); \( p_7, p_6 \mapsto 3\), and \(p_8, p_9 \mapsto 4\).
Given a process \(\varPi = (P',T',F',\lambda ,\mu )\) of (N, M) we define the schedule \(f_{\min }\) as follows: if \(p' \in \min (\varPi )\) then \(f_{\min }(p')=0\), otherwise define \(f_{\min }(p') = \max \{f_{\min }(p'') + \tau (\lambda (p'')) \mid p'' \preceq p' \}\). Further, we define the minimal execution time \(t_{\min }(\varPi ) =\max \{ f(p') + \tau (\lambda (p'')) \mid p' \in \max (\varPi ) \}\). In the process in Fig. 1(b), the schedule \(f_{\min }\) is the function that assigns \(i, p_1, p_2, p_7~\mapsto ~0\), \(p_3, p_4~\mapsto ~1\), \(p_6, p_8~\mapsto ~3\), \(p_9~\mapsto ~4\), and \(o~\mapsto ~6\), and so \(t_{\min }(\varPi )= 6\). We have:
Lemma 1
A process \(\varPi = (P',T',F',\lambda ,\mu )\) of (N, M) can be executed within time \(t_{\min }(\varPi )\) with \(P'\) resources, and cannot be executed faster with any number of resources.
Proof
For \(k \ge P'\) resources condition (3) of Definition 2 holds vacuously. \(\varPi \) is executable within time t iff conditions (1) and (2) hold. Since \(f_{\min }\) satisfies (1) and (2) for \(t = t_{\min }(\varPi )\), \(\varPi \) can be executed within time \(t_{\min }(\varPi )\). Further, \(t_{\min }(\varPi )\) is the smallest time for which (1) and (2) can hold, and so \(\varPi \) cannot be executed faster with any number of resources.
3 Resource Threshold
We define the resource threshold of a run of a workflow net, and of the net itself. Intuitively, the resource threshold of a run is the minimal number of resources that allows one to execute it as fast as with unlimited resources, and the resource threshold of a workflow net is the minimal number of resources that allows one to execute every run as fast as with unlimited resources.
Definition 3
Let N be a workflow net, and let \(\varPi \) be a run of N. The resource threshold of \(\varPi \), denoted by Open image in new window is the smallest number k such that \(\varPi \) can be executed in time \(t_{\min }(\varPi )\) with k resources. A schedule of \(\varPi \) realizes the resource threshold if it is a Open image in new window schedule.
The resource threshold of N, denoted by Open image in new window , is defined by Open image in new window . A schedule ofN is a function that assigns to every process \(\varPi \in \mathcal{NP}(N, M)\) a schedule of \(\varPi \). A schedule of N is a (k, t)schedule if it assigns to every run \(\varPi \) a (k, t)schedule of \(\varPi \). A schedule of N realizes the resource threshold if it assigns to every run \(\varPi \) a Open image in new window schedule.
Example 1
We have seen in the previous section that for the process in Fig. 1(b) we have \(t_{\min }(\varPi )= 6\), and a schedule with two resources already achieves this time. So the resource bound is 2. The workflow net of Fig. 1 has infinitely many runs, in which loosely speaking, the net executes \(t_4\) arbitrarily many times, until it “exits the loop” by choosing \(t_6\), followed by \(t_7\) and \(t_8\). It can be shown that all processes have resource threshold 2, and so that is also the resource threshold of the net.
In the rest of the section we obtain two negative results about the result threshold. First, it is difficult to compute: Determining if the resource threshold exceeds a given threshold is NPcomplete even for acyclic marked graphs, a very simple class of workflows. Second, we show that even for acyclic freechoice workflow nets the resource threshold may not be realized by any online scheduler.
3.1 Resource Threshold Is NPcomplete for Acyclic Marked Graphs
More formally, the problem is defined as follows: Given jobs \(\mathcal{J}=\{J_1, \ldots , J_n\}\), where \(J_i\) has duration \(\tau (J_i)\) for every \(1 \le i \le n\), and a partial order \(\preceq \) on \(\mathcal{J}\), does there exist a function Open image in new window such thatGiven: a finite, partially ordered set of jobs with nonnegative integer durations, and nonnegative integers t and k.
Decide: Can all jobs can be executed with k machines within t time units in a way that respects the given partial order, i.e., a job is started only after all its predecessors have been finished?
 (1)
for every \(1\le i, j \le n\): if \(J_i \prec J_j\) then \(f(J_i) + \tau (J_i) \le f(J_j)\);
 (2)
for every \(1 \le i \le n\): \(f(J_i) + \tau (J_i) \le t\); and
 (3)
for every \(0 \le u < t\) there are at most k indices i such that \(f(J_i) \le u < f(J_i) + \tau (J_i)\).
These conditions are almost identical to the ones we used to define if a nonsequential process can be executed within time t with k resources. We exploit this to construct an acyclic workflow marked graph that “simulates” the scheduling problem. For the detailed proof, we refer to the full version of this paper [15].
Theorem 1
Given: An acyclic, sound workflow marked graph N, and a number k.
Decide: Does Open image in new window hold?
3.2 Acyclic FreeChoice Workflow Nets May Have no Optimal Online Schedulers
A resource threshold of k guarantees that every run can be executed without penalty with k resources. In other words, there exists a schedule that achieves optimal runtime. However, in many applications the schedule must be determined at runtime, that is, the resources must be allocated without knowing how choices will be resolved in the future. In order to formalize this idea we define the notion of an online schedule of a workflow net N.
Definition 4
Let N be a Petri net, and let \(\varPi \) and \(\varPi '\) be two processes of (N, M). We say that \(\varPi \) is a prefix of \(\varPi '\), denoted by \(\varPi \lhd \varPi '\), if there is a sequence \(\varPi _1, \ldots , \varPi _n\) of processes such that \(\varPi _1=\varPi \), \(\varPi _n=\varPi '\), and \(\varPi _{i+1}\) extends \(\varPi _i\) by one transition for every \(1 \le i \le n1\).
Let f be a schedule of (N, M), i.e., a function assigning a schedule to each process. We say that f is an online schedule if for every two runs \(\varPi _1, \varPi _2\), and for every two prefixes \(\varPi _1' \lhd \varPi _1\) and \(\varPi _2' \lhd \varPi _2\): If \(\varPi _1'\) and \(\varPi _2'\) are isomorphic, then \(f(\varPi _1') = f(\varPi _2')\).
Proposition 1
There is an acyclic, sound freechoice workflow net for which no online schedule realizes the resource threshold.
Proof
Indeed, observe that \(f_g\) and \(f_r\) execute \(\varPi _g\) and \(\varPi _r\) within time 5, and even with unlimited resources no schedule can be faster because of the task \(p_4\), while two or fewer resources are insufficient to execute either run within time 5.
 (a)
Every schedule \(f_g\) of \(\varPi _g\) that realizes the resource threshold (time 5 with 3 resources) satisfies \(f_g(p_3)=0\). Indeed, if \(f_g(p_3)\ge 1\), then \(f_g(p_5) \ge 3\), \(f_g(p_9) \ge 6\), and finally \(f_g(o)\ge 6\), so \(f_g\) does not meet the time bound.
 (b)
Every schedule \(f_r\) of \(\varPi _r\) that realizes the resource threshold (time 5 with 3 resources) satisfies \(f_r(p_3) > 0\).
Observe first that we necessarily have \(f_r(p_4)=0\), and so a resource, say \(R_1\), is bound to \(p_4\) during the complete execution of the workflow, leaving two resources left. Assume \(f_r(p_3) = 0\), i.e., a second resource, say \(R_2\), is bound to \(p_3\) at time 0, leaving one resource left, say \(R_3\). Since both \(p_1\) and \(p_2\) must be executed before \(p_8\), and only \(R_3\) is free until time 2, we get \(f_r(p_8) \ge 2\). So at time 2 we still have to execute \(p_6, p_7, p_8\) with resources \(R_2, R_3\). Therefore, two out of \(p_6, p_7, p_8\) must be executed sequentially by the same resource. Since \(p_6, p_7, p_8\) take 2 time units each, one of the two resources needs time 4, and we get \(f_r(o) \ge 6\).
By this claim, at time 0, an online schedule has to decide whether to allocate a resource to \(p_3\) or not, without knowing which of \(t_3\) or \(t_4\) will be executed in the future. If it schedules \(f(p_3)=0\) and later \(t_4\) occurs, then \(\varPi _r\) is executed and the deadline of 5 time units is not met. The same occurs if it schedules \(f(p_3)>0\), and later \(t_3\) occurs.
4 Concurrency Threshold
Due to the two negative results presented in the previous section, we study a different parameter, introduced in [4], called the concurrency threshold. During execution of a business process, information on the resolution of future choices is often not available, and further no information on the possible duration of a task (or only weak bounds) are known. Therefore, the scheduling is performed in practice by assigning a resource to a task at the moment some resource becomes available. The question is: What is the minimal number of resources needed to guarantee the optimal execution time achievable with an unlimited number of resources?
The answer is simple: since there is no information about the duration of tasks, every reachable marking of the workflow net without durations may be also reached for some assignment of durations. Let M be a reachable marking with a maximal number of tokens, say k, in places with positive duration, and let \(d_1 \le d_2 \le \cdots \le d_k\) be the durations of their associated tasks. If less than k resources are available, and we do not assign a resource to the task with duration \(d_k\), we introduce a delay with respect to the case of an unlimited number of resources. On the contrary, if the number of available resources is k, then the scheduler for k resources can always simulate the behaviour of the scheduler for an unlimited number of resources.
Definition 5
The following lemma follows easily from the definitions.
Lemma 2
For every workflow net N: Open image in new window .
Proof
Follows immediately from the fact that for every schedule f of a run of N, there is a schedule g with Open image in new window machines such that \(g(p) \le f(p)\) for every place p.
In the rest of the paper we study the complexity of computing the concurrency threshold. In [4], it was shown that the threshold can be computed in polynomial time for regular workflows, a class with a very specific structure, and the problem for the general freechoice case was left open. In Sect. 4.1 we prove that the concurrency threshold of marked graphs can be computed in polynomial time by reduction to a linear programming problem over the rational numbers. In Sect. 4.2 we study the freechoice case. We show that deciding if the threshold exceeds a given value is NPcomplete for acyclic, sound freechoice workflow nets. Further, it can be computed by solving the same linear programming problem as in the case of marked graphs, but over the integers. Finally, we show that in the cyclic case the problem remains NPcomplete, but the integer linear programming problem does not necessarily yield the correct solution.
4.1 Concurrency Threshold of Marked Graphs

If \(M_2\) is reachable from \(M_1\) in N, then \({\varvec{M}}_\mathbf{2} = {\varvec{M}}_\mathbf{1} + {\varvec{N}} \cdot {\varvec{X}}\) for some integer vector \({\varvec{X}} \ge 0\).

If N is a marked graph and \({\varvec{M}}_\mathbf{2} = {\varvec{M}}_\mathbf{1} + {\varvec{N}} \cdot {\varvec{X}}\) for some rational vector \({\varvec{X}} \ge 0\), then \(M_2\) is reachable from \(M_1\) in N.

If N is acyclic and \({\varvec{M}}_\mathbf{2} = {\varvec{M}}_\mathbf{1} + {\varvec{N}} \cdot {\varvec{X}}\) for some integer vector \({\varvec{X}} \ge 0\), then \(M_2\) is reachable from \(M_1\) in N.
Theorem 2

If N is a marked graph, then Open image in new window .

If N is acyclic, then Open image in new window .
In particular, it follows that Open image in new window can be computed in polynomial time for marked graphs, acyclic or not. (The result about acyclic nets is used in the next section.)
4.2 Concurrency Threshold of FreeChoice Nets
We study the complexity of computing the concurrency threshold of freechoice workflow nets. We first show that, contrary to numerous other properties for which there are polynomial algorithms, deciding if the concurrency threshold exceeds a given value is NPcomplete.
Theorem 3
Given: A sound, freechoice workflow net \(N=(P,T,F,I,O)\), and a number \(k \le T\).
Decide: Is the concurrency threshold of N at least k?
Proof
Given a graph \(G=(V,E)\), we construct a sound freechoice workflow net \(N_G\) in polynomial time as follows:Given: An undirected graph \(G=(V, E)\), and a number \(k \le V\).
Decide: Is there a set Open image in new window such that Open image in new window and Open image in new window for every Open image in new window ?

For each \(e=\{v, u\} \in E\) we add to \(N_G\) the “gadget net” \(N_e\) shown in Fig. 5(a), and for every node v we add the gadget net \(N_v\) shown in Fig. 5(b).

For every \(e=\{v, u\} \in E\), we add an arc from the place \([e,v]^4\) of \(N_e\) to the transition \(v^1\) of \(N_v\), and from \([e,u]^4\) to the transition \(u^1\) of \(N_u\).

The set I of initial places contains the place \(e^0\) of \(N_e\) for every edge e; the set O of output places contains the places \(v^2\) of the nets \(N_v\).
It is easy to see that \(N_G\) is freechoice and sound, and in [15] we show the result of applying the reduction to a small graph and prove that G has an independent set of size at least k iff the concurrency threshold of \((N_G,M_I)\) is at least \(2E+k\). The intuition is that for each edge \(e \in E\), we fire the transition \([e,u]^1\) where Open image in new window , and for each Open image in new window , we fire the transition \(v^1\), thus marking one of \([e,u]^2\) or \([e,v]^2\) for each edge \(e \in E\) and the place \(v^2\) for each Open image in new window .
4.3 Approximating the Concurrency Threshold
Recall that the solution of problem (1) over the rationals or the integers is always an upper bound on the concurrency threshold for any Petri net (Theorem 2). The question is whether any stronger result holds when the workflows are sound and freechoice. Since computing the concurrency threshold is NPcomplete, we cannot expect the solution over the rationals, which is computable in polynomial time, to provide the exact value. However, it could still be the case that the solution over the integers is always exact. Unfortunately, this is not true, and we can prove the following results:
Theorem 4
Given a Petri net N, let Open image in new window and Open image in new window be as in Theorem 2.
 (a)
There is an acyclic sound freechoice workflow net N such that Open image in new window .
 (b)
There is a sound freechoice workflow net N such that and let Open image in new window .
Proof
For (a), we can take the net obtained by adding to the gadget in Fig. 5(a) a new transition with input places \([e,v]^4\) and \([e,u]^4\), and an output place o with weight 2. We take \(e^0\) as input place. The concurrency threshold is clearly 2, reached, for example, after firing \([e,v]^1\). However, we have Open image in new window , reached by the rational solution \({\varvec{X}} = (1/2, 1/2, \ldots , 1/2)\). Indeed, the marking equation then yields the marking M satisfying \(M([e,v]^2) = M([e,u]^2)=M(o)=1/2\).
For (b), we can take the workflow net of Fig. 6. It is easy to see that the concurrency threshold is equal to 1. The marking \({\varvec{M}}\) that puts one token in each of the two places with weight 1, and no token in the rest of the places, is not reachable from \(M_I\). However, it is a solution of the marking equation, even when solved over the integers. Indeed, we have \({\varvec{M}} = {{\varvec{M}}}_{{\varvec{I}}} + {\varvec{N}} \cdot {\varvec{X}}\) for X=(1,0,1,1,0,0,1). Therefore, the upper bound derived from the marking equation is 2.
5 Concurrency Threshold: A Practical Approach
We have implemented a tool^{1} to compute an upper bound on the concurrency threshold by constructing a linear program and solving it by calling the mixedinteger linear programming solver Cbc from the COINOR project [14]. Additionally, fixing a number k, we used the stateofthe art Petri net model checker LoLA [19] to both establish a lower bound, by querying LoLA for existence of a reachable marking M with Open image in new window ; and to establish an upper bound, by querying LoLA if all reachable markings \(M'\) satisfy Open image in new window .
We evaluated the tool on a set of 1386 workflow nets extracted from a collection of five libraries of industrial business processes modeled in the IBM WebSphere Business Modeler [9]. For the concurrency threshold, we set \(D = P \setminus O\). These nets also have multiple output places, however with a slightly different semantics for soundness allowing unmarked output places in the final marking. We applied the transformation described in [12] to ensure all output places will be marked in the final marking. This transformation preserves soundness and the concurrency threshold.
On all 642 sound nets, we computed an upper bound on the concurrency threshold using our tool, both using rational and integer variables. We computed lower and upper bounds using LoLA with the value Open image in new window of the concurrency threshold. We report the results for computing the lower and upper bound separately.
All experiments were performed on the same machine equipped with an Intel Core i76700K CPU and 32 GB of RAM. The results are shown in Table 1. Using the linear program, we were able to compute an upper bound for all nets in total in less than 7 s, taking at most 30 ms for any single net. LoLA could compute the lower bound for all nets in 6 s LoLA fails to compute the upper bound in three cases due to reaching the memory limit of 32 GB. For the remaining 639 nets, LoLA could compute the upper bound within 7 min in total.
Statistics on the size and analyis time for the 642 nets analyzed. The times marked with \(^*\) exclude the 3 nets where LoLA reaches the memory limit.
Net size  Analysis time (sec)  

Median  21  14  16  3  0.01  0.01  0.01  0.01 
Mean  28.4  18.6  \(3\cdot 10^{14}\)  3.7  0.01  0.01  0.01  0.58\(^*\) 
Max  262  284  \(2\cdot 10^{17}\)  66  0.03  0.03  1.18  307.76\(^*\) 
Comparing the values of the upper bound, first we observed that we obtained the same value using either rational or integer variables. The time difference between both was however negligible. Second, quite surprisingly, we noticed that the upper bound obtained from the linear program is exact in all of our cases, even for the cyclic ones. Further, it can be computed much faster in several cases than the upper bound obtained by LoLA and it gives a bound in all cases, even when the statespace exploration reaches its limit. By combining linear programming for the upper bound and statespace exploration for the lower bound, an exact bound can always be computed within a few seconds.
6 Conclusion
Planning sufficient execution resources for a business or production process is a crucial part of process engineering [3, 13, 20]. We considered a simple version of this problem in which resources are uniform and tasks are not interruptible. We studied the complexity of computing the resource threshold, i.e., the minimal number of resources allowing an optimal makespan. We showed that deciding if the resource threshold exceeds a given bound is NPhard even for acyclic marked graphs. For this reason, we investigated the complexity of computing the concurrency threshold, an upper bound of the resource threshold introduced in [4]. Solving a problem left open in [4], we showed that deciding if the concurrency threshold exceeds a given bound is NPhard for general sound freechoice workflow nets. We then presented a polynomialtime approximation algorithm, and showed experimentally that it computes the exact value of the concurrency threshold for all benchmarks of a standard suite of freechoice workflow nets.
Footnotes
 1.
The tool is available from https://gitlab.lrz.de/i7/macaw.
References
 1.van der Aalst, W.M.P.: Verification of workflow nets. In: Azéma, P., Balbo, G. (eds.) ICATPN 1997. LNCS, vol. 1248, pp. 407–426. Springer, Heidelberg (1997). https://doi.org/10.1007/3540631399_48CrossRefGoogle Scholar
 2.van der Aalst, W.M.P.: Workflow verification: finding controlflow errors using Petrinetbased techniques. In: van der Aalst, W., Desel, J., Oberweis, A. (eds.) Business Process Management. LNCS, vol. 1806, pp. 161–183. Springer, Heidelberg (2000). https://doi.org/10.1007/3540455949_11CrossRefGoogle Scholar
 3.Bessai, K., Youcef, S., Oulamara, A., Godart, C., Nurcan, S.: Resources allocation and scheduling approaches for business process applications in cloud contexts. In: 4th IEEE International Conference on Cloud Computing Technology and Science Proceedings, CloudCom 2012, Taipei, Taiwan, 3–6 December 2012, pp. 496–503 (2012)Google Scholar
 4.Botezatu, M., Völzer, H., Thiele, L.: The complexity of deadline analysis for workflow graphs with multiple resources. In: La Rosa, M., Loos, P., Pastor, O. (eds.) BPM 2016. LNCS, vol. 9850, pp. 252–268. Springer, Cham (2016). https://doi.org/10.1007/9783319453484_15CrossRefGoogle Scholar
 5.Chuzhoy, J., Codenotti, P.: Resource minimization job scheduling. In: Dinur, I., Jansen, K., Naor, J., Rolim, J. (eds.) APPROX/RANDOM 2009. LNCS, vol. 5687, pp. 70–83. Springer, Heidelberg (2009). https://doi.org/10.1007/9783642036859_6CrossRefGoogle Scholar
 6.Chuzhoy, J., Guha, S., Khanna, S., Naor, J.: Machine minimization for scheduling jobs with interval constraints. In: 45th Symposium on Foundations of Computer Science (FOCS 2004), 17–19 October 2004, Rome, Italy, Proceedings, pp. 81–90 (2004)Google Scholar
 7.Desel, J., Esparza, J.: Free Choice Petri Nets. Cambridge University Press, Cambridge (1995)CrossRefGoogle Scholar
 8.Esparza, J., Hoffmann, P., Saha, R.: Polynomial analysis algorithms for free choice probabilistic workflow nets. In: Agha, G., Van Houdt, B. (eds.) QEST 2016. LNCS, vol. 9826, pp. 89–104. Springer, Cham (2016). https://doi.org/10.1007/9783319434254_6CrossRefGoogle Scholar
 9.Fahland, D., Favre, C., Jobstmann, B., Koehler, J., Lohmann, N., Völzer, H., Wolf, K.: Instantaneous soundness checking of industrial business process models. In: Dayal, U., Eder, J., Koehler, J., Reijers, H.A. (eds.) BPM 2009. LNCS, vol. 5701, pp. 278–293. Springer, Heidelberg (2009). https://doi.org/10.1007/9783642038488_19CrossRefGoogle Scholar
 10.Favre, C., Fahland, D., Völzer, H.: The relationship between workflow graphs and freechoice workflow nets. Inf. Syst. 47, 197–219 (2015)CrossRefGoogle Scholar
 11.Hall, N.G., Sriskandarajah, C.: A survey of machine scheduling problems with blocking and nowait in process. Oper. Res. 44(3), 510–525 (1996)MathSciNetCrossRefGoogle Scholar
 12.Kiepuszewski, B., ter Hofstede, A.H.M., van der Aalst, W.M.P.: Fundamentals of control flow in workflows. Acta Inf. 39(3), 143–209 (2003)MathSciNetCrossRefGoogle Scholar
 13.Liu, L., Zhang, M., Lin, Y., Qin, L.: A survey on workflow management and scheduling in cloud computing. In: 14th IEEE/ACM International Symposium on Cluster, Cloud and Grid Computing, CCGrid 2014, Chicago, IL, USA, 26–29 May 2014, pp. 837–846 (2014)Google Scholar
 14.LougeeHeimer, R.: The common optimization interface for operations research: promoting opensource software in the operations research community. IBM J. Res. Dev. 47(1), 57–66 (2003)CrossRefGoogle Scholar
 15.Meyer, P.J., Esparza, J., Völzer, H.: Computing the concurrency threshold of sound freechoice workflow nets. arXiv:1802.08064 [cs.LO] (2018)
 16.Murata, T.: Petri nets: properties, analysis, and applications. Proc. IEEE 77(4), 541–576 (1989)CrossRefGoogle Scholar
 17.Pinedo, M.L.: Scheduling: Theory, Algorithms, and Systems. Springer, New York (2016). https://doi.org/10.1007/9781461423614CrossRefMATHGoogle Scholar
 18.Ullman, J.: NPcomplete scheduling problems. J. Comput. Syst. Sci. 10(3), 384–393 (1975)MathSciNetCrossRefGoogle Scholar
 19.Wolf, K.: Generating Petri net state spaces. In: Kleijn, J., Yakovlev, A. (eds.) ICATPN 2007. LNCS, vol. 4546, pp. 29–42. Springer, Heidelberg (2007). https://doi.org/10.1007/9783540730941_5CrossRefGoogle Scholar
 20.Xu, J., Liu, C., Zhao, X.: Resource allocation vs. business process improvement: how they impact on each other. In: Dumas, M., Reichert, M., Shan, M.C. (eds.) BPM 2008. LNCS, vol. 5240, pp. 228–243. Springer, Heidelberg (2008). https://doi.org/10.1007/9783540857587_18CrossRefGoogle Scholar
Copyright information
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made. The images or other third party material in this book are included in the book's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the book's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.