Online machine scheduling with batch setups


 Emery Jones
 2 years ago
 Views:
Transcription
1 Online machine scheduling with batch setups Lele Zhang, Andrew Wirth Department of Mechanical Engineering The University of Melbourne, VIC 3010, Australia Abstract We study a class of scheduling problems with batch setups for the onlinelist and onlinetime paradigms. Jobs are to be scheduled in batches for processing. All obs in a batch start and complete together, and a constant setup is prior to each batch. The obective is to minimize the total completion time of all obs. We primarily consider the special cases of these problems with identical processing times, for which efficient online heuristics are proposed and their competitive performance is evaluated. eyword: online scheduling; batch setups. 1 Introduction This paper considers a class of online scheduling problems with batch setups, where n independent nonpreemptive obs are to be processed on a single machine or one of m identical parallel machines. A machine can process at most one ob at a time. Each ob i must be assigned to a batch, which consists of a set of obs processed consecutively on a machine. A batch setup s is incurred at the start of each batch. The completion time of a ob i is the time when the last ob of the batch that includes ob i completes its processing, that is, the completion time of the batch. The obective is to minimize the total completion time of all the obs. Thus, our problem can be stated as online machine scheduling with sequenceindependent batch setup times in the batch availability model, and also denoted as 1, P m s, F = 1 C i [1,, 3], where F represents the number of families. This sort of problem is motivated by various real life applications in manufacturing areas and storage systems. One application, mentioned in [4, 5], is the logging of tass in a storage system. Log data can be consecutively Corresponding author. Tel.: ; 1
2 written on diss. Each writeon can be considered as a batch of obs and a constant setup for diswrite occurs for each batch. Another example is from parttype production in flexible manufacturing systems [6]. All parttypes must be mounted on a pallet for processing. The obs on the same pallet form a batch and are completed at the same time. A standardized pallet setup time precedes each batch. Previous wor mainly considered the single machine scheduling problem. Coffman et al. [7] provided two main properties of optimal solutions for their batchsizing problem to minimize the total flow time 1 s, F = 1 F i : (i) obs are processed in shortest processing time order, and (ii) batch sizes, the numbers of obs in batches, are in nonincreasing order. Further, they proposed a dynamic programming algorithm which could optimally solve the problem in O(n log n) time. Webster and Baer [8] summarized the results in [7, 9, 10], and for the problem with identical processing times, they showed that, ignoring the integer requirement, the optimal number of batches η = np s 1 and the optimal batch size of the th batch is n η + s(η +1) p s p, where p and s are the common processing and setup times. An online variant of the problem in the presence of release dates while minimizing the total flow time 1 r i, s, F = 1 F i was discussed by Gfeller et al. [4]. They proposed a competitive GREEDY algorithm as well as two lower bounds for the special case with identical processing times. In the online general case, they showed that any online algorithm is at best ( n ɛ)competitive for any ɛ > 0, and any online algorithm without unnecessary idle time cannot be better than ncompetitive. Besides, they introduced an O(n 5 ) dynamic programming algorithm for the offline version of their problem with a fixed ob sequence in [5]. For the parallel machine problem, Cheng et al. [6] studied the offline problem of batching and scheduling simultaneously available obs on identical parallel machines to minimize total completion time. They developed an O(mn (m+1) ) dynamic programming algorithm to solve the general problem. They also showed that the special case of identical processing times reduces to single machine scheduling problem, which can be solved optimally by previous algorithms [9, 11]. The comprehensive review papers [1,, 3, 8] provide general definitions and realistic applications for machine scheduling problems with setups for various machine configurations, performance measures and availabilities. In this paper, we consider the problems of single machine and identical parallel machine scheduling with batch setups for the online list and onlinetime paradigms. More precisely, in the onlinelist paradigm, all obs are available at time zero and are presented one by one in some sequence. Once a ob is presented, an online algorithm must assign it to some machine and a batch immediately, without any the information about any subsequent obs. In the onlinetime paradigm, a ob becomes available at its release time r i (also nown as its arrival time), and an online algorithm may schedule the
3 ob upon its arrival or delay the decisionmaing until a later time. From the nature of scheduling with batch setups, we see that in the onlinetime setting obs processed in a batch must arrive before the start of that batch s processing. Furthermore, once that processing starts, no extra obs can be added to that batch. We evaluate online algorithms in terms of their competitive performance. Let I be a problem instance, A(I) be the obective function value by algorithm A for I and OP T (I) be the optimal offline value. We say A is a ccompetitive algorithm if A(I) OP T (I) c for all I. Furthermore, we say A has a competitive performance ratio of R A if R A = inf{c 1 : c, for all I}. The general notation is listed below. Notation: A I m n s M l B (l) A(I) OP T (I) an online algorithm; an instance of obs; the number of parallel machines; the number of obs; the constant setup requirement, which is incurred before the processing of each batch; the lth machine; the th batch on machine l; the start time of the setup for batch B (l) S (l) n (l) the number of obs in batch B (l) ; J i the ith presented or released ob; C i the completion time of J i, equal to the completion time of the batch B (l) such that i B (l) ; p i the processing time of J i ; r i the release time of J i ; superscript the corresponding optimal value. Table 1 summarizes the the bounds proven in this paper. OnlineList Paradigm Add Table 1 here. Now we consider the problems in the onlinelist paradigm, where obs are presented one by one and an online algorithm has to schedule a ob to a machine and a batch immediately upon its presentation. It is clear that for identical processing times, it is unnecessary to have any idle times in this paradigm. So we shall assume that none of our algorithms for the problems with identical processing times allow idle time. ; 3
4 .1 Lower bounds for a single machine scheduling The following proposition provides a lower bound on the competitive performance ratio for the a single machine scheduling problems with identical processing times and arbitrary processing times, respectively. Proposition 1. No online algorithm is better than competitive for the 1 onlinelist, p i = p, s, F = 1 C i problem. Proof. Two problem instances are constructed to be scheduled by an online algorithm A. If A assigns the first two obs in separate batches, then the third ob shows up. Otherwise, no more ob is presented. Without loss of generality, we assume that s < p. For the instance I with two obs, the A schedule is sj 1 J and so A(I) = (s + p), whereas an optimal offline solution is sj 1 sj and then OP T (I) = 3(s + p). On the other hand, for the instance I with three obs, the A schedule may be either sj 1 sj sj 3 or sj 1 sj J 3 and the obective function value satisfies A(I ) min {6s + 6p, 5s + 7p} = 6s + 6p by the assumption that s < p. An optimal solution for I is sj 1 J sj 3 as s < p, and so OP T (I ) = 4s + 7p. Now we minimize the following expression { } s + 4p 6s + 6p max,, 3s + 3p 4s + 7p and find that R A , where the optimum is achieved when s p = Hence the result follows. 4
5 . P m onlinelist, p i = p, s, F = 1 C i In this section, we consider scheduling obs on m identical parallel machines. First we define a positive integer λ as follows: λ = ηs p, where η is a constant and will be set to different values at various stages of the following discussion. Then, it immediately follows from the definition function that p(λ 1) < sη λp. (1) Now we introduce a simple heuristic for this case and then consider its competitive performance. Heuristic UniformBatchSize (U BS): Set η =. Whenever a new ob is presented, assign it to the machine with the least number of obs. In case of a tie, choose the machine with a smaller index. Every λ obs assigned to the same machine form a batch. (An example of a UBS schedule is given in Figure 1. Add Figure 1 here. Proposition. R UBS = 3 problem. for the P m onlinelist, p i = p, s, F = 1 C i Proof. Before we consider the competitive performance of heuristic U BS, we remind ourselves of some properties of optimal solutions. We recall Lemma 3 of [6] which states that, for the equal processing time problem, there exists an optimal solution for which the difference in the numbers of obs between any two machines is not greater than 1. It is easy to see that, for any input instance, the number of obs processed on some machine in the UBS schedule may be assumed to be the same as that in an optimal solution, since all the machines are identical. In addition, because all obs have identical processing times, we also assume that the obs assigned to some machine in the UBS schedule are processed on the same machine in the optimal solution, and that their relative processing positions are also the same. In the remainder of the proof, we shall mainly consider the scenario of an arbitrary machine l for 1 l m, and similar arguments can be applied to other machines. We let a + 1 and d be the number of batches on machine l and the number of obs in the last batch, respectively. Obviously, the number of obs in a 5
6 batch can never exceed λ. In addition, the first a batches are the same size, and each of them contains exactly λ obs. Therefore, the number of the obs processed on machine l can be written as aλ + d, if a 0. Note that if a = 1, which implies that n < m, then the UBS schedule, which has each ob processed on different machines, is exactly the same as the optimal solution, and then the problem becomes trivial. Thus, without loss of generality, we may assume that a 0 for the rest of the discussion. Now we consider the obs completed in the first a batches and in the last batch, respectively. For batch B (l), 1 a if a 1: The completion time of B (l), the th batch processed on machine l for 1 a and a 1, is (s + λp). Thus the total completion time of the obs in batches B (l) 1,..., B(l) a is given by i B (l),1 a C i = aλ(a + 1)(s + λp). We see that the number of the obs in the first a batches is aλ, and the completion time of the ith ob in this set in an optimal solution is at least s+ip. It follows that the optimal total completion time of these obs satisfies i B (l) C,1 a i aλs + aλp(aλ + 1) = 1 (a λ p + aλp + aλs). Let 1 = 6 i B (l),1 a C i 4 i B (l),1 a C i. By (1) and η =, we have λp p < s λp. It follows that So we obtain 1 a λ(λp s) + aλ(s + p λp) + aλp > 0. For batch B (l) a+1 : i B (l) C,1 a i i B (l),1 a C i < 3. The completion time of batch B (l) a+1 is s(a + 1) + p(aλ + d) and so the total completion time of the obs in the batch is written as: C i = d[s(a + 1) + p(aλ + d)] = aλdp + ads + ds + d p. i B (l) a+1 Recall that we assume that the processing order in the optimal solution is the same as that in the UBS schedule, because all obs are identical. That is 6
7 to say, in the optimal offline schedule, the obs in the batches B (l) 1,..., B(l) a precede the obs in B (l) a+1. Hence, the optimal completion time of the ith ob in B (l) a+1 cannot be smaller than s + p(aλ + i), and then the optimal total completion time of the obs in B (l) a+1 satisfies i B (l) a+1 C i ds + aλdp + dp(d + 1) = 1 (aλdp + ds + d p + dp). Again, we let = 6 i B (l) Ci 4 a+1 i B (l) C i. Since d λ, similar to a+1 the preceding case we have Hence we prove that ad(λp s) + d(s + p dp) + dp > 0. i B (l) a+1 i B (l) a+1 C i C i < 3. It is derived from the above cases that i B (l) i B (l) applicable to any machine, and so we draw the conclusion that n i=1 C i n < 3 i=1 C i. C i C i < 3. This inequality is Furthermore, we see that R UBS 3. Now we consider a lower bound of the competitive performance ratio for U BS. Again, we consider the situation of an arbitrary machine l. First we recall some properties of optimal solutions for the corresponding single machine scheduling problem from [8]: the optimal number of batches η = np s 1 and the size of the th batch is n η + s(η +1) p s p. We can rewrite the number of obs in the th batch as: n = n η s p. Next, we consider an instance for which s = p and n = m(g +G) for a large integer G such that G + G is divisible by 4. Then, by UBS, the parameter λ equals and (G +G) 4 batches are formed and processed on machine l. It follows that the total completion time of the obs on machine l is given by i M l C i = 3s 16 (G + G)(G + G + 4) = 3sG O(G3 ). As for an optimal solution, by the optimal properties, G batches are formed and the number of obs in the th batch is G + 1. So the optimal total 7
8 completion time is obtained by Ci = M l = η n (s + s =1 n ) =1 [ η (G + 1 ) s + s =1 ] (G + 1 ) as η = G, =1 = s η η η (G + 5G + 3) (3G + 4) + =1 =1 =1 3 So we have = sg4 8 + O(G3 ). i M l C i i M l C i 3, as G. The proof is now complete. We note that UBS can be applied to the 1 onlinelist, p i = p, s, F = 1 C i problem as well and, besides, the results of Proposition is also valid for that problem. 3 OnlineTime Paradigm In this section, we adopt the problem in the onlinetime scheduling environment. We recall the statement of the problems in the onlinetime paradigm that any ob in a batch must arrive before that batch s processing starts, and no extra obs can be added to that batch, once its processing starts. In this section, we shall discuss parallel machine scheduling first. We propose a new heuristic for the P m onlinetime, r i, p i = p, s, F = 1 C i and a lower bound of the competitive ratio for the case of m =. Furthermore, we consider the problem of a single machine scheduling, for which we introduce an online heuristic as well as a lower bound. 3.1 P m onlinetime, r i, p i = p, s, F = 1 C i A New Heuristic We remind ourselves of the definition expression of the parameter λ, that is, λ = ηs p. Now we set η = 1 for the following heuristic. 8
9 Heuristic Sync (Sy): Do not schedule until all machines are idle and some obs are available. Let n t be the number of unscheduled obs at some scheduling time t. Start a new batch on each machine, and then assign min{λ, nt m } obs to each of machines 1,..., n t m n t m and assign min{λ, n t m } obs to each of the remaining machines. In case there are more than mλ obs, select the earliest released obs. Remars: Without loss of generality, we may assume that the first released ob arrives at time 0. Suppose that there is an instance I for which the first ob is available after time 0, that is, r 1 > 0. Then we can obtain another instance I which is derived by decreasing the release time of each ob in I by r 1. It is easy to see that the online and the optimum total completion times both decrease by r 1 n from I to I. Thus, Sy(I) OP T (I) < Sy(I ) OP T (I ) Sy, we may assume that r 1 = 0.. As we are interested in the worst performance of We notice that there may exist a scenario in which, at time t, a batch B 1 starts on machine 1 whilst there is no batch that starts on machines l 0,..., m for some l 0 with l 0 m. For convenience, we may add a dummy batch on each of those machines such that n (l) = 0 for l 0 l m. Now we give some preliminaries for the following discussion of Sy s performance. Classifications: Let Ī = {I 1, I,... } be a set of time intervals defined as follows: Machine 1 is never idle during any interval I q and all machines are idle throughout the time period between I q and I q+1. Let BI = {BI 1, BI,... } be a set of time intervals such that a batchinterval BI contains the processing period of B (1), the th batch on machine 1, that is, BI = [S (1), S (1) s + n (1) p], where S (1) and n (1) are the start time and the number of obs of batch B 1. We divide an interval I q into batchgroups in the following way. A batchgroup G q in I q, = 1,,..., ends with a batchinterval BI in which the number of obs is smaller than mλ, or possibly ends with the last batchinterval of I q. Figure gives an example of Sy schedules. 9
10 Add Figure here. Proposition 3. 1 m R Sy for the P m onlinetime, r i, p i = p, s, F = 1 C i problem. Proof. Consider an arbitrary interval I q. For a batchgroup G q we let a + 1 be the number of batches on machine 1 and let h be the index of its first batch. Also, we let N (l) be the total number of obs in the batchgroup on machine l. If a 1, by the definition of batchgroups we see that each of batches 1,..., a on any machine must have λ obs. Hence, N (l) = aλ+n (l) h+a, where n (l) h+a is the number of obs in the last batch of Gq on machine l. Then, the total completion time of the obs in G q on machine l of the Sy schedule is i G q,m l C i = S (1) h N (l) a(a + 1) + λ(s + λp) + n (l) h+a [a(s + λp) + s + n(l) h+a p]. On the other hand, a lower bound for the optimal total completion time of the obs in G q satisfies Ci m r i + s i G q i G q l=1 N (l) + p m l=1 N (l) m l=1 ( N (l) m + 1)( m m l=1 + p m l=1 N (l) m l=1 ( N (l) + 1)(m m m N (l) m l=1 m N (l) m l=1 N (l) ) m m l=1 + m N (l) ). m We notice that the difference between N l 1 and N l for l 1, l with 1 l 1, l m, is at most 1. Also, because all the obs have the same processing time and the machines are all identical, we can rewrite the above lower bound as Ci = i G q m l=1 i G q,m l m l=1 i G q,m l C i [ m r min N (l) l=1 r i + sn (l) + sn (l) + + (l) (l) pn (N + 1) ] + 1), (l) (l) pn (N where r min = min r i for i G q. We let l be the difference between i G q,m l C i and i G q,m l C i for some machine l. By (1) and η = 1, we 10
11 have λp p < s λp. In addition to N (l) = aλ + n (l) h+a, we have l r min N (l) + sn (l) + pn (l) (N (l) n (l) h+a (as + aλp + s + n(l) h+a p) = r min N (l) + sn (l) S (1) h aλ(s + p λp) + > r min N (l) + sn (l) N (l) + aλp + pn(l) h+a S (1) h N (l). + 1) S (1) h N (l) aλ + (λp s)( a λ + an(l) h+a ) (s + λp)(a + 1) If = 1, then r min = S 1 h and thus l > r min N (l) + sn (l) > 0. Otherwise, from the definition of G q as well as heuristic Sy, we see that the earliest released ob in G q must arrive after the start time of B1 h 1. That is to say, r min > Sh 1 1 = S1 h s n1 h 1p. In this scenario, if h 3 or q, then we have r min > Sh 1 1 > s + p. Furthermore, we can obtain l > r min N (l) > r min N (l) + sn (l) S (1) h + sn (l) + (S (1) > (s + p)n (l) n (1) (l) h 1pN (s + p)n (l) λpn (l) N (l) h s n1 h 1 p)n S (1) h > 0, since n (1) h 1 λ and λp < s + p. Therefore, for the cases of = 1 q, h 3 and q, we obtain the following result: N Ci C i = i G q i G q m l > 0. l=1 Thus, for the rest of the proof we need only consider the case of q = 1 and h = for G 1, that is, the situation in which the first batch processed on machine 1 in G 1 is B1. For this case, we consider the first two batchgroups together. Again, we let a + 1 be the number of batches on machine 1 in G 1. Let u and v be the smallest superscripts of batches B(l 1) 1 and B (l ) a+ such that n (l 1) 1 < n (1) 1 and n (l ) a+ < n(1) a+ respectively. If n(l) 1 = n (1) 1 and/or n (l) for all l with 1 l m, then let u = m + 1 and/or v = m + 1. a+ = n(1) a+ We note that there are m possible cases with regard to the values of u and v. We shall analyze the case of n (1) 1 = n () = = n (m) and 11
12 n (1) a+ = = n(m 1) a+ = n (m) a+ + 1, that is, u = and v = m. Then, for the other cases, analogous arguments can lead to the same result as below. By the assumption that the first ob is released at time 0, the total completion time of the obs in the two batchgroups is given by m C i = i G 1 1 G1 = l=1 i G 1 1 G1,M l m l=1 [n (l) 1 C i aλ(a + 1)(s + λp) (s + n(l) 1 p) + aλ(s + n(1) 1 p) + + n (l) a+ (as + aλp + s + n(1) 1 p + n(l) a+ p)]. Let N 1 denote the total number of obs in the two batchgroups. By hypothesis, N 1 = amλ+mn 1 1 +mn1 a+ m. Then the optimal total completion time of these obs satisfies Ci sn 1 + p N 1 m ( N 1 m + 1)(N 1 m N 1 m ) i G 1 1 G1 + p N 1 m ( N 1 m + 1)(m N 1 + m N 1 m ) = m(aλ + n (1) n (1) a+ )[s + p (aλ + n(1) 1 + n (1) a+ )]. We let = i G 1 1 G1 C i i G 1 1 G1 C i. Since λp s and n (1) a+ 1, > amλ(s + λp as λp) + amn (1) a+ (λp s) + (n(1) 1 1)(sm + amλp + mpn (1) a+ ) + mpn (1) 1 mp + pn (1) a+ pn(1) amλ(a 1)(λp s) p(m 1)(n (1) 1 1) aλp + as + s + mpn (1) 1 mp + p pn (1) 1 Consequently, we have shown that i I q C i < i I q Ci R Sy. for all l. Thus Next we provide a lower bound for R Sy. Consider an instance I consisting of m obs with r 1 = 0, r i = ɛ for i m and s < p. The Sy schedule is sj 1 sj (on machine 1) and [idle]sj i (on machine i 1) for 3 i m, where the idle period equals s + p. On the other hand, an optimal solution is sj 1 (on machine 1) and [ɛ]sj i (on machine i) for i m. Thus, Sy(I) = (s + p) + (m 1)(s + p), OP T (I) = (s + p) + (m 1)(ɛ + s + p) m(s + p) as ɛ 0. 1
13 It follows that R Sy Sy(I) OP T (I) 1 m. Now the proof of this proposition is complete. We notice that an obvious variant of heuristic Sy can be applied to the 1 r i, p i = p, s, F = 1 C i problem with a competitive ratio of A Lower Bound for P r i, p i = p, s, F = 1 C i Proposition 4. No online algorithm is better than the P r i, p i = p, s, F = 1 C i problem competitive for Proof. We construct instances based on the scheduling mechanism of an online algorithm A. The constructed instances all start with obs available at time 0 and all have s = p. We let X 1 and X be the start times of J 1 and J by A, respectively. Without loss of generality, we assume that X 1 X. Now we consider the following possible cases. 1. J 1 and J are assigned to the same machine. In this case, there are no further obs. Let the instance of this case be I. Then the total completion time satisfies A(I) (X 1 + s + p) and the optimum is OP T (I) = (s + p). Hence A(I) OP T (I) = X 1 + s + p s + p s + p s + p = 4, as s = p. 3. J 1 and J are scheduled to different machines. Depending on the value of X, either no further ob arrives or two obs arrive at time X + ɛ, where ɛ is a small positive number. Let I denote the instance consisting of two obs, and then we have A(I ) = X 1 + X + (s + p) whereas OP T (I ) = (s + p). Besides, let I be the instance of four obs. For I A can either schedule J 3 and J 4 on the same machine after J 1 or J, or put them on different machines. We remind ourselves that obs processed in a batch must arrive before the start of the batch processing. Thus, J 3, J 4 must be assigned to new batches for r 3 = r 4 = X + ɛ > X X 1. Also, we note that if obs 3 and 4 are placed on the same machine, the total completion time is smaller if they are processed within a batch, since s = p. Thus, A(I ) min{3(x 1 + s + p) + (s + p) + X + s + p, 3(X + s + p) + (s + p) + X 1 + s + p, (X 1 + s + p) + (X + s + p) + (s + p)} min{3x 1 + X + 6s + 8p, X 1 + X + 6s + 6p}, 13
14 by the assumption that X 1 X. An optimal solution of instance I is sj 1 J on M 1 and [X + ɛ]sj 3 J 4, where [X + ɛ] is an idle period for M. Thus, we have OP T (I ) = (s + p) + (X + s + p). The maximum of A(I ) and A(I ) OP T (I ) OP T (I ) { X1 + X + (s + p) max, (s + p) min is greater than or equal to { 3X1 + X + 6s + 8p, X 1 + X + 3s + 3p X + 4s + 8p X + s + 4p Since X 1 0 and s = p, we rewrite the above expression as: { } A(I ) max OP T (I ), A(I ) OP T (I ) { { X + 6p X + 0p max, min 6p X + 16p, X }} + 9p. X + 8p We minimize the maximum, and by some algebra we find that { } A(I ) max OP T (I ), A(I ) +, OP T (I ) 6 }}. where the minimum is achieved when X 1 = 0 and X = ( 4)p. From the discussion above, we see that R A min{ 4 3, + 6 } = 1.115) for any online algorithm A. Hence the result follows. + 6 ( 3. 1 onlinetime, r i, p i = p, s, F = 1 C i Now we turn our attention to the shop type of a single machine. We notice that the online scheduling environment of the case considered in this section is the same as the corresponding problem 1 onlinetime, r i, s, F = 1 F i in [4], except for the obective function. The two obectives of the total completion time and the total flow time are equivalent when all obs are available at the same time or for offline scheduling problem. We cannot say which one is more appropriate, since one may be more realistic in some cases, and may be less in others. As an essential part of our discussion, we consider this 1 onlinetime, r i, p i = p, s, F = 1 C i problem in this section. In [4], the authors proposed a competitive online heuristic for the problem, and also showed that no online algorithm could achieve better results. We shall prove that for 1 onlinetime, r i, s, F = 1 C i this lower bound on the competitive ratio for all online algorithms is 5+1, and further develop a simply implemented heuristic, which guarantees a competitive ratio bounded between 5 3 and
15 3..1 A Lower Bound Proposition 5. No online algorithm is better than the 1 onlinetime, r i, p i = p, s, F = 1 C i problem competitive for Proof. We consider the following scenario. The first ob is released at time 0 with p = ɛs, where ɛ is an arbitrary small number which we shall let tend to zero. Suppose that an online algorithm A allocates the machine to J 1 at time X 1. Depending on the value of X 1, either no further ob arrives or n 1 obs arrive at time X 1 + ɛ. In the latter case, as ɛ 0, A can at best assign the last n 1 obs in one batch immediately succeeding the first batch, whereas an offline optimal solution may have all n obs processed in one batch starting at time X 1 +ɛ. Hence the online total completion time of the n obs is greater than or equal to n(x 1 + s) + (n 1)[s + (n 1)p], whilst the offline optimum is not greater than n(x 1 + ɛ + s + np). A may choose the best value of X 1 to minimize max{ X 1+s+p s+p, n(x 1+s)+(n 1)[s+(n 1)p] n(x 1 +ɛ+s+np) }. Now we let ɛ 0, then p 0 and finally let n. It follows that { X1 + s + p max, n(x } 1 + s) + (n 1)[s + (n 1)p] max s + p { 1 + X 1 s, 1 + s X 1 + s n(x 1 + ɛ + s + np) }. Some algebra shows that the minimum is obtained when X 1 = s( 5 1) and equals 5+1. Therefore, we conclude that R A 5+1 for all online algorithms. 3.. A New Heuristic Now we introduce a new heuristic for the 1 r i, p i = p, s, F = 1 C i problem, which is a variant of heuristic Sy, and then evaluate its performance. Again we recall the definition function of λ, λ = ηs, and further we set η = 5 3 for the following heuristic. Heuristic WaitHalfSetup (W HS): Keep the machine idle until time s. If both the machine and some unscheduled ob(s) are available, then assign the λ earliest released obs to the machine to form a new batch. If the number of available obs is smaller than λ, then schedule all of them to the machine in a new batch. Proposition R W HS 1 + problem. p 3 5 for the 1 r i, p i = p, s, F = 1 C i 15
16 Proof. Let c = η. The proof of this proposition is analogous to that of Proposition 3, and thus we adopt its notation. We assume, without loss of generality, that the first ob is released at time 0. We say a batch is full if it consists of λ obs; otherwise, we say it is nonfull. Then we recall that, in an arbitrary processing interval I q without idle time, batchgroup G q contains a group of batches ending with a nonfull batch or possibly with the last batch of I q. We let a + 1 be the number of batches in G q and let h be the index of the first batch. As we now consider the a single machine scheduling case, we omit the superscript of the symbols for batches. Thus, the number of obs in G q is given by N = aλ + n a+h, where n a+h is the number of obs in the last batch of G q. Furthermore, the total completion time of the obs in G q can be written as: C i = S h N + i G q aλ(a + 1)(s + λp) A lower bound for the optimum satisfies Ci r i + sn + pn (N + 1) i G q i G q where r min = min r i for i G q. + n a+h (as + s + pn ). () r min N + sn + pn (N + 1), (3) With regard to the values of h, and q, we consider the following cases. (A summary of the analysis is given in Table 1??.) We let l be the difference between c i G q Ci and i G q C i in Case l, as set out below. 1. h = 1 ( = 1, q = 1) Add Table 1 here. By the assumption that the first ob arrives at time 0, we have r min = 0 and S 1 = s in this case. By () and (3), the difference 1 satisfies 1 aλ s η λp p(1 + η) + + a η (λp η s ) } {{ } } {{ } δ (1) 1 δ (1) + n a+h a(λp η s) + s η s pn a+h(η 1) + η 16 p(1 + η) η } {{ } δ (1) 3.
17 Now we consider the values of δ (1) 1, δ(1) and δ (1) 3 respectively. If follows from (1) that δ (1) 0 and also δ (1) p(λ 1) 1 > η λp p(1 + η) + η = λp( 1 η 1 ) + + η p(η > 0, since η = 5 3. η ) Finally, we consider δ (1) 3. A batch may contain at most λ obs, and so n a+h λ. Thus, δ 3 must satisfy Consequently, we see that 1 > 0. δ (1) 3 > s η ηs + p η > 0.. h = 3, = and q = 1 In this case, batchgroup G 1 1 contains two batches B 1 and B. By the definitions of W HS and G q, we see that B 1 is full whereas B is nonfull, that is, n 1 = λ, n λ 1, and also the earliest released ob in G 1 must arrive after the start time of B, that is, r min > S = 3s +λp. Thus, we rewrite () and (3) for this case as: i G q i G q C i = ( 5s + λp + n aλ(a + 1)(s + λp) p)n + + n a+h (as + s + pn ) ( 5s + λp p)n aλ(a + 1)(s + λp) + + n a+h (as + s + pn ), Ci ( 3s + λp)n + sn + pn (N + 1). Then we obtain 3. h, = 1 N [ (1 + 1 η )(3s ] + λp) s λp + p ] [ 3s > N η + λp η s p(λ 1) > sn η (3 + η η ) by (1), >
18 In the scenario of = 1 and h, B h is the first batch of I q, and the obs processed in or after B h cannot arrive before time S h and thus the smallest release date of G q satisfies r min S h > 3s. Then the following inequality bounds the value of h 4, 3 S h N (1 + 1 η ) S hn + sn + 1 > 0. When h 4 and 1, we observe that the batch immediately preceding of G q, B h 1, is a nonfull batch and thus we have n h 1 λ 1 and r min > S h s pn h 1 > 5s. Also by ((1)), the difference 4 for this case is bounded below by 5. h =, = and q = 1 4 N (r min + r min η S h + s ) + 1 > N (S h s pn h 1 + 5s η S h + s ) > sn η (5 η η ) > 0. From this case onwards, we shall consider the batchgroups G q 1 and G q together instead of Gq only. That is to say, in this case we consider G 1 1 and G1 consisting of the first a + h batches. The online total completion time is C i = (N 1 + N )( 3s + pn 1) i G 1 1 G1 + aλ(a + 1)(s + λp) + n a+h (as + s + pn ). Now we provide two lower bounds for the optimal total completion time of the obs in G 1 1 and G1. If the optimal solution starts all the obs in G 1 3s 1 at or after time, then a lower bound for the optimum is given by LB (5) 1 = 3s(N 1 + N ) + p(n 1 + N )(N 1 + N + 1). Otherwise, if the optimal solution starts processing some of the obs in G 1 3s 1 before time, then it must start another batch for the processing of the obs in G 1. Thus, another lower bound is LB (5) = s(n 1 + N ) + sn + p(n 1 + N )(N 1 + N + 1). 18
19 If LB (5) 1 LB (5), then the difference 5 between c i G 1 1 G1 C i and i G 1 C 1 G1 i satisfies 5 3s(N 1 + N ) η pn 1 = N 1 η + aλ + p(n 1 + N )(N 1 + N + 1) η aλ(as + s + λp) n a+h (as + s + n a+hp ) p(n + η + 1) + (pn 1 pηn 1 + 3s ) } {{ } δ (5) 1 3s η s λp + p + p η + a(λp η s) } {{ } δ (5) + n a+h 3s η s pn a+h(η 1) + η p(1 + η) η } {{ } δ (5) 3 + p(n 1 + N ) +a( λp η s). We consider the signs of δ (5) 1, δ(5) and δ (5) 3. Since n a+h λ and N 1 = n h 1 λ 1, by (1) we see that Thus, we obtain 5 > 0. δ (5) 1 > s(3 + η η ) > 0, δ (5) > s η (3 η η ) > 0, δ (5) 3 > s η (3 η η ) > 0. Otherwise, if LB (5) 1 > LB (5), then we have 6. h = 3, = and q = 5 s(1 + η)(n N 1 ) η > N 1(δ (5) 1 sη s) η > sn 1( η ) η >
20 As for Case 5, we consider the obs in G 1 and G. For q =, there must be machine idle time preceding G 1. Also, by W HS we see that the earliest released ob in G 1 arrives at S the start time of B and the earliest released ob in G must arrive after time S. We let r min be the smallest release time of the obs in G 1 and G, and then we have r min S > 3s + pn 1. Thus we can write the online total completion time and a lower bound for the optimum as: aλ(s + λp)(a + 1) C i = (S + s + pn 1 )(N 1 + N ) + i G 1 G i G 1 G + n a+h (as + s + pn ), Ci (r min + s)(n 1 + N ) + p(n 1 + N )(N 1 + N + 1). Since r min S > 3s + pn 1, the difference 6 satisfies 6 (N 1 + N )(r min S + r min s ) + 5 > 0. η 7. h = 3, = 3 and q = 1 In this case, G 1 and G1 3 consist of B and B 3,..., B a+3, respectively. From the values of h, and q, we see that G 1 1 and G1 both contain a nonfull batch, namely, N 1 = n 1 λ 1 and N = n λ 1. This implies that the obs in the batchgroups G 1 and G1 3 must arrive after time s 3s and + pn, respectively. Following observations similar to those in Case 5, we can write the online total completion time of the obs in G 1 and G1 3 and two lower bounds for the optimum as: C i = (N 1 + N )( 5s + pn aλ(a + 1)(s + λp) + pn 1 ) + i G 1 G1 3 i G 1 G1 3 C i min + n a+h (as + s + pn ), (N 1 + N )( 5s + pn ) + p(n 1 + N )(N 1 + N + 1), } {{ } LB (7) 1 (N 1 + N )( 3s + pn ) + sn + p(n 1 + N )(N 1 + N + 1), } {{ } 0 LB (7)
21 where LB (7) 1 and LB (7) are defined similar to those in Case 5. We note that the term pn (N 1 + N ) of LB (7) comes from the processing requirements of the obs in G 1 1. Since we may assume that obs are processed in order of nondecreasing release dates in an optimal solution, the obs in G 1 and G1 3 must be processed after those in G1 1. An argument analogous to that in Case 5 leads to the same result that the difference 7 between c i G 1 G1 C 3 i and i G 1 C G1 i is greater 3 than 0. We conclude from the above arguments that the inequality c i G q C Gq i +1 i G q C Gq i > 0 is valid for all h, q and. Hence we obtain c n +1 i=1 C i n i=1 C i > 0. Now we consider the following instance I to determine a lower bound for the competitive ratio R W HS. The first ob of I is available at time 0 and n 1 obs arrives at time s + ɛ. We let the positive number ɛ 0 and then the identical processing time p 0. The online and the optimal offline obective function values are given below: W HS(I) = 3s (n 1)(5s + np) + p +, n(3s + ɛ + np) OP T (I) =. Let ɛ 0, then p 0 and finally n, and it follows that Hence the result follows. W HS(I) OP T (I) Conclusions and Future Wor In this paper we discuss four problems of scheduling machines with a common batch setup in the batch availability model. For both the onlinelist and the onlinetime scenarios, we introduce lower bounds and new online heuristics to each case with equal processing requirements. Furthermore, we establish the lower and upper bounds on the competitive performance ratio for each heuristic. Table summarizes the the bounds proven in this paper. Future wor can extend the above discussion in various ways. With tighter lower bounds on the optimal obective function values or better defined properties of the optimal (offline) solutions, can the competitive ratios of the proposed heuristics be improved? Also, given the general case with arbitrary processing times, is there any online algorithm which can guarantee rcompetitive for any r? Moreover, issues with other obectives lie the minimization of total weighted flow time or the total weighted completion time remain to be approached in online scheduling environments. 1
22 References [1] A. Allahverdi, J. N. Gupta, T. Aldowaisan, A review of scheduling research involving setup considerations, Omega, The International Journal of Management Science 7 (1999) [] A. Allahverdi, C. T. Ng, T. C. E. Cheng, M. Y. Kovalyov, A survey of scheduling problems with setup times or costs, European Journal of Operational Research 187 (008) [3] C. N. Potts, M. Y. Kovalyov, Scheduling with batching: a review, European Journal of Operational Research 10 (000) [4] B. Gfeller, L. Peeters, B. Weber, P. Widmayer, Online single machine batch scheduling, LNCS 416 (006) [5] B. Gfeller, L. Peeters, B. Weber, P. Widmayer, Single machine batch scheduling with release times, Institute of Theoretical Computer Science, ETH Zurich Technical Report 514 (006) 1 3. [6] T. C. E. Cheng, Z. L. Chen, M. Y. Kovalyov, B. M. T. Lin, Parallelmachine batching and scheduling to minimize total completion time, IIE Transactions 8 (1996) [7] E. G. Coffman, M. Yannaais, M. J. Magazine, C. Santos, Batch sizing and ob sequencing on a single machine, Annals of Operations Research 6 (1990) [8] S. Webster, K. R. Baer, Scheduling groups for obs on a single machine, Operations Research 43 (4) (1995) [9] D. F. Shallcross, A polynomial algorithm for a one machine batching problem, Operations Research Letters 11 (199) [10] C. Santos, M. J. Magazine, Batching in single operation manufacturing systems, Operations Research Letters 4 (1985) [11] E. G. Coffman, A. Nozari, M. Yannaais, Optimal scheduling of products with two subassembles on a single machine, Operations Research Letters 37 (3) (1989)
An improved online algorithm for scheduling on two unrestrictive parallel batch processing machines
This is the PrePublished Version. An improved online algorithm for scheduling on two unrestrictive parallel batch processing machines Q.Q. Nong, T.C.E. Cheng, C.T. Ng Department of Mathematics, Ocean
More informationSingle Machine Batch Scheduling with Release Times
Single Machine Batch Scheduling with Release Times Beat Gfeller Leon Peeters Birgitta Weber Peter Widmayer Institute of Theoretical Computer Science, ETH Zurich Technical Report 514 pril 4, 2006 bstract
More informationResearch Article Batch Scheduling on TwoMachine Flowshop with MachineDependent Setup Times
Hindawi Publishing Corporation Advances in Operations Research Volume 2009, Article ID 153910, 10 pages doi:10.1155/2009/153910 Research Article Batch Scheduling on TwoMachine Flowshop with MachineDependent
More informationOffline sorting buffers on Line
Offline sorting buffers on Line Rohit Khandekar 1 and Vinayaka Pandit 2 1 University of Waterloo, ON, Canada. email: rkhandekar@gmail.com 2 IBM India Research Lab, New Delhi. email: pvinayak@in.ibm.com
More informationJUSTINTIME SCHEDULING WITH PERIODIC TIME SLOTS. Received December May 12, 2003; revised February 5, 2004
Scientiae Mathematicae Japonicae Online, Vol. 10, (2004), 431 437 431 JUSTINTIME SCHEDULING WITH PERIODIC TIME SLOTS Ondřej Čepeka and Shao Chin Sung b Received December May 12, 2003; revised February
More informationSingle machine parallel batch scheduling with unbounded capacity
Workshop on Combinatorics and Graph Theory 21th, April, 2006 Nankai University Single machine parallel batch scheduling with unbounded capacity Yuan Jinjiang Department of mathematics, Zhengzhou University
More informationOptimal Onlinelist Batch Scheduling
Optimal Onlinelist Batch Scheduling Jacob Jan Paulus a,, Deshi Ye b, Guochuan Zhang b a University of Twente, P.O. box 217, 7500AE Enschede, The Netherlands b Zhejiang University, Hangzhou 310027, China
More informationA binary search algorithm for a special case of minimizing the lateness on a single machine
Issue 3, Volume 3, 2009 45 A binary search algorithm for a special case of minimizing the lateness on a single machine Nodari Vakhania Abstract We study the problem of scheduling jobs with release times
More informationScheduling Single Machine Scheduling. Tim Nieberg
Scheduling Single Machine Scheduling Tim Nieberg Single machine models Observation: for nonpreemptive problems and regular objectives, a sequence in which the jobs are processed is sufficient to describe
More informationClassification  Examples
Lecture 2 Scheduling 1 Classification  Examples 1 r j C max given: n jobs with processing times p 1,...,p n and release dates r 1,...,r n jobs have to be scheduled without preemption on one machine taking
More informationThe Goldberg Rao Algorithm for the Maximum Flow Problem
The Goldberg Rao Algorithm for the Maximum Flow Problem COS 528 class notes October 18, 2006 Scribe: Dávid Papp Main idea: use of the blocking flow paradigm to achieve essentially O(min{m 2/3, n 1/2 }
More informationCompletion Time Scheduling and the WSRPT Algorithm
Completion Time Scheduling and the WSRPT Algorithm Bo Xiong, Christine Chung Department of Computer Science, Connecticut College, New London, CT {bxiong,cchung}@conncoll.edu Abstract. We consider the online
More informationarxiv:1112.0829v1 [math.pr] 5 Dec 2011
How Not to Win a Million Dollars: A Counterexample to a Conjecture of L. Breiman Thomas P. Hayes arxiv:1112.0829v1 [math.pr] 5 Dec 2011 Abstract Consider a gambling game in which we are allowed to repeatedly
More informationDuplicating and its Applications in Batch Scheduling
Duplicating and its Applications in Batch Scheduling Yuzhong Zhang 1 Chunsong Bai 1 Shouyang Wang 2 1 College of Operations Research and Management Sciences Qufu Normal University, Shandong 276826, China
More informationDimensioning an inbound call center using constraint programming
Dimensioning an inbound call center using constraint programming Cyril Canon 1,2, JeanCharles Billaut 2, and JeanLouis Bouquard 2 1 Vitalicom, 643 avenue du grain d or, 41350 Vineuil, France ccanon@fr.snt.com
More informationA simple criterion on degree sequences of graphs
Discrete Applied Mathematics 156 (2008) 3513 3517 Contents lists available at ScienceDirect Discrete Applied Mathematics journal homepage: www.elsevier.com/locate/dam Note A simple criterion on degree
More informationThe EpsilonDelta Limit Definition:
The EpsilonDelta Limit Definition: A Few Examples Nick Rauh 1. Prove that lim x a x 2 = a 2. (Since we leave a arbitrary, this is the same as showing x 2 is continuous.) Proof: Let > 0. We wish to find
More information2.3 Scheduling jobs on identical parallel machines
2.3 Scheduling jobs on identical parallel machines There are jobs to be processed, and there are identical machines (running in parallel) to which each job may be assigned Each job = 1,,, must be processed
More informationCost Model: Work, Span and Parallelism. 1 The RAM model for sequential computation:
CSE341T 08/31/2015 Lecture 3 Cost Model: Work, Span and Parallelism In this lecture, we will look at how one analyze a parallel program written using Cilk Plus. When we analyze the cost of an algorithm
More informationMinimizing the Number of Machines in a UnitTime Scheduling Problem
Minimizing the Number of Machines in a UnitTime Scheduling Problem Svetlana A. Kravchenko 1 United Institute of Informatics Problems, Surganova St. 6, 220012 Minsk, Belarus kravch@newman.basnet.by Frank
More informationThe Conference Call Search Problem in Wireless Networks
The Conference Call Search Problem in Wireless Networks Leah Epstein 1, and Asaf Levin 2 1 Department of Mathematics, University of Haifa, 31905 Haifa, Israel. lea@math.haifa.ac.il 2 Department of Statistics,
More information3. Mathematical Induction
3. MATHEMATICAL INDUCTION 83 3. Mathematical Induction 3.1. First Principle of Mathematical Induction. Let P (n) be a predicate with domain of discourse (over) the natural numbers N = {0, 1,,...}. If (1)
More informationWORSTCASE PERFORMANCE ANALYSIS OF SOME APPROXIMATION ALGORITHMS FOR MINIMIZING MAKESPAN AND FLOWTIME
WORSTCASE PERFORMANCE ANALYSIS OF SOME APPROXIMATION ALGORITHMS FOR MINIMIZING MAKESPAN AND FLOWTIME PERUVEMBA SUNDARAM RAVI, LEVENT TUNÇEL, MICHAEL HUANG Abstract. In 1976, Coffman and Sethi conjectured
More informationR u t c o r Research R e p o r t. A Method to Schedule Both Transportation and Production at the Same Time in a Special FMS.
R u t c o r Research R e p o r t A Method to Schedule Both Transportation and Production at the Same Time in a Special FMS Navid Hashemian a Béla Vizvári b RRR 32011, February 21, 2011 RUTCOR Rutgers
More informationOnline Scheduling for Cloud Computing and Different Service Levels
2012 IEEE 201226th IEEE International 26th International Parallel Parallel and Distributed and Distributed Processing Processing Symposium Symposium Workshops Workshops & PhD Forum Online Scheduling for
More informationTHE SCHEDULING OF MAINTENANCE SERVICE
THE SCHEDULING OF MAINTENANCE SERVICE Shoshana Anily Celia A. Glass Refael Hassin Abstract We study a discrete problem of scheduling activities of several types under the constraint that at most a single
More informationScheduling Realtime Tasks: Algorithms and Complexity
Scheduling Realtime Tasks: Algorithms and Complexity Sanjoy Baruah The University of North Carolina at Chapel Hill Email: baruah@cs.unc.edu Joël Goossens Université Libre de Bruxelles Email: joel.goossens@ulb.ac.be
More informationA class of online scheduling algorithms to minimize total completion time
A class of online scheduling algorithms to minimize total completion time X. Lu R.A. Sitters L. Stougie Abstract We consider the problem of scheduling jobs online on a single machine and on identical
More informationDouble Sequences and Double Series
Double Sequences and Double Series Eissa D. Habil Islamic University of Gaza P.O. Box 108, Gaza, Palestine Email: habil@iugaza.edu Abstract This research considers two traditional important questions,
More informationApproximability of TwoMachine NoWait Flowshop Scheduling with Availability Constraints
Approximability of TwoMachine NoWait Flowshop Scheduling with Availability Constraints T.C. Edwin Cheng 1, and Zhaohui Liu 1,2 1 Department of Management, The Hong Kong Polytechnic University Kowloon,
More informationThe Trip Scheduling Problem
The Trip Scheduling Problem Claudia Archetti Department of Quantitative Methods, University of Brescia Contrada Santa Chiara 50, 25122 Brescia, Italy Martin Savelsbergh School of Industrial and Systems
More informationIntroduction to Scheduling Theory
Introduction to Scheduling Theory Arnaud Legrand Laboratoire Informatique et Distribution IMAG CNRS, France arnaud.legrand@imag.fr November 8, 2004 1/ 26 Outline 1 Task graphs from outer space 2 Scheduling
More informationClassification  Examples 1 1 r j C max given: n jobs with processing times p 1,..., p n and release dates
Lecture 2 Scheduling 1 Classification  Examples 11 r j C max given: n jobs with processing times p 1,..., p n and release dates r 1,..., r n jobs have to be scheduled without preemption on one machine
More information14.1 Rentorbuy problem
CS787: Advanced Algorithms Lecture 14: Online algorithms We now shift focus to a different kind of algorithmic problem where we need to perform some optimization without knowing the input in advance. Algorithms
More informationOnline Scheduling with Bounded Migration
Online Scheduling with Bounded Migration Peter Sanders, Naveen Sivadasan, and Martin Skutella MaxPlanckInstitut für Informatik, Saarbrücken, Germany, {sanders,ns,skutella}@mpisb.mpg.de Abstract. Consider
More informationPh.D. Thesis. Judit NagyGyörgy. Supervisor: Péter Hajnal Associate Professor
Online algorithms for combinatorial problems Ph.D. Thesis by Judit NagyGyörgy Supervisor: Péter Hajnal Associate Professor Doctoral School in Mathematics and Computer Science University of Szeged Bolyai
More informationChapter 4, Arithmetic in F [x] Polynomial arithmetic and the division algorithm.
Chapter 4, Arithmetic in F [x] Polynomial arithmetic and the division algorithm. We begin by defining the ring of polynomials with coefficients in a ring R. After some preliminary results, we specialize
More informationApproximation Algorithms
Approximation Algorithms or: How I Learned to Stop Worrying and Deal with NPCompleteness Ong Jit Sheng, Jonathan (A0073924B) March, 2012 Overview Key Results (I) General techniques: Greedy algorithms
More information20 Selfish Load Balancing
20 Selfish Load Balancing Berthold Vöcking Abstract Suppose that a set of weighted tasks shall be assigned to a set of machines with possibly different speeds such that the load is distributed evenly among
More informationTheory of Computation Prof. Kamala Krithivasan Department of Computer Science and Engineering Indian Institute of Technology, Madras
Theory of Computation Prof. Kamala Krithivasan Department of Computer Science and Engineering Indian Institute of Technology, Madras Lecture No. # 31 Recursive Sets, Recursively Innumerable Sets, Encoding
More information1. R In this and the next section we are going to study the properties of sequences of real numbers.
+a 1. R In this and the next section we are going to study the properties of sequences of real numbers. Definition 1.1. (Sequence) A sequence is a function with domain N. Example 1.2. A sequence of real
More informationNotes 11: List Decoding Folded ReedSolomon Codes
Introduction to Coding Theory CMU: Spring 2010 Notes 11: List Decoding Folded ReedSolomon Codes April 2010 Lecturer: Venkatesan Guruswami Scribe: Venkatesan Guruswami At the end of the previous notes,
More informationEMBEDDING COUNTABLE PARTIAL ORDERINGS IN THE DEGREES
EMBEDDING COUNTABLE PARTIAL ORDERINGS IN THE ENUMERATION DEGREES AND THE ωenumeration DEGREES MARIYA I. SOSKOVA AND IVAN N. SOSKOV 1. Introduction One of the most basic measures of the complexity of a
More informationDynamic TCP Acknowledgement: Penalizing Long Delays
Dynamic TCP Acknowledgement: Penalizing Long Delays Karousatou Christina Network Algorithms June 8, 2010 Karousatou Christina (Network Algorithms) Dynamic TCP Acknowledgement June 8, 2010 1 / 63 Layout
More informationOPTIMAL CONTROL OF FLEXIBLE SERVERS IN TWO TANDEM QUEUES WITH OPERATING COSTS
Probability in the Engineering and Informational Sciences, 22, 2008, 107 131. Printed in the U.S.A. DOI: 10.1017/S0269964808000077 OPTIMAL CONTROL OF FLEXILE SERVERS IN TWO TANDEM QUEUES WITH OPERATING
More informationAn example of a computable
An example of a computable absolutely normal number Verónica Becher Santiago Figueira Abstract The first example of an absolutely normal number was given by Sierpinski in 96, twenty years before the concept
More informationApproximation Algorithms. Scheduling. Approximation algorithms. Scheduling jobs on a single machine
Approximation algorithms Approximation Algorithms Fast. Cheap. Reliable. Choose two. NPhard problems: choose 2 of optimal polynomial time all instances Approximation algorithms. Tradeoff between time
More informationOptimal Online Preemptive Scheduling
IEOR 8100: Scheduling Lecture Guest Optimal Online Preemptive Scheduling Lecturer: Jir Sgall Scribe: Michael Hamilton 1 Introduction In this lecture we ll study online preemptive scheduling on m machines
More informationMechanisms for Fair Attribution
Mechanisms for Fair Attribution Eric Balkanski Yaron Singer Abstract We propose a new framework for optimization under fairness constraints. The problems we consider model procurement where the goal is
More informationBatch Scheduling of Deteriorating Products
Decision Making in Manufacturing and Services Vol. 1 2007 No. 1 2 pp. 25 34 Batch Scheduling of Deteriorating Products Maksim S. Barketau, T.C. Edwin Cheng, Mikhail Y. Kovalyov, C.T. Daniel Ng Abstract.
More informationBargaining Solutions in a Social Network
Bargaining Solutions in a Social Network Tanmoy Chakraborty and Michael Kearns Department of Computer and Information Science University of Pennsylvania Abstract. We study the concept of bargaining solutions,
More informationThe Relative Worst Order Ratio for OnLine Algorithms
The Relative Worst Order Ratio for OnLine Algorithms Joan Boyar 1 and Lene M. Favrholdt 2 1 Department of Mathematics and Computer Science, University of Southern Denmark, Odense, Denmark, joan@imada.sdu.dk
More informationIN THIS PAPER, we study the delay and capacity tradeoffs
IEEE/ACM TRANSACTIONS ON NETWORKING, VOL. 15, NO. 5, OCTOBER 2007 981 Delay and Capacity TradeOffs in Mobile Ad Hoc Networks: A Global Perspective Gaurav Sharma, Ravi Mazumdar, Fellow, IEEE, and Ness
More informationMATH10040 Chapter 2: Prime and relatively prime numbers
MATH10040 Chapter 2: Prime and relatively prime numbers Recall the basic definition: 1. Prime numbers Definition 1.1. Recall that a positive integer is said to be prime if it has precisely two positive
More informationHOMEWORK 5 SOLUTIONS. n!f n (1) lim. ln x n! + xn x. 1 = G n 1 (x). (2) k + 1 n. (n 1)!
Math 7 Fall 205 HOMEWORK 5 SOLUTIONS Problem. 2008 B2 Let F 0 x = ln x. For n 0 and x > 0, let F n+ x = 0 F ntdt. Evaluate n!f n lim n ln n. By directly computing F n x for small n s, we obtain the following
More informationFACTORING POLYNOMIALS IN THE RING OF FORMAL POWER SERIES OVER Z
FACTORING POLYNOMIALS IN THE RING OF FORMAL POWER SERIES OVER Z DANIEL BIRMAJER, JUAN B GIL, AND MICHAEL WEINER Abstract We consider polynomials with integer coefficients and discuss their factorization
More informationDEGREES OF ORDERS ON TORSIONFREE ABELIAN GROUPS
DEGREES OF ORDERS ON TORSIONFREE ABELIAN GROUPS ASHER M. KACH, KAREN LANGE, AND REED SOLOMON Abstract. We construct two computable presentations of computable torsionfree abelian groups, one of isomorphism
More informationCHAPTER II THE LIMIT OF A SEQUENCE OF NUMBERS DEFINITION OF THE NUMBER e.
CHAPTER II THE LIMIT OF A SEQUENCE OF NUMBERS DEFINITION OF THE NUMBER e. This chapter contains the beginnings of the most important, and probably the most subtle, notion in mathematical analysis, i.e.,
More informationSubsets of Euclidean domains possessing a unique division algorithm
Subsets of Euclidean domains possessing a unique division algorithm Andrew D. Lewis 2009/03/16 Abstract Subsets of a Euclidean domain are characterised with the following objectives: (1) ensuring uniqueness
More informationPricing of Limit Orders in the Electronic Security Trading System Xetra
Pricing of Limit Orders in the Electronic Security Trading System Xetra Li Xihao Bielefeld Graduate School of Economics and Management Bielefeld University, P.O. Box 100 131 D33501 Bielefeld, Germany
More informationSome Polynomial Theorems. John Kennedy Mathematics Department Santa Monica College 1900 Pico Blvd. Santa Monica, CA 90405 rkennedy@ix.netcom.
Some Polynomial Theorems by John Kennedy Mathematics Department Santa Monica College 1900 Pico Blvd. Santa Monica, CA 90405 rkennedy@ix.netcom.com This paper contains a collection of 31 theorems, lemmas,
More informationMIPBased Approaches for Solving Scheduling Problems with Batch Processing Machines
The Eighth International Symposium on Operations Research and Its Applications (ISORA 09) Zhangjiajie, China, September 20 22, 2009 Copyright 2009 ORSC & APORC, pp. 132 139 MIPBased Approaches for Solving
More informationList Scheduling in Order of αpoints on a Single Machine
List Scheduling in Order of αpoints on a Single Machine Martin Skutella Fachbereich Mathematik, Universität Dortmund, D 4422 Dortmund, Germany martin.skutella@unidortmund.de http://www.mathematik.unidortmund.de/
More information8.1 Min Degree Spanning Tree
CS880: Approximations Algorithms Scribe: Siddharth Barman Lecturer: Shuchi Chawla Topic: Min Degree Spanning Tree Date: 02/15/07 In this lecture we give a local search based algorithm for the Min Degree
More informationOn the kpath cover problem for cacti
On the kpath cover problem for cacti Zemin Jin and Xueliang Li Center for Combinatorics and LPMC Nankai University Tianjin 300071, P.R. China zeminjin@eyou.com, x.li@eyou.com Abstract In this paper we
More informationWeek 1: Introduction to Online Learning
Week 1: Introduction to Online Learning 1 Introduction This is written based on Prediction, Learning, and Games (ISBN: 2184189 / 2184189 CesaBianchi, Nicolo; Lugosi, Gabor 1.1 A Gentle Start Consider
More informationMathematical Induction
Chapter 2 Mathematical Induction 2.1 First Examples Suppose we want to find a simple formula for the sum of the first n odd numbers: 1 + 3 + 5 +... + (2n 1) = n (2k 1). How might we proceed? The most natural
More informationa 11 x 1 + a 12 x 2 + + a 1n x n = b 1 a 21 x 1 + a 22 x 2 + + a 2n x n = b 2.
Chapter 1 LINEAR EQUATIONS 1.1 Introduction to linear equations A linear equation in n unknowns x 1, x,, x n is an equation of the form a 1 x 1 + a x + + a n x n = b, where a 1, a,..., a n, b are given
More informationSection 3 Sequences and Limits, Continued.
Section 3 Sequences and Limits, Continued. Lemma 3.6 Let {a n } n N be a convergent sequence for which a n 0 for all n N and it α 0. Then there exists N N such that for all n N. α a n 3 α In particular
More informationPrime Numbers. Chapter Primes and Composites
Chapter 2 Prime Numbers The term factoring or factorization refers to the process of expressing an integer as the product of two or more integers in a nontrivial way, e.g., 42 = 6 7. Prime numbers are
More information(Refer Slide Time: 01.26)
Discrete Mathematical Structures Dr. Kamala Krithivasan Department of Computer Science and Engineering Indian Institute of Technology, Madras Lecture # 27 Pigeonhole Principle In the next few lectures
More informationLecture 4: BK inequality 27th August and 6th September, 2007
CSL866: Percolation and Random Graphs IIT Delhi Amitabha Bagchi Scribe: Arindam Pal Lecture 4: BK inequality 27th August and 6th September, 2007 4. Preliminaries The FKG inequality allows us to lower bound
More informationRonald Graham: Laying the Foundations of Online Optimization
Documenta Math. 239 Ronald Graham: Laying the Foundations of Online Optimization Susanne Albers Abstract. This chapter highlights fundamental contributions made by Ron Graham in the area of online optimization.
More informationLecture 4 Online and streaming algorithms for clustering
CSE 291: Geometric algorithms Spring 2013 Lecture 4 Online and streaming algorithms for clustering 4.1 Online kclustering To the extent that clustering takes place in the brain, it happens in an online
More informationMATH10212 Linear Algebra. Systems of Linear Equations. Definition. An ndimensional vector is a row or a column of n numbers (or letters): a 1.
MATH10212 Linear Algebra Textbook: D. Poole, Linear Algebra: A Modern Introduction. Thompson, 2006. ISBN 0534405967. Systems of Linear Equations Definition. An ndimensional vector is a row or a column
More informationLoad balancing of temporary tasks in the l p norm
Load balancing of temporary tasks in the l p norm Yossi Azar a,1, Amir Epstein a,2, Leah Epstein b,3 a School of Computer Science, Tel Aviv University, Tel Aviv, Israel. b School of Computer Science, The
More informationModule MA1S11 (Calculus) Michaelmas Term 2016 Section 3: Functions
Module MA1S11 (Calculus) Michaelmas Term 2016 Section 3: Functions D. R. Wilkins Copyright c David R. Wilkins 2016 Contents 3 Functions 43 3.1 Functions between Sets...................... 43 3.2 Injective
More informationA REMARK ON ALMOST MOORE DIGRAPHS OF DEGREE THREE. 1. Introduction and Preliminaries
Acta Math. Univ. Comenianae Vol. LXVI, 2(1997), pp. 285 291 285 A REMARK ON ALMOST MOORE DIGRAPHS OF DEGREE THREE E. T. BASKORO, M. MILLER and J. ŠIRÁŇ Abstract. It is well known that Moore digraphs do
More informationMathematical Induction
Mathematical Induction Victor Adamchik Fall of 2005 Lecture 2 (out of three) Plan 1. Strong Induction 2. Faulty Inductions 3. Induction and the Least Element Principal Strong Induction Fibonacci Numbers
More informationChapter 6. Number Theory. 6.1 The Division Algorithm
Chapter 6 Number Theory The material in this chapter offers a small glimpse of why a lot of facts that you ve probably nown and used for a long time are true. It also offers some exposure to generalization,
More information8.1 Makespan Scheduling
600.469 / 600.669 Approximation Algorithms Lecturer: Michael Dinitz Topic: Dynamic Programing: MinMakespan and Bin Packing Date: 2/19/15 Scribe: Gabriel Kaptchuk 8.1 Makespan Scheduling Consider an instance
More informationResource Allocation with Time Intervals
Resource Allocation with Time Intervals Andreas Darmann Ulrich Pferschy Joachim Schauer Abstract We study a resource allocation problem where jobs have the following characteristics: Each job consumes
More informationOPRE 6201 : 2. Simplex Method
OPRE 6201 : 2. Simplex Method 1 The Graphical Method: An Example Consider the following linear program: Max 4x 1 +3x 2 Subject to: 2x 1 +3x 2 6 (1) 3x 1 +2x 2 3 (2) 2x 2 5 (3) 2x 1 +x 2 4 (4) x 1, x 2
More information1 Approximating Set Cover
CS 05: Algorithms (Grad) Feb 224, 2005 Approximating Set Cover. Definition An Instance (X, F ) of the setcovering problem consists of a finite set X and a family F of subset of X, such that every elemennt
More informationApplied Algorithm Design Lecture 5
Applied Algorithm Design Lecture 5 Pietro Michiardi Eurecom Pietro Michiardi (Eurecom) Applied Algorithm Design Lecture 5 1 / 86 Approximation Algorithms Pietro Michiardi (Eurecom) Applied Algorithm Design
More informationEnergyEfficient Mobile Data Transport via Online MultiNetwork Packet Scheduling
EnergyEfficient Mobile Data Transport via Online MultiNetwork Packet Scheduling Aaron Coté, Adam Meyerson, Brian Tagiku Department of Computer Science University of California, Los Angeles Los Angeles,
More informationWhy? A central concept in Computer Science. Algorithms are ubiquitous.
Analysis of Algorithms: A Brief Introduction Why? A central concept in Computer Science. Algorithms are ubiquitous. Using the Internet (sending email, transferring files, use of search engines, online
More informationCONTINUED FRACTIONS AND PELL S EQUATION. Contents 1. Continued Fractions 1 2. Solution to Pell s Equation 9 References 12
CONTINUED FRACTIONS AND PELL S EQUATION SEUNG HYUN YANG Abstract. In this REU paper, I will use some important characteristics of continued fractions to give the complete set of solutions to Pell s equation.
More informationIrreducibility criteria for compositions and multiplicative convolutions of polynomials with integer coefficients
DOI: 10.2478/auom20140007 An. Şt. Univ. Ovidius Constanţa Vol. 221),2014, 73 84 Irreducibility criteria for compositions and multiplicative convolutions of polynomials with integer coefficients Anca
More informationLecture 6: Approximation via LP Rounding
Lecture 6: Approximation via LP Rounding Let G = (V, E) be an (undirected) graph. A subset C V is called a vertex cover for G if for every edge (v i, v j ) E we have v i C or v j C (or both). In other
More informationOptimized Asynchronous Passive MultiChannel Discovery of BeaconEnabled Networks
t t Technische Universität Berlin Telecommunication Networks Group arxiv:1506.05255v1 [cs.ni] 17 Jun 2015 Optimized Asynchronous Passive MultiChannel Discovery of BeaconEnabled Networks Niels Karowski,
More informationFairness in Routing and Load Balancing
Fairness in Routing and Load Balancing Jon Kleinberg Yuval Rabani Éva Tardos Abstract We consider the issue of network routing subject to explicit fairness conditions. The optimization of fairness criteria
More informationLoad Balancing. Load Balancing 1 / 24
Load Balancing Backtracking, branch & bound and alphabeta pruning: how to assign work to idle processes without much communication? Additionally for alphabeta pruning: implementing the youngbrotherswait
More informationInformation Theory and Coding Prof. S. N. Merchant Department of Electrical Engineering Indian Institute of Technology, Bombay
Information Theory and Coding Prof. S. N. Merchant Department of Electrical Engineering Indian Institute of Technology, Bombay Lecture  17 ShannonFanoElias Coding and Introduction to Arithmetic Coding
More informationON THE FIBONACCI NUMBERS
ON THE FIBONACCI NUMBERS Prepared by Kei Nakamura The Fibonacci numbers are terms of the sequence defined in a quite simple recursive fashion. However, despite its simplicity, they have some curious properties
More informationNotes on Chapter 1, Section 2 Arithmetic and Divisibility
Notes on Chapter 1, Section 2 Arithmetic and Divisibility August 16, 2006 1 Arithmetic Properties of the Integers Recall that the set of integers is the set Z = f0; 1; 1; 2; 2; 3; 3; : : :g. The integers
More informationFactors to Describe Job Shop Scheduling Problem
Job Shop Scheduling Job Shop A work location in which a number of general purpose work stations exist and are used to perform a variety of jobs Example: Car repair each operator (mechanic) evaluates plus
More informationQuotient Rings and Field Extensions
Chapter 5 Quotient Rings and Field Extensions In this chapter we describe a method for producing field extension of a given field. If F is a field, then a field extension is a field K that contains F.
More informationNotes on Factoring. MA 206 Kurt Bryan
The General Approach Notes on Factoring MA 26 Kurt Bryan Suppose I hand you n, a 2 digit integer and tell you that n is composite, with smallest prime factor around 5 digits. Finding a nontrivial factor
More informationFrequency Capping in Online Advertising
Frequency Capping in Online Advertising Niv Buchbinder Moran Feldman Arpita Ghosh Joseph (Seffi) Naor July 2, 2014 Abstract We study the following online problem. There are n advertisers. Each advertiser
More information