Externalities, Tangible Externalities, and Queue Disciplines

Author(s): Moshe Haviv and Ya'acov Ritov

Source: Management Science, Vol. 44, No. 6 (Jun., 1998), pp. 850-858
Published by: INFORMS
Stable URL: http://www.jstor.org/stable/2634653
Accessed: 19-02-2018 22:23 UTC

REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/2634653?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms

INFORMS is collaborating with JSTOR to digitize, preserve and extend access to Management
Science

This content downloaded from 201.234.181.53 on Mon, 19 Feb 2018 22:23:54 UTC
All use subject to http://about.jstor.org/terms
 Externalities, Tangible Externalities, and
Queue Disciplines

Moshe Haviv * Ya'acov Ritov

Department of Statistics, The Hebrew University of Jerusalem, 91905 Jerusalem, Israel
Department of Econometrics, The University of Sydney, Sydney NSW 2006, Australia

E xternalities are the (marginal) costs that a user of a common resource imposes on others.
We introduce the efficient measure of tangible externalities that are the costs that a user
imposes on others while being served. Then, for a single server queueing system under various
service disciplines, we compute the expected externalities and the expected tangible externalities
conditioning on the length of the service requirement.
(Externalities; Cognestion Costs; Queue Discipline)

1. Introduction The second purpose for explicitly defining extemalities

Users of a common facility may impose costs on each is that various ways of managing a common facility
other, costs that are known as negative externalities or may result in different amounts of extemalities imposed

congestion costs. In many cases, these externalities are in by the same user. Hence, the resulting pricing mecha-
the form of delays that are imposed on users prior to nism can serve as an additional criterion for deciding

the completion of their tasks. This is, for example, the which management scheme to adopt. We next elaborate
case when airplanes take off from a runway, when com- on these issues while considering arrivals to an M / M / 1
muters cross a bridge, when jobs share a common CPU, queue under various service disciplines. In many cases

and when messages are routed through the same data
network. These models and many others are examples
of queueing systems. There is hence a need to quantify
this notion. This is in particular the case due to the re-
cent growth in the use of the Internet. For this context
see Walrand and Varaiya (1996).
The amount of externalities that a user imposes on
others is typically a function of his own requirement.
our results are applicable for M / G / 1 queues, too.
Recall that M / M/ 1 stands for a single server queue
with a Poisson arrival process and exponential service
requirement. The waiting room is unlimited, all random
variables involved are independent, and the system is
considered to be in steady state. The entrance to service
discipline is not specified yet. Let X be the rate of arrival.
Each arrival requires an exponential amount of service

Specifically, in the context of service systems, the longer with an expected value of ,p-1. Denote X /,u by p, which
his service requirement is (or its expected value, re- is referred to as the traffic intensity. Of course, 0 c p

spectively), the larger are the negative externalities (or < 1. Finally, M/G/1 queues are defined as M/M/1

their expected value, respectively) he imposes. Defining queues but without the assumption of exponential ser-
quantitatively the externalities can serve two purposes. vice time. In this case, too, p stands for the traffic inten-

First, a zero profit operator who charges users for the
use of a common facility usually likes to do so in accor-
dance with the congestion costs that they impose on
others. This is, of course, on top of other charges such
as access charges, user charges, or service quality
charges. See Walrand and Varaiya (1996) for details.
sity, namely AK1 where Kl is the expected service re-
quirement.
The standard definition for externalities that an in-
dividual imposes on others is the difference between the
actual total queueing time of the others and the corre-
sponding value in a (simulated) realization in which
everything stays the same but without the presence of

0025-1909 /98/4406/ 0850$05.00

Copyright X 1998, Institute for Operations Research
850 MANAGEMENT SCIENCE/Vol. 44, No. 6, June 1998 and the Management Sciences

This content downloaded from 201.234.181.53 on Mon, 19 Feb 2018 22:23:54 UTC
All use subject to http://about.jstor.org/terms
 HAVIV AND RITOV
Externalities, Tangible Externalities, and Queue Disciplines

the user in question. By saying that everything stays the
same we mean that all other arrivals occur at the same
instants and with the same workload requirements. For-
mally, we consider a system, as described above, that is
in a steady state at time 0, and at that time a customer
with a fixed service requirement of x is added to the
system. We analyze the externalities imposed by this
customer. See ?2 for details. For more on this measure
of externalities and its applications see Carlin and Park
(1970), Dewan and Mendelson (1990), Dolan (1978),
Hassin (1983), and Hassin and Haviv (1994), or Loch
(1991), Mendelson and Whang (1990), and Newbery
(1990).
We like to mention that another approach for the han-served earlier. Of course, other customers might be
dling of externalities appears in the literature which is
the marginal added waiting cost per unit of time to the
society due to the infinitesimal last arrival. Specifically,
if W(A) is the expected waiting time when the arrival
rate is X (and hence the total waiting cost per unit of
time is XW(A)), one looks at
dA (AW(A)) = W(A) + X d W(A).
For this approach see Mendelson and Whang (1990),
Dewan and Mendelson (1990) or Walrand and Varaiya
(1996, p. 386).
Externalities as defined above suffer from the lack of
efficiency, namely, their sum does not coincide with the
total queueing time. Actually, this sum overestimates
the total queueing time. For example, consider a single
server, first come first served (FCFS) queue where three
customers A, B, and C with identical service require-
ments are the first to arrive, and in this order, at an
empty queue. Also, assume that the arrival times of cus-
tomers A and B (almost) coincide, while customer C
arrives when customer A is ready to clear the system.
Hence, the externalities of both customers A and B in-
clude the waiting time of customer C. We then conclude
that externalities cannot be used as a pricing mechanism
for charging customers for their share of the total
queueing cost.
We suggest here an alternative measure. For each ar-
rival of a customer we associate what we call a tangible
externality. This is defined as the total queueing time
added to the others while this customer is in service. This as it is strong and work conserving. There, for p as the
measure is efficient since any part of the total queueing
time is now associated with one and only one customer.
In particular, the sum of tangible externalities across
customers equals the total queueing time. Hence, tan-
gible externalities can be used as an efficient pricing
mechanism for payment to be made by the customers.
The above suggested mechanism, used as is, might
seem a bit unfair for some customers. For example, in a
FCFS system, a customer who is asked to pay a bundle
might claim that it was not his fault that many custom-
ers arrived during his service period: he was served
during a particular period in which the number of ar-
rivals was high by a mere chance or just because he was
delayed service by other customers who have been
more lucky. A better measure is the expected tangible
externalities that a customer imposes as a function of
his (actual) service requirement. This measure has the
same expectation (over all customers) as the tangible
externalities, but its variance is smaller.
Besides variance reduction, there is an additional ad-
vantage to the expected tangible externalities over tan-
gible externalities. This is the simplicity of using them
as a pricing mechanism: there is no need to monitor how
many customers are in the system upon each service
commencement, how many arrive, and at exactly which
instants during this service period. All that is required
is to monitor the length of the service requirements
themselves. Actually, the queue operator can place a
sign that tells each customer how much to pay as a func-
tion of his service requirement.
Before describing the content of the rest of the paper,
we need the following terminology and results.
A queueing discipline is said to be strong if the order
in which customers receive service is not a function of
their actual service requirements (i.e., future values
from the operator's point of view). Put differently, the
queue operator is not informed of these values while
deciding who gets service when. Also, a queueing sys-
tem is said to be work-conserving if the resulting process
of total (residual) service requirement due to customers
currently in the system (also known as the virtual wait-
ing time) coincides with that of a FCFS system having
the same characteristics. Consider now an M / M /1
queueing system with any entrance discipline so long
traffic intensity and for ,-1 as the expected service

MANAGEMENT SCIENCE/ Vol. 44, No. 6, June 1998 851

This content downloaded from 201.234.181.53 on Mon, 19 Feb 2018 22:23:54 UTC
All use subject to http://about.jstor.org/terms
 HAVIV AND RITOV
Externalities, Tangible Externalities, and Queue Disciplines

requirement, the expected externalities per arrival (in a for the variability of the prices E(C I x). Here again, the
steady state) equals p / (,(1 - p)2). (See Proposition4.1.) values for E(C I x) (and hence, of course, for
This value is larger than its expected queueing time, p / Var(E(T I x))), for the last two queueing disciplines co
(,(1 - p)). In other words, the sum of expected exter- incide. Finally, for an M / M /1 queue under all three
nalities overestimates the total cost by a ratio of (1 cases we considered,
-p)-1. E(T x) = 1
The rest of the paper contains our detailed analysis
E(CIx) 1 -p' ?<o
of expected externalities and of tangible externalities
and it is organized as follows. In ?2 we derive the ex- However, this ratio does not hold for all strong work-
pected externalities that a customer imposes on others conserving disciplines. A counterexample and some
given his own service requirement. This is done sepa- other remarks are given in our final section, ?4.
rately for the following three cases:
1. Any strong and work-conserving discipline which
2. The Expected Externalities
does not allow preemption. Among them, FCFS, ran-
dom order queues and last come first served without 2.1. General Results
preemption (LCFS-HOL).1 In this section we compute the expected externalities
2. Last come first served with preemptive resume imposed by an x-customer on the others, E(T I x), and
(LCFS-PR). the resulting variance, Var(E(T l x)), in an M / G/1
3. Processor sharing (PS). queue under the first and the second disciplines and in
Detailed description of the last two disciplines are an M / M / 1 queue under the third discipline. However,
given in the next section. some notation and preliminary results are needed first.
Call a customer whose service requirement is x an x- Our analysis is conditioned on an arrival of a cus-
customer and let T I x be the (random) externalities that tomer with a service requirement of x at time 0. To be
he imposes on others. Then, let E(T I x) be the corre- more exact, we condition on an arrival of a customer
sponding conditional expected value and let Var(E(T I x))with service requirement in the interval (x - c, x + e)
be the variance of the conditional expected values. Note during the interval (- e, e) and consider the limit as
that Var(E(T I x)) (as oppose to Var(T)), for example) is e -- 0. Note that since the arrivals during disjoint inter-
the right measure for the variability for the prices paid vals are independent, the x-customer observes a Poisson
by customers in the case that the price mechanism of arrival process with independent service requirements
externalities is adopted. In the next section we compute distributed according to some distribution function G.
these values for an M / G/1 queue for the first two of We next refer to a simulated realization in which all
the above mentioned cases and for an M / M / 1 queue random variables get the same values but without the
for the last. As it turns out, the values E(T I x) (and presence of the x-customer. The case with him will be
hence, of course for Var(E(T I x))), for the last two cases referred to as the original case. Let M denote the number
coincide. of customers (other than the x-customer) who exit the
Section 3 considers tangible externalities. Specifically, original system from the (first) service commencement
of the x-customer and until the first time the server is
let C I x be the (random) tangible externalities incurred
by an x-customer. Our first interest is in E(C I x), whichidle. For 1 c i c M, let ri be the total amount of service
is the conditional expected tangible externalities im- received by the x-customer until the i-th departure time
posed by an x-customer. In ?3 we compute E(C I x) and
(in the original system). Finally, let Ii be the total idle
Var(E(C I x)) for the three abovementioned cases for M /time (in the simulated system) of the server between the
arrival of the x-customer and the arrival of the ith cus-
G / 1 queues. As said above with regard to Var(E(T I x)),
we can say here that Var(E(C I x)) is the right measure
tomer in the simulated system.
We next exemplify the notation just introduced. Sup-
pose that x = 5 and that the tagged x-customer arrives
1 HOL stands for head of the line. at an empty FCFS queue. Suppose a customer arrives at

852 MANAGEMENT SCIENCE/Vol. 44, No. 6, June 1998

This content downloaded from 201.234.181.53 on Mon, 19 Feb 2018 22:23:54 UTC
All use subject to http://about.jstor.org/terms
 HAVIV AND RITOV
Externalities, Tangible Externalities, and Queue Disciplines

time 1 and requires service of 2 units. Another customer tI = i - Ti +Ili. (4)

arrives at time 2 and requires service of 3 units. Finally,
at time 9, a third customer arrives with demand of 2 Note also that in this case Tr is eith

units. At time 12, after the service requirements of these served before the tagged customer, or x, for those who

4 customers are met, the queue is empty again for the are served after his departure. This completes our proof

first time. Then the externalities imposed by the tagged for (1), case (i). For case (ii), Equation (4) does not hold

customer are (7 - 3) + (10 - 6) + (12 - 11) = 9. Clearly, as is. However, we can assume that all service between

M = 3. Since the other three customers are served after two consecutive exits from the simulated system was
given to the next to depart customer. This statistically
the tagged customer is served completely, ri = 5 for all
changes nothing due to the memoryless property of the
of them. Finally, I, = 1, I2 = 1, (the length of the first
time period before the arrival of the first customer) and service requirement, but now (4) holds for case (ii) as
I3 = 4 (the above time plus the idle time between the well.

departure of the second customer in the simulated sys- Next we prove (2) (for both cases (i) and (ii)). Let Ji
tem and the arrival of the third). Hence, we obtain that be the length of the ith idle period in the simulated sys-

the externalities are tem occurring during the time interval between the first
service commencement of the x-customer and the first
M M
time in which the server is idle in the original system,
Externalities = T i Ii.
i=l i=l for i, 1 ? i ? L, where L denotes the number of such
idle periods. Note that L = inf{j: ELji i 2 x}, that L
This is an example of the following proposition.
2 1 and that IM = EL-1 Ji. Clearly, Jl, . . ., JL are inde-
PROPOSITION 2.1. For (i) any strong and work- pendent and exponentially distributed random vari-
conserving M / G / 1 queue without preemption, or ables with expected value of A-1. Thus, L - 1 follows a
(ii) any
strong and work-conserving M / M / 1 queue, Poisson distribution with parameter Xx. In particular,
E(L) = Xx + 1. Note that the time interval between two
E(Tlx) =E , Ti IiF (1) idle periods in the simulated system corresponds to a
i=l i=l
__ 2

busy period in an M / G / 1 queue. Thus, for j, 1 ? j -

L, let Kj be the number of customers to arrive at the
= Et Ti} - 2(1 )X (2) simulated system (and be served) during the busy pe-
riod between the jth and the (j + 1)th idle periods. In
PROOF. Lett, _ t2 c- tM (tlc t2' < t'M particular, E(Kj) = (1 - p)-1. Also, note that M = >=1
respectively) be the departure instants in the original Kj. Hence,
(simulated, respectively) system since the service com-
M L-1 i
mencement of the x-customer until the end of the orig-
E E I, = E E Ki E
inal busy period in which x is served while in the sim- i=1 i=1 j=1
ulated system the departure of the x customer is ex- L-1 i

cluded. Note that M denotes the number of such

=(1 - p)-1 E E E Jj
departing customers. Of course, i=1 j=1

M L

E(TIx) = E E (ti - t[). (3) = (1 - p)-1 E E (L -1)11

i=l j=1

Note that E(T I x) does not depend on the order in which

the M customers leave service in the two systems, but
only on the marginal distribution of the departure
times. In the nonpreemptive case, we can simulate the
departure times such that the orders of departures in
the two systems coincide and hence
= (1- P)-1 [E{L E J4 -E E{JjjD(L j)
where D( ) is the indicator random variable. Now, for
any j, j 2 1, D(L 2 j) is determined by JI, . . ., Iji-1, and,
hence, Jj is independent of D(L 2 j). Also, because Jj, 1

MANAGEMENT SCIENCE/Vol. 44, No. 6, June 1998 853

This content downloaded from 201.234.181.53 on Mon, 19 Feb 2018 22:23:54 UTC
All use subject to http://about.jstor.org/terms
 HAVIV AND RITOV
Externalities, Tangible Externalities, and Queue Disciplines

? j c L, are exponentially distributed, L is independentNote that the time between his departure and the end
of 4=1 Ji. Hence, of the busy period is statistically identical to the sum of
M busy periods, each of which corresponds to one of the
E , Ii = (1 - p)-1[X-1(Xx + 1)2 customers present in the system at that instant. Since
the expected number of customers to be served in a busy
period is 1 / (1 - p), we conclude that
_-1 2 E{L(L + 1)}] (5)

= 1X-1(1 - p)-1(Xx)2, (6) E(M) = 2 + Xx

since L - 1 is Poisson with mean AX. D
The rest is simple algebra. D
2.2. The Case of M / G / 1 Queues Without
2.3. The LCFS-PR M / G / 1 Queue
Preemption
In the LCFS-PR discipline (also called a stack), an arrival
We need the following notation. Denote by K, the nth
starts receiving service as soon as he arrives, possibly
moment of the service requirement. In particular, p
preempting the customer who is in service. Preempted
equals XK < 1. Note that in the case of exponential ser-
customers return to service in a reverse order of their
vice Kn = n! /,un.
arrival. Finally, service is resumed from the point where
THEOREM 2.1. For any strong work-conservingMIG/ it was interrupted.
1 queue without preemption, in particular for FCFS, LCFS-
THEOREM 2.2. For the LCFS-PR M / G / 1 system,
HOL, and nonpreemption random M / G / 1 queues,

E(TIx)= X,
E(TIx) x+ ( 2 x (7)
2 ( - p) 2(1l-p)2
and
and

2 2 p22.
Var(E(T I x)) = ) (K2 - K).
Var(E(T x)) = 4(1 - (K4 - K2

PROOF. Note that in this case, only customers that

13K2
are in queue at the time of arrival of the x-customer
+ 2(1 _ P)3 - 1211)
suffer any additional delay due to his arrival. Their ex-
pected number is p / (1 - p) and each one of them has
+ \L )4 (K2 -K
to wait during the busy period that initializes with the
arrival of the x-customer and ends upon his departure.
PROOF. Under the nonpreemption assumption, no
The expected length of this period (which coincides
one leaves the (original) system during the service pe-
with the expected busy period that begins with a service
riod of the x-customer. Using the notation of Proposi-
of length x) is x / (1 - p), and the theorem follows. [
tion 2.1, we get that Tr = x for all i = 1, ..., M, and
then, by (2) 2.4. The Processor Sharing M / M / 1 Queue
In this queueing model the server, at any time, shares
E(TIx) = E(M)x- x2. (8) evenly its capacity among all customers present in the
2(1 - p)
system. This model is the limit of the round-robin
Thus, we now turn our attention to E(M). Note that for scheme. Specifically, in a round-robin scheme all cus-
the queueing system considered here, the expected tomers present in the system alternate receiving service
number of customers in the queue upon service com- of some quantum. So if the time length of a cycle is A
mencement is well-known to equal X2K2 /2(1 - p). and
To a fixed number of n + 1 customers are present dur-
this value we have to add Xx in order to get the number
ing this cycle, then each of them receives an amount of
in the system at the time of departure of the x-customer.
service of A / (n + 1) (and imposes tangible externalities

854 MANAGEMENT SCIENCE/Vol. 44, No. 6, June 1998

This content downloaded from 201.234.181.53 on Mon, 19 Feb 2018 22:23:54 UTC
All use subject to http://about.jstor.org/terms
 HAVIV AND RITOV
Externalities, Tangible Externalities, and Queue Disciplines

of nA/l(n + 1) on the others). Note that in order to
receive A units of service, a customer has to spend (n
dr as the service increment for the x-customer and not
+ 1 )A units of time in the system. The processor sharing
model corresponds to the case where A goes to zero. A as the real time increment. Specifically, denote the dis-
way to interpret this model is by assuming that a cus-
tomer has his own watch which runs only while he is
in service. In particular, an x-customer leaves when his
watch says x. In comparison with the natural clock, this
watch progresses at a variable pace: when n additional
customers are in the system, his watch moves (n + 1)
times slower.
Before stating our main result for this subsection we
need the following lemma. It is stated and proved for
ers in the system are
the M / G / 1 case (rather than the simpler M / M / 1 case
dealt with) since the general result is needed in the next
section.
LEMMA 2.1. For an MI G / 1-PS queueing system (under
stationary conditions), let N(r) be the number of customers
(not including the x-customer) in queue as soon as the x-
customer has been servedfor T units, 0 ? T ? x. Then N(T)
follows a geometric distribution with parameter p. Moreover,
the exit process of the other customers, with time measured
by the amount of service the x-customer has already received,
is Markovian with intensity (K&)WN(T).
that this is so also for r E (0, x]. Our proof follows Ross'
proof exactly with the understanding of the increment
tribution of the service requirement of a customer by F
and the corresponding probability density function by
f. If there are n additional customers in the system at
service time T for the x-customer, let y = (yi, ..., yn), y
= y(r), be the vector of their completed workloads (yi
_ . yn), let y-i = (yl, . . ., yi-1, Yi+l, *., yn) and
finally, let yo = (0, y). Clearly, y(r), r E (0, x), is a Mar-
kov process. In particular, the transition rates q(, )
which correspond to a change in the number of custom-
q(y, y-j) - lF(y )
1 - F(yi)
and
q(y, yo) = An.
(The factor n, the number of customers in the system,
on the RHS, is because dr = n-ldt, where t is the real
time.) We claim that p(y), the density of state y is,
n

p(y) = (1 - p)pnn!Kj 1n7 (1 - F(yi)). (9)

The lemma seems paradoxical: while the x customer i=l

is in the system, the service to other customers will be

Indeed, it is easy to check that
slower than at a random time. Therefore, it may seem
that N(Tr) should be stochastically increasing in r. How-
ever, this is not the case, due to the change of time scale. P(Y) - F(y ) = p(y_j)Xf(yj)n, (10
If we visit the system at a random (real) time, then it is
true that we would find more customers due to the pres- and
ence of the x customer. But here we sample the system
p(y)X(n + 1) = P(Yo). (11)
according to the "adaptive" clock of the tagged cus-
tomer. It is more likely that the tagged customer will getThese equations establish that indeed p() is the station-
to the r service point when there are fewer customers ary distribution of the process and that the process is
in the system. This stochasticly balances the effect of the time reversible. For details see Ross (1983). In particu-
slower service. Note that in particular, because of time lar, the marginal distribution of N(r) is geometric with
reversibility, the number of customers when the x- parameter p. Moreover, P({N(Tr + Ar) = N(r)
customer exits behaves as the number of customers - 1} N(T)) = (Kj) -N(r)as required. E
upon his arrival. Now, N(r) is the number of customers
THEOREM 2.3. For the MIMI 1-PS queueing system
at the time of exit if it was a r-customer. Here is the
formal argument.
E(TIx)= P2X, (12)
PROOF. It is well known that N(O) has the stated dis-
tribution, e.g., Ross (1983, p. 172). It remains to prove and

MANAGEMENT SCIENCE/Vol. 44, No. 6, June 1998 855

This content downloaded from 201.234.181.53 on Mon, 19 Feb 2018 22:23:54 UTC
All use subject to http://about.jstor.org/terms
 HAVIV AND RITOV
Externalities, Tangible Externalities, and Queue Disciplines

p22.
Var(E(T Ix)) = P), (K2 - K1). (13) E(C I X) =2(1 K2 )X 2 ' (17)
and
In particular, both E(T I x) and Var(E(T I x)) have the same
values in LCFS-PR and PS M / M / 1 queueing systems.
x4 2
PROOF. We investigate the three terms in (2). Let M* Var(E(C x)) = (K2 - K1)
4(1 - p)2
be the number of customers who receive service before
the tagged customer clears the system. Any customer + - (K4 - K2)
4
that leaves the system as soon as the x-customer receives
r units of service, adds r to the first term on the RHS of A3K2
(2). It follows from Lemma 2.1 that + 2 ) (K3 - K2Kj). (18)
2(1 - p)
M* rx
E Ti =E TjILN(T)dT (14) PROOF. It is well-known that under these disciplines
i=l1 the expected number of customers in the queue when
the x-customer commences service is A2K2 / 2(1 - p). An
- X x~2. (15) expected number of AT are added until he is served for
2(1 p) x
r units of time, 0 - r T x. Hence,
On the other hand, 2MM*+1 Ti is clearly x times the
expected number of customers who leave from a ran- V(T) = X2K2/2(1-p) + XT, O 5 T 5 X.
dom time to the end of the busy period, that is,
The theorem easily follows now by Lemma 3.1. [l

p - x. (16) 3.2. The Case of M / G /1 LCFS-PR Queue

(1 - 2
In a LCFS-PR M-I G /1 queueing system the x-customer
Finally, (12) follows from Equations (2), (15), and finds, upon his arrival, p / (1 - p) customers on the av-
(16). D erage. These are the only other customers in the queue
while he is served. Hence, v(r) = p / (1 - p). This, cou-
pled with Lemma 3.1, immediately implies the follow-
3. The Expected Tangible
ing theorem.
Externalities
In this section we compute E(C I x), the expected tangi- THEOREM 3.2. For a LCFS-PR M / G 1 queue,
ble externalities incurred by an x-customer and the cor-
responding variance Var(E(C I x)). We start with a triv- E(Clx)= Px, (19)
ial result.

and
LEMMA 3.1. For any T, T C (0, x), let v(r) be the ex-
pected number of customers in the queue, excluding the x-
customer, at the time in which the x-customer is servedfor Var(E(C Ix))=p22.
2 2 (K2 Ki) (20)
(1 - p)
his r unit of service. Then,
3.3. The Case of M / G / 1 PS Queue
E(C I x) = f v(T)dT. The following theorem is an immediate conclusion from
Lemma 2.1 and Lemma 3.1.

3.1. The Case of M / G / 1 Queues Without THEOREM 3.3. For a processor sharing MI G / 1 queue,
Preemption

THEOREM 3.1. For any strong work-conserving MI G I E(C Ix) p

- P __x, (21)
1 queue without preemption, in particular for the FCFS,
LCFS-HOL, and random order M / G / 1 queueing systems, and

856 MANAGEMENT SCIENCE/Vol. 44, No. 6, June 1998

This content downloaded from 201.234.181.53 on Mon, 19 Feb 2018 22:23:54 UTC
All use subject to http://about.jstor.org/terms
 HAVIV AND RITOV
Externalities, Tangible Externalities, and Queue Disciplines

they are served on a FCFS basis. Now, a customer with

Var(E(C Ix))= 1 2 (K2 - K1). (22)
(1- p)
In particular, the expected tangible externalities imposed by
an x-customer in a processor sharing and in a LCFS-PR M I
G/ 1 queues coincide.
4. Some Final Remarks
REMARK 1. Note that under M /M / 1, the ratio be-
tween Var(E(C x)) for the FCFS discipline (and the
other nonpreemptive disciplines) and the correspond-
ing variance under the LCFS-PR and the PS disciplines
is 2p2 - 6p + 5 > 1 for any p E (0, 1). This is also the
ratio between the corresponding values of Var(E(T I x)).
Of course, we cannot say which discipline is better since
it depends on the objectives of the society or the queue
operator. However, the ratio can be smaller than 1 un-
der M/ G /1.
REMARK 2. Note that although the tangible external-
ities implied in an FCFS queue by two customers are
additive, the expected tangible externalities are not ad-
ditive. Hence, in a FCFS queue in which the expected
tangible externalities pricing mechanism is applied, an
x-customer has an incentive to disguise himself as, say,
two x / 2-customers. This is not the case under the LCFS-
PR or the PS disciplines in which the expected tangible
externalities are linear functions of the service require-
ment. Finally, note that everything said here applies
verbatim when externalities replace tangible externali-
ties.
REMARK 3. It was found in all examined disciplines
that for any service requirement x, the ratio between thePROOF. Tag a customer in the "original" system as
expected tangible externalities and the expected exter-
nalities equals 1 - p. It is tempting to conjecture that
this is the case in general. However, as our next coun-
terexample shows, this is not the case. Specifically, we
next show a strong work-conserving queue discipline
for which the ratio is not 1 - p. Actually, the ratio in
this example depends on the value of x. Consider an
M / M /1 system that operates as LCFS-PR, as long as
the customer gets less than y units of service for some
fixed value of y. As soon as a customer receives a service
of length y, he gets the lowest priority and is served onlyessarily equal. This is permitted since we calculate the
when there are no other customers in the system. If
there are more than one such customers in queue, then
service requirement larger than y stays in the system an
amount of time of Op (y) (Recall that {Ty} is Op(y) if Ty/y
is uniformly stochasticly bounded, i.e., for any e > 0
there is an M < oo such that supy P( I Ty / y I > M) < e.
It is Op(l) if Ty is uniformly stochasticly bounded.) This
is the case since (i) it takes an expected y/ (1 - p(y))
units of time until he gets his first y units of service,
where p(y) = X[fx0o xf(x)dx + y(l - F(y))] and where
f(x) and F(x) are the density and distribution functions
of the service requirement, and (ii) then he should wait
an additional amount of Op(l) of time until the end of
the busy period. Note, that his residual requirement is
exponential with mean Au-1. On the other hand, such
customers arrive at a rate of Xe-6Y, and hence the prob-
ability that a customer with service requirement greater
that y meets another such customer converges to 0 as
y -- oo. Hence, for large values of y, almost all external-
ities imposed by an x-customer, with x > y, are accu-
mulated while he is in service. In other words, he im-
poses externalities of Op(l) in addition to the Op(y) ex-
ternalities that he imposes while in service. Hence, for
such a customer the ratio between the expected tangible
externalities and the expected externalities converges to
1 as y --* x.
We conclude with the following proposition. Since we
are unable to cite a reference for this claim, we prove it.
PROPOSITION 4.1. Consider an MIMIl queueing sys-
tem with any strong and work conserving entrance discipline.
The expected externalities for an arrival (in steady state)
equals p / (,u(l - p)2).
before. Construct a simulated system that starts upon
the arrival of the tagged customer, with the same num-
ber of other customers as in the original system (but
without the tagged customer). In the new system all
arrival and departure times are the same as in the orig-
inal system, except for the exit of the customer that
leaves the original system empty for the first time after
the arrival of the tagged customer. The two systems
are coupled from the time of this exist. Note that the
service requirements in the two systems are not nec-
expected externalities and in the M / M / 1 queue, the
service time is exponential. Thus we compare the

MANAGEMENT SCIENCE/Vol. 44, No. 6, June 1998 857

This content downloaded from 201.234.181.53 on Mon, 19 Feb 2018 22:23:54 UTC
All use subject to http://about.jstor.org/terms
 HAVIV AND RITOV
Externalities, Tangible Externalities, and Queue Disciplines

expected waiting time in the two systems conditioned
on the initial state, where in one system there is an
extra tag customer.
We compare the total waiting time of all customers
in the systems until the first time that both of them are
empty. Until this time, there is exactly one more cus-
tomer in the original queue than in the simulated one.
Thus, the total waiting time of all the customers in the
2Helpful discussions with I. Adan, G. Mills, and J. van der Wal im-
proved the content of this paper. An associate editor and a referee
helped in the much needed improvement of the presentation.
References
Carlin, A. and R. R. Park, "Marginal Cost Pricing of Airport Runway
Capacity," American Econ. Rev., 60 (1970).
Dewan, S. and H. Mendelson, "User Delay Costs and Internal Pricing
for Service Facility," Management Sci., 36 (1990), 1502-1517.
Dolan, R. J., "Incentive Mechanism for Priority Queues," Bell J. Eco-

original system, is larger than the total waiting time in nomics, 9 (1978), 421-436.
Hassin, R., "On the Optimality of the First Come Last Served Queues,"
the other system, exactly by the length of time until the
Econometrica, 53 (1983), 201-202.
coupling. The expected value of the length of time until
and M. Haviv, "Equilibrium Strategies and the Value of Infor-
the coupling is 1/ (,u(l - p)2). This is the product of the mation in a Two Line Queueing System with Threshold Jockey-
expected number of customer in the original system just ing," Comm. in Statistics -Stochastic Models, 10 (1994), 415-436.

after the arrival of the tagged customer, 1/ (1 - p), and
the the expected time it takes to reduce the queue size
by one, 1/ (,u(l - p)).
However, the externalities imposed by the tagged
customer equal the extra time except the time he is in
the system. Therefore we have to subtract the expected
value of the time the tagged customer is in the system,
1/ (y.(l - p)), from the expected value of the total extra
waiting time. Thus, the expected externalities are 1/
Kleinrock, S., Queueing Systems, Vol 2: Computer Applications, Wiley-
Interscience, New York, 1976.
Loch, C., "Pricing in Markets Sensitive to Delay," Ph.D. Thesis, Stan-
ford University, Stanford, CA, 1991.
Mendelson, H. and S. Whang, "Optimal Incentive Compatible Priority
Pricing for the M/M/ 1 Queue," Oper. Res., 38 (1990), 870-883.
Newbery, D. M., "Pricing and Congestion: Economic Principles Rele-
vant to Pricing Roads," Oxford Rev. of Econ. Policy, 6 (1990), 22-
38.
Ross, S. M., Stochastic Processes, John Wiley, New York, 1983.
Walrand, J. and P. Varaiya, High-Performance Communication Networks,

( -(l - p)2) - 1/(A(l - p)). D2 Morgan Kaufmann Publishers, San Francisco, CA, 1996.

Accepted by Linda V. Green; received in March 1996. This paper has been with the authors 12 months for 2 revisions.

858 MANAGEMENT SCIENCE/Vol. 44, No. 6, June 1998

This content downloaded from 201.234.181.53 on Mon, 19 Feb 2018 22:23:54 UTC
All use subject to http://about.jstor.org/terms