Constructing Transaction Serialization Order for Incremental. Data Warehouse Refresh. Ming-Ling Lo and Hui-I Hsiao. IBM T. J. Watson Research Center

Size: px
Start display at page:

Download "Constructing Transaction Serialization Order for Incremental. Data Warehouse Refresh. Ming-Ling Lo and Hui-I Hsiao. IBM T. J. Watson Research Center"

Transcription

1 Construting Transation Serialization Order for Inremental Data Warehouse Refresh Ming-Ling Lo and Hui-I Hsiao IBM T. J. Watson Researh Center July 11, 1997 Abstrat In typial pratie of data warehouse, the warehouse data is kept in a separate site from the operation data. Changes to the operation data to propagated to the warehouse site periodially, usually by shipping the new log entries in the operation site. When the operation site is a parallel a parallel DBMS, it is neessary to arrange the log entries into an oherent order for the warehouse data to maintain onsisteny and integrity. This paper solves the problem by presenting a simple method for onstruting a global transation order out of a set of loal logs of a parallel or distributed system. For serial database systems, transation order an easily be inferred from loal database logs. For parallel or distributed database systems, sine eah loal log les has only partial information of eah transation, this problem beomes diulty. In this paper, we present a simple method that determines the order of transations by examining only the relative position of their vote and ommit entries, and needs not looking into what data items eah transation aesses. Based on this mehanism, we present a method that eiently onstruts a global transation order. Coneptually, our method rst onnets the vote and ommit symbols of loal log les into a network. Then it traverse the network to detet inonsisteny in the ordering of ommit symbols in dierent log. It then resolves the inonsisteny by adjusting the position of ommit symbols in the network. Our method has the following features. First, it examines only the vote and ommit entries of transations, but not the data aessing entries. Seond, for vote or ommit 1

2 entries, only their relative positions and their transation ID (to identify from whih transation an entry omes) requires attention. No other information reorded in the vote or ommit entry needs be examined. Finally, it has low omplexity. For a set of loal logs with l totally, only O(l) ost is required both in terms of time and spae. 1 Introdution Data warehouses [Inm96] has beome an inreasingly important faility for deision support and data mining appliations. In suh appliations, information are gather from operation sites for study and analysis. While it is possible to install the warehouse data on the same site as the operation databases, the typial pratie in data ware housing is to separate the warehouse data from operation databases [CD97]. Change on the operation side are later shipped or propagated to the warehouse at onvenient time to reet new information gathered. One reason for doing this is that operational databases are typially used by mission ritial OLTP appliations and any degradation in OLTP performane ould potentially redue ompany's revenue and inome. A seond reason is that OLTP and deision support appliations have very dierent funtional and performane harateristis, and it is very diult to optimizing a system to satisfy both types of workload. One key issue in an environment where warehouse site and operation site is separated is how to propagate eiently the hanges in the operational database to the warehouse database. Copying the omplete operational database to the warehouse site is apparently ineient. Another solution is to propagate the update from operational database to warehouse database ontinuously to keep the warehouse database losely up-to-date. The ARM mehanism proposed by Bhide et al. [BGHJ92], for example, an be applied in data warehouse environment. In most of the data warehousing environment, however, absolute up-todate information is not required. The most ost eetive approah therefore is to propagate the updates to warehouse database periodially, whih an be ahieved by shipping the (possibly ompressed) log reords of operation site and apply them to warehouse database. When the operational database is on a parallel or distributed database system, as is inreasingly ommon, the problem beomes more omplex. The operation site and ware- 2

3 house site may have dierent number of nodes, and/or dierent data partitioning. Unless the operation site and the warehouse site has exatly the same number of nodes and the same data partitioning, applying the parallel logs in the operational site to the warehouse independently may result in inonsisteny and violate data integrity. For instane, on operation site a reord an be reated, moved to another node, and deleted there. When propagating logs from operation site to the warehouse site, if the operation logs are not applied in orret order, it is possible that the delete will be enountered rst. It is therefore neessary to ombine the logs from the operation site into a oherent log before shipping to the warehouse site. This paper solves this problem by presenting an eient method for onstruting a global transation order out of a set of loal logs [MHLP91, Gra78] in a parallel or distributed database system. In addition to failitating warehouse update, onstruting global transation order has the added benet of providing transation exeution history in a distributed or parallel database systems. Suh information is often the objet of mining itself, and an also be a part of historial data. For serial database systems, transation order an easily be inferred from loal database logs. For parallel or distributed database systems, sine eah loal log les has only partial information, this problem beomes diulty. Some systems reord additional information in the log entries that an be used for global transation order onstrution, suh as virtual timestamps [Hsi94]. However, not all systems have the help of additional information. In this paper, we present a method that onstrut a global transation order from the loal logs of a parallel or distributed database system by examining the patterns in the log along. It does not require modiation to the log entry format or to the logging subsystem. Speially, our method onstruts a global transation order by examining only the relative pattern of the vote symbols and ommit symbols in the loal logs, and enfores an ordering of \transation A before transation B" when there is a relative pattern of C A? V B? C B in some loal log, where C X and V X is the ommit and vote symbols of transation X, respetively. Based on our basi mehanism for ordering two arbitrary transations, we present a method that eiently onstruts a global transation order by sanning eah eah loal 3

4 log les only one. Coneptually, our method rst onnets the vote and ommit symbols of the loal logs into a network. Then it resolves ommit symbol yles using our ommit symbol sliding mehanism, and merge the network into a serial sequene. Our method has the following features: 1. It onstruts global transation orders by examining the relative position of the vote and ommit symbols alone. There is no need to look into what data items eah transation has aessed. 2. For eah vote or ommit symbol, we only need to examine its transation ID. No other information reorded in the log entries needs be examined. For example, we don't need to hek the reorded timestamps, if any. 3. We requires no speial treatment of the oordinators of transations. This simplies the algorithm. 4. We only need to read eah loal log les one, and an thus onstrut transation order very eiently. To the best of our knowledge, this is the rst paper to oer solution to suh problem. The rest of the paper is organized as follows. Setion 2 outlines the general operating environment and desribed the problem we try to solve. In setion 3, we introdue the basi mehanism used in ordering transations. Setion 4 presents our main algorithm and Setion 5 disusses the omplexity of our algorithm. Finally, Setion 6 onludes this paper. 2 Preliminaries We assume a database system that (1) is a parallel or distributed database system, (2) uses two-phase ommit protool [BHG87], and (3) uses two-phase loking protool [BHG87]. In this system, eah node has a loal log, and reords in the log the operations that are invoked on the node. For eah transation there is a oordinator and a set of subordinates, possibly on dierent nodes, working on the transation's behalf. The oordinator initiates and 4

5 manages the transation. The subordinates aess the database and perform omputation. Note that oordinators of dierent transations may reside on dierent nodes. Many variants of the two-phase ommit protool have been proposed, inluding the presumed abort and presumed ommit [MLO86], and the early prepare [SC90] protools. In general, these protools use the two-phase loking protool as their onurreny ontrol protool. Also, when a variant of the two-phase ommit protool is used, in addition to the operational log entries, the oordinator and partiipants of a transation also reord in their respetive loal logs the following transation management entries: in its loal log [BHG87]: Coordinator: 1. start-2pc, 2. Commit or Abort 3. End/Forget Transation. Partiipants: 1. Vote: yes/no, 2. Commit or Abort. Dierent variants of the protools may reord dierent subsets of the above entries in the log, depending on the aspets of performane they try to optimize. However, for most of the variants, the following statement is true: in the absene of failure, or, if there is a failure, after the failure is reovered, for eah ommitted transation there is a vote entry and a ommit entry reorded in eah partiipant log, and vote entries preedes the ommit entry. Our analysis and method apply to any two-phase ommit protool variant that has the above property. The problem we propose to solve is to obtain a global transation order from a set of loal logs, so that the global order does not onit the transation order implied by the loal nodes. Our design goals are: (1) To inur as little overhead as possible in onstruting the ordering. (2) To support the onstruting of both total and partial global order of transation. (3) The onstruted global order must be able to be used in onstruting a global log. Note that to solve this problem, we need only look at the ommitted transations, and need not examine log entries written on behalf of aborted transations at all. 5

6 2.1 Serialization Order In the ourse of our desription, we also draw upon the following onepts: Serialization Order: Given a set of objets, we an order them into a total order (a sequene) or a partial order. When the set of objets are transations, we all this order a "serialization order" of the set of transations. The order in whih two transations appear in the total or partial order is also alled the "serialization order" of the two transations. Serialization Requirement: In a DBMS, the system may perform its operations in ertain way, so that it is only meaningful to view one transation as preeeding another. In this ase we say that the DBMS requires one transation be serialized before another transation. This is alled the "serialization requirement" of the two transations by the system. In other words, the serialization requirement between two transations is the order in whih two transations should be serialized. For example, in a DBMS using the two-phase loking protool, if two transations aess a ommon data item, the database system will grant the lok to the data item to one transation rst then the other. In this ase there is a serialization requirement between the two transations. Serialization Impliation: A database log reords the operations that are performed on its assoiated node. The sequene and type of operations, together with the semantis of the system in whih they are exeuted, arry information regarding the serialization requirement by the system. This information is alled the "serialization impliation" of the log le. 2.2 Alternatives The most obvious solutions to this problem are either to examine the timestamp information for eah transation, or the data items aessed by eah transations. All suh information is reorded in the loal logs. However, global timestamps are not always available in every DBMS, and examining data items aessed by eah transation requires examining almost all log entries and performing ompliated analysis. 6

7 Another possibility is to examine the order of ommit entries in the loal logs. However, the ommit entries alone does not provide enough information for onstruting a global transation order. In a parallel or distributed database system, a database node logs its ommit entries in the order it proesses them, instead of in the order imposed by the system serialization requirements. Therefore, even if the ommit of transation A is logged before the ommit of transation B in some log, any one of the following situations an still happen: 1. A must be serialized before B. 2. B must be serialized before A. 3. There is no serialization requirement between A and B. The reason is that the ommit entry of one transation may preeed that of another simply beause it is proessed earlier, and the two transations may not have aess any ommon data item at all. Also, the ommit entries of the same two transations may appear in different order in dierent log les, even when there exists a serialization requirement between them. This happens when dierent nodes in the parallel or distributed system experiene dierent ommuniation delays. A third alternative is to order transation by the order of their vote entries. This alternative relies on (1) partiipants gaining no new loks after voting, and (2) partiipants to synhronize themselves with the oordinators before vote. That is, it requires partiipants, after nishing their jobs, to send a \done" message to the oordinator, and the oordinator, upon reeiving suh messages, sends out a \prepare to ommit" message to all partiipants. However, in pratie a oordinator an ombine the last request to a partiipant and and the \prepare to ommit" ommand in one message to optimize on performane. This is partiularly useful for short update-only transations where eah partiipant performs only one write. In general, the order of ommit or vote entries alone does not always provide enough information to onstrut a global serialization order. In the following setion, we present a method that examines a partiular pattern of vote and ommit entries, whih leads to a simple implementation for onstruting global transation order. 7

8 3 Basi Ordering Method Our method uses a mehanism that examining only the pattern of the vote and ommit entries of loal logs to onstrut a global transation order. We assume that the implementation of the two-phase loking protool is suh that a transation annot aquire loks on a node after it asts its vote on that node. This requirement is implied by most two-phase ommit protool variants and is in fat the way most database systems are implemented. It is stated here expliitly for larity. When used with two-phase ommit protools, the two-phase loking protool an in general release loks in two ways. The rst possibility is to release read loks after the vote, and write loks after the ommit point. This is what is desribed or implied by most two-phase ommit protool literature [GR93, MLO86, SC90]. The seond possibility is that both read loks and write loks are released after the ommit point. This is in fat the strit two-phase loking protool [BHG87]. Although this alternative is more restritive in allowing onurreny, many real-life systems, inluding IBM DB2/PE, implemented this lok release strategy. The reason for this hoie is twofold. First, if the ommit operation is very eient, it may not be ost eetive to release read loks before ommit and write lok afterwards [BHG87]. Seond, it is diult to detet the end of the voting phase of ommit, sine one partiipant may invoke another as part of the voting phase. A partiipant may be alled to do additional work after that partiipant has voted to ommit the transation. This situation is alled rejoining the transation [GR93]. In the following disussions, we will rst present our analysis assuming strit two-phase loking, and disuss the impliation of releasing read loks after vote later. We denote the vote and ommit entries for a transation A as V A and C A, and for two entries E 1 and E 2 in the same log, use E 1? E 2 to mean that E 1 preedes E 2. Note that E 1? E 2 only means E 1 omes before E 2 in the sequene. It does not mean E 1 is the symbol immediately before E 2. We say E 1? E 2 is a relative pattern (pattern for short) onerning E 1 and E 2. A seond subsript is used to denote the node (or log) at whih an operation ours. Therefore, C A1 means the ommit operation of transation A at node (or log) 1, and V B2 means the vote operation of transation B at node (or log) 2. 8

9 Our basi mehanism to onstrut a global transation order is as follows: For any two transations A and B, if there is a relative pattern C A? V B? C B, we order A before B in our global transation order. No other pattern ditates our global transation order. 3.1 Ordering under Strit Two-Phase Loking First we assume strik two-phase loking. This mehanism works orretly beause of the following theorems. Theorem 1: If there is a serialization requirement of A before B by the system, then there is a relative pattern of C A? V B? C B in at least one loal log. Proof: Under two-phase loking protool, a serialization requirement of transation A before B means transation A and B aess a ommon data item D in some node, and A aquire the lok on the data item before B does. Sine transation B annot aess any more data item after the vote V B is asted, it must aquire the lok on and aess D before its vote. Beause the system uses strik two-phase loking, A releases the lok to D only after its ommit. Thus the vote of B must follow the ommit of A. That is, C A? V B. 2 Theorem 2: For any two transations A and B, if the relative pattern C A? V B? C B appears in some loal log, the pattern C B? V A? C A annot appear in any other loal logs. Proof: We use E 1! E 2 to denote that the event E 1 is written to log ours before the event E 2 is written to log. Note the denition does not require E 1 and E 2 to belong to the same log le. When E 1 and E 2 do belong to the same the log, E 1! E 2 and E 1? E 2 are equivalent. The two-phase ommit protool guarantees that for any given transation, the votes at all nodes are written to log before the ommit at any node is written to log. This is beause 9

10 the oordinator must ollet all votes before it sends out its deision to any node. Thus, V T X! C T Y for any transation T and any pair of nodes X and Y. Suppose there is a pattern C AL? V BL in some log L. For an arbitrary log X, we have V AX! C AL! V BL! C BX. This in turn means V AX! C BX and V AX? C BX. Thus the vote symbol of A must preeed the ommit symbol of B in any log. 2 Note that two transations A and B may not aess any ommon data item and still have the pattern C A? V B? C B on a log by hane. However, from the theorems above, we see that the mehanism has the following properties: No miss: If there is a serialization requirement of A before B by the system, there is a pattern of C A? V B? C B in at least one loal log, thus our mehanism will identify it. No fault: If there is a relative pattern C A? V B? C B in some loal logs, then either the serialization requirement between transations A and B is A before B, or there is no serialization requirement at all. Either way, it is safe to order A before B. No hange: If we see C A? V B? C B in one loal log, we will never see C B? V A? C A in any other loal log. Thus, we never need to hange the global order one it is determined. 3.2 Ordering under Early Read Lok Release When read loks are release after vote, vote and ommit pattern annot orret identify all serialization requirements. To see this, onsider two transation A and B. Transation A, among other things, aess data item D. After A votes and releases the read lok to D, transation B aquires write lok to D, writes to D, ommits and releases the write lok to D. At this point A reahes its ommit point. The pattern for A and B will be V A? V B? C B? C A. Though there is a serialization requirement A? B, our method does not identify it. That is, under early release of read loks, examining vote and ommit pattern annot identify read-write onits. However, it an still identify write-read and write-write onits. 10

11 Theorem 3: In a system where the write loks of a transation are released after its ommit point, if there is a serialization requirement of A before B by the system due to write-read or write-write onit, then there is a relative pattern of C A? V B? C B in at least one loal log. Proof: The proof is similar to that of Theorem 3.1, exept that now A release its write lok after C A, and B must aquire the oniting share lok or write lok before its vote. 2 Although our method does not produe an ordering that does not reet all serialization impliation in the logs, for the purpose of warehouse database updates, the transation ordering produed will still result in orret warehouse databases. There are two reasons. First, our method arranges all updates to a partiular data item in the same order as they ourred during their original exeution. Most importantly, in our ordering the nal write to every data item is the same as in the original exeution. Seond, when apply write operations to the warehouse database opy, the arguments of write operations are taken from what is reorded in the log, instead of determined by previous reads as in ordinary exeution. Thus the ontents to be written to the warehouse database the write operations are always the same as that written to the operation database. Missing read-write onits is therefore immaterial in the data warehouse updates. 4 Transation Ordering Algorithm Based on the above theorem, algorithms an be designed to onstrut global transation orders by examining only the vote and ommit symbols of transation partiipants. At a rst look, one an devise a simple algorithm as follows: Consider eah transation as a node in a graph, then examine the loal log les. For eah pattern of C A? V B? C B deteted, add an direted edge from T A to T B. When all logs are sanned and edges added, sort the (ayli) graph into a sequene. On loser examinations, however, suh an straightforward algorithm involves extremely 11

12 high spae and time omplexity, and is thus impratial to implement. Assuming there are n transations, sine the number of serialization requirements between transations is O(n 2 ), to maintain the edges expliitly we need O(n 2 ) spae, and at least O(n 2 ) time. With large number of transations, the quadrati spae and time omplex will be prohibitive expensive. To overome this problem, instead of maintaining ordering information expliitly, we adopt the following general strategy in designing our algorithm: 1. We use the ommit symbols as the marker of transation order. That is, transation A is onsideration to be ordered before transation B if its ommit entry appears before that of B in some log. 2. In ases where ommit symbols of a set of transations have inonsistent order in dierent logs, we adjust the positions of ommit symbols in some logs to make all log having the same order for the set of transations. Eventually this will results in all logs having onsistent orders of ommit symbols. 3. The loal logs an then be merged easily to produe a global order. Suh an approah has the following advantages advantages. First, There is no need for maintaining large expliit data strutures. Seond, although the order of ommit symbols does not guarantee serialization requirements, in pratie, for the majority for the transations, espeially when the system is not heavily loaded, the order of ommit symbols does reet their serialization requirements. We thus only need to proess those ases when the ommit symbol positions ontradit the serialization requirements. Details of the algorithm is desribed in the following setions. 4.1 Adjusting Symbol Positions The key steps of this approah are the identiation of inonsisteny in ommit symbol ordering, and adjusting the positions of ommit symbols to ahieve ordering onsisteny. 12

13 L1 Ci Vj Cj L2 Cj Ci vote or ommit symbol Figure 1: Example of a ommit symbol yles involve log les L 1 and L 2. Inonsisteny in ommit symbol ordering in dierent logs reets in ommit symbol yles. That is, inonsisteny ours when there are ommit symbol yles in the logs and vie versa. The ation of adjusting ommit symbol positions to make the order of ommit symbols onsistent on all loal logs amounts to resolving ommit symbol yles. Formally, ommit symbol yles are dened as follows: Denition: Given a set of loal logs, if there exist some ommits symbols C 1 ; C 2 ; :::; C n?1 ; C n, suh that relative patterns C 1? C 2 ; C 2? C 3 ; :::; C n?1? C n ; C n? C 1 exist in some loal logs, we say that there is ommit symbol yle or simply a yle onsisting of C 1 ; C 2 ; :::; C n. Eah of C k? C k+1, for 1 <= k <= n? 1, and C n? C? 1 is alled a segment of the yle. Figure 1 shows an example of a two-segment ommit symbol yle. There are two logs involved in this yles. On log L 1, C i? C j forms a segment of the yle. On log L 2, C j? C i is the other segment. Denition: Let C i? C j in some loal log be a segment of some yle. If relative pattern C i? V j? C j exist in the log, we say the segment is diretly supported by serialization impliation of the log. In the example in Figure 1, sine there is a pattern of C i?v j?c j on log L 1, the segment C i? C j is a supported segment. When a ommit symbol yle is identied, we must adjust the positions of the ommit 13

14 L1 Ci Vj Cj L2 Ci Cj vote or ommit symbol Figure 2: Example of a ommit symbol yles resolved. symbols on some unsupported segment to resolve the yle and make the transation order onsistent in all logs. Sine the relative patterns of vote and ommit symbols imply serialization impliations, adjusting positions of ommit symbols may hange the relative patterns and reate or destroy the serialization impliations of a loal log. The ation of adjusting ommit symbols must therefore be done in a disiplined and well-ontrolled fashion. We adhere to the following priniples when adjusting symbol positions: 1. The serialization impliation of any loal log should be hanged as little as possible. 2. If a piee of serialization impliation is deleted, it must still exist in some other logs. The total serialization impliation in the set of loal logs is never redued. 3. If a piee of serialization impliation is reated, no other log an have oniting serialization impliation The Slide Operation We use the bakward slide (or slide) operation to ahieve a disiplined way of adjusting ommit symbols. The slide operation only moves, or slides, ommit symbols. Further, it slides a ommit symbol only in in the bakward diretion along the same log, and never slides a ommit symbol past its orresponding vote symbol. Figure 2 shows the ommit symbol yle resolved by slide C i on L2 to the position immediate before C j. 1 This priniple is neessary sine we try to onstrut a total order out of a set of loal logs whih implies only a partial order, and thus inevitably need to reate serialization impliation beyond the system's serialization requirements. 14

15 The slide operation has an important property: it only reates serialization impliations, but never destroys them. During the slide operation, the ommit symbol being moved an slide past some other ommit symbols, or the vote symbols of some other transations. From the theorems we proved above, it is easy to see that swithing the order of any two adjaent ommits does not hange any serialization impliation. Lemma 1: For any log, swithing the order of any two adjaent ommits (or two adjaent votes) does not hange the serialization impliation of the log. Proof: We have shown that serialization impliation between two transation A and B is arried only by the pattern C AX? V BX? C BX or C BX? V AX? C AX. That is, if there is a serialization impliation between A and B, their ommit symbols annot be adjaent. Thus swithing two adjaent ommit annot destroy a serialization impliation. Sine the ation of swithing does not introdue a vote symbol between two ommits, it annot reate a serialization impliation either. 2 On the other hand, swithing the order of two adjaent vote and ommit symbols will either reate or destroy a serialization impliation in a log. However, if the swith is moving a ommit symbol that was originally positioned after a vote symbol to the position in front, then no serialization impliation an be destroyed. Lemma 2: The ation of sliding a ommit symbol bakward does not delete any serialization impliation. Proof: Sine a serialization impliation between transation A and B is reeted by the relative pattern C A? V B? C B, it an be delete only if we move C A to positions before V B, whih the slide operation never does. 2 In the following setions, we will disuss how to detet ommit symbol yles, and how to resolve them using the slide operation. 15

16 L1 v v L2 v v L3 v v merge point Figure 3: Example of a set of loal logs onneted into a network. 4.3 Deteting Commit Symbol Cyles To detet ommit symbol yles, rst we onnet all loal logs into a network at ommit symbols. The vote and ommit symbols of transations are onsidered nodes in the network: 1. For every log, eah symbol within the log is onsidered onneted to the symbols immediately before and after the symbol. 2. Eah transation is represented by a merge point, whih has a wait list and an auxiliary support list assoiated with it. The wait list is essential for yles detetion. The support list makes yles resolution more eient. 3. If a transation T has vote and ommit symbols in multiple logs L 1 ; L 2 ; :::; L k, we onnet all these ommit symbols to the merge point of T, and enter L 1 ; L 2 ; :::; L k into the wait list of T. The vote symbols of the orresponding are also onnet to the merge point, and is reorded in the support list. The vote symbols are not onsidered part of the merge point, are is reorded for eieny in yle resolution. Figure 3 show an example of a network onneted from three loal logs. To detet ommit symbol yles, we need to tranverse all logs, from beginning to end. Eah log is onsidered to have a urrent traversal point to keep trak of its urrent position in traversal. The traversal of eah log is in either the stalled or the ative state. The basi idea for merge point is that one the traversal of a log reahes a merge point, it must stall 16

17 L1 v v L2 v m1 v m2 L3 v m3 v merge point urrent traversal point target point Figure 4: A snapshot of log traversal. and wait for all other logs onneted to the merge point to also reah the point. The details of yle detetion are as follows: 1. Pik an arbitrary log L in ative state and advane its urrent traverse point till a merge point is enountered. 2. Delete votes from the support list of their orresponding merge points as they are enounter. This step is not related to yle identiation, but is useful in yle resolution later. 3. When the traversal point reahes a merge point C m : (a) Delete L from the wait list of C m. (b) If the wait list is non-empty, put L into the stalled state. If the wait list is empty, return all logs onneted to C m to the ative state. 4. If all logs are ompletely traversed, end proessing. Otherwise, if there is at least one log that is not stalled, go to step If all logs are in stalled state, at least one ommit symbol yle is enountered. Identify and resolve the yle. 6. Go to step 2. 17

18 Figure 4 show a snapshot during traversing the network presented in gure 3. Note that merge point m 1 and m 2 are bloking eah other, and yles resolution is required before traversal an ontinue. When it happens that all logs stall at some merge points, we know that eah log L must have stopped at some merge point C m (its urrent traversal point) with a non-empty wait list. Further, for eah log L w in the wait list of C m, we know that C m must lie beyond the urrent traversal point of L w (otherwise L w would have been deleted from the wait list of C m ). We would need to wait till L w reahes C m, before traversal of L an proeed again. We all C m a target point of L w (see Figure 4). Whenever a log has a target point, it means that some other logs are waiting for the log to ome up to its target point before the other logs proeed further. At the time when all logs are stalled, the urrent traversal point of eah log must onnet to some merge point whih in turn onnets to the target points of some other logs (see Figure 4). However, not every log has a target point when all logs are stalled. Also, the urrent traversal point and the target point of a log annot be the same, beause that would mean a transation is ommitted twie on a single database node. 4.4 Cyle Resolution When all logs stall, we know that ommit symbol yles are enountered. We need not, however, atually identify the segments of the yles before we an resolve the yles. The basi idea is to nd a merge point C m so that: if C m is involved in ommit symbol yles with some other merge point C x, then all segments C x?c m are unsupported segments. We will then resolve all these yles, by sliding C m to the position immediately in front of C x. Using our urrent framework, this an be easily done by nding a merge point suh that for every target point C ml on log L that is onneted to this merge point, V ml does not lie between the urrent traverse point of of L and C ml. We all this an unsupported merge point. Finding an unsupported merge point is easy: we simple nd a merge point 18

19 with empty support list. Next we prove there will be at least one suh merge point. Theorem 4: When all log traversal stops, there is at least one merge point C m suh that for every target points C ml on log L that is onneted to this merge point, V ml does not lie between the urrent traverse point of of L and C ml. Proof: If a merge point C m does not satisfy the desribed ondition, that means there is a pattern C X?V ml?c ml on some log L, where C X is another merge point. This in turn means that transation orresponding to C m, T m is serialized after T X orresponding to C x. If all merge points do not satisfy the aforementioned ondition, that means the orresponding transation of every merge point is serialized after the transation orresponding to some other merge point, whih is impossible given the original exeution is serializable. 2 One an unsupported merge point is identied. We slide every target point C ml on log L that is onneted to this merge point to the position before the urrent traversal point of L, and make it the urrent traversal point of L. After this operation, there is at least a merge point all of whose ommit symbols are un-bloked, and traversal an be ontinue on logs with these ommit symbols. As stated earlier, when slide operation is performed, we may slide a ommit symbol past a vote symbol (of some other transation), whih may reate a serialization impliation. However, it an be shown that, given our way of traversing the logs, the serialization impliation reated will never ontradit serialization impliations already exist. Theorem 5: Slide a target point C ml of log L to the position immediately before the merge point C x of L will never reate serialization impliations ontraditing existing ones. Proof: Suppose before the slide, there is a vote symbol V yl of a this transation between C xl and C ml, i.e., a relative pattern C xl? V yl? C ml. After we will reate a new serialization impliation of T m? T y, by reating the pattern C ml? C xl? V yl? C yl. For this to be 19

20 ontradition, there must be a relative pattern C yk? V mk? C mk on some other log K. There are two possibilities: (1) C mk is a urrent traversal point, or (2) C mk is a target point, and V mk lies before the urrent traversal point C zk of log K. If C mk is a urrent traversal point, then C yk would have being traversed before we reah C mk. But on log L, V yl and hene C yl lie after the urrent traversal point C xl. This means that traverse on log K would have been bloked at C yk, and C mk annot possibly be the urrent traversal point of log K. If C mk is a target point, and V mk lies before the urrent traversal point C zk of log K, then will be a pattern C yk? V mk? C zk. Again, this is impossibly sine we would have being bloked at C yk, and C zk annot be the urrent traversal point Global Order Constrution Algorithm With the mehanism for yles resolution in plae, the algorithm for onstruting a global transation order an be outlined as follows: 1. Connet all logs into a network. 2. Traverse the logs, detet, and resolve ommit symbol yles in the network. 3. Merge the network into a serial sequene. The rst step onnet loal logs into a network. The seond step detets, and resolve ommit symbol yles. The merging proessing is straightforward. Starting from the beginning of the logs, we traverse logs and output ommit symbols in arbitrary fashion, logs as long as their relative orders in the original logs are not violated. One a merge point is enountered, we annot traverse and output beyond this point before all other log onneted to the merge point has always reahed the merge point. A merge point is output only one. Steps two and three an be merged into one pass of the network in atual implementation, but are stated as separate proess here for larity. 20

21 5 Disussions Assume there are n transations and p nodes, whih means p loal logs, and that the total number of log entries in all nodes are l. Assuming the average number of nodes a transation aesses is f, we know that n f = l. The omplexity of our algorithm is analyzed as follows: 1. To onnet logs into a network at ommit symbol, we need to traverse all logs, and maintain the logs that onnet to a merge point. This require O(l) time and O(l) spae. 2. For yle detetion, we need to traverse all log entries, this is O(l) time. At eah merge point (one per transation), we need to proess all logs that onnet to the merge point. This takes O(n f) = O(l) time. 3. To an identify unsupported merge point, we need to examine at worst all merge points present at the time of the traversal stall. There are at most p merge points present. Eah time this is done we have one less transation to onsider, so the total omplexity of this step is O(n p), whih is (l p ). f 4. One the yle is identied, the slide operation an be implement by a delete and an insert of the network. This is take onstant time for eah yle. Sine resolving eah yle results at least one symbol being un-reahable (the symbol that is moved by the slide operation), there an at least be O(l) yle resolution. 5. Merging the network into a sequene take O(l) time. The total time omplexity is thus O(l p ), or O(n p), where p is the number of nodes in f the database system. 21

22 6 Conlusions In typial pratie of data warehouse, the warehouse data is kept in a separate site from the operation data. Changes to the operation data to propagated to the warehouse site periodially, usually by shipping the new log entries in the operation site. When the operation site is a parallel DBMS, it is neessary to arrange the log entries into an oherent order for the warehouse data to maintain onsisteny and integrity. This paper solves the problem by presenting a simple method for onstruting a global transation order out of a set of loal logs of a parallel or distributed system. In this paper, we present a simple mehanism that determines the order of transations by examining only the relative position of their vote and ommit entries. We show that if there is a serialization requirement of transation A before transation B, then the relative pattern C A?V B?C B must exist in some loal log. Furthermore, when suh a pattern exist, it is always safe to order A before B. Based on this mehanism, we present a method that eiently onstruts a global transation order. Our method uses the position of ommit symbols to denote impliitly the transation order. It rst onnets the vote and ommit symbols of loal log les into a network. Dierent logs are onneted at the merge points. The logs are then traversed to detet inonsisteny in the ordering of ommit symbols in dierent log. The traversal of logs must synhronize at merge point, and ordering inonsisteny an thereby be deteted. Inonsisteny in transation ordering is resolution using a very simple slide operation, whih always moves a ommit symbol bakward in its log, to the position immediately before a oniting ommit symbol. Our method has the following features. First, it examines only the vote and ommit entries of transations, but not the data aessing entries. Seond, for vote or ommit entries, only their relative positions and their transation ID (to identify from whih transation an entry omes) requires attention. Other information reorded in the vote or ommit entry need not be examined. Finally, our method has low omplexity. For a set of loal logs with l totally, only O(l) ost is required both in terms of time and spae. In summary, the 22

23 method we presented is oneptually simple, only requires aessing minimal information, and an leads to eient implementations. Referenes [BGHJ92] A. Bhide, A. Goyal, H. Hsiao, and A. Jhingran. An eient sheme for providing high availability. In Proeedings of 1992 ACM SIGMOD Conferene, San Diego, [BHG87] P. A. Bernstein, V. Hadzilaos, and N. Goodman. Conurreny Control and Reovery in Database Systems. Addison-Wesley, [CD97] S. Chaudhuri and U. Dayal. An overview of data warehousing and olap tehnology. ACM SIGMOD Reord, 26(1):65{74, [GR93] J. Gray and Andreas Reuter. Transation Proessing: Conepts and Tehniques. Morgan Kaufmann, [Gra78] J. Gray. Notes on database operating systems. Operating Systems: An Advanes Course, [Hsi94] H. Hsiao. Virtual timestamp for poin-in-time reovery in db2 pe. IBM DB2 Parallel Edition design doument, [Inm96] W. Inmon. The data warehous and data mining. Communiations of the ACM, Nov [MHLP91] C. Mohen, D. Haderle, B. Linsay, and H. Pirahesh. Aries: a transation reovery method supporting ne-granularity loking and partial rollbaks using writeahead logging. ACM, TODS, 17, [MLO86] C. Mohan, B. Lindsay, and R. Obermark. Transation management in the R distributed database management system. ACM Transation on Database Systems, 11(4):378{96, Deember

24 [SC90] J. Stamos and F. Cristian. A low-ost atomi ommit protool. In Proeedings of Ninth Symposium on Reliable Distributed Systems, pages 66{75. IEEE, Otober

A Novel Timestamp Ordering Approach for Co-existing Traditional and Cooperative Transaction Processing

A Novel Timestamp Ordering Approach for Co-existing Traditional and Cooperative Transaction Processing A Novel Timestamp Ordering Approah for Co-existing Traditional and Cooperative Transation Proessing Author Sun, Chengzheng, Zhang, Y., Kambayashi, Y., Yang, Y. Published 1998 Conferene Title Proeedings

More information

Extracting Partition Statistics from Semistructured Data

Extracting Partition Statistics from Semistructured Data Extrating Partition Statistis from Semistrutured Data John N. Wilson Rihard Gourlay Robert Japp Mathias Neumüller Department of Computer and Information Sienes University of Strathlyde, Glasgow, UK {jnw,rsg,rpj,mathias}@is.strath.a.uk

More information

On - Line Path Delay Fault Testing of Omega MINs M. Bellos 1, E. Kalligeros 1, D. Nikolos 1,2 & H. T. Vergos 1,2

On - Line Path Delay Fault Testing of Omega MINs M. Bellos 1, E. Kalligeros 1, D. Nikolos 1,2 & H. T. Vergos 1,2 On - Line Path Delay Fault Testing of Omega MINs M. Bellos, E. Kalligeros, D. Nikolos,2 & H. T. Vergos,2 Dept. of Computer Engineering and Informatis 2 Computer Tehnology Institute University of Patras,

More information

Outline: Software Design

Outline: Software Design Outline: Software Design. Goals History of software design ideas Design priniples Design methods Life belt or leg iron? (Budgen) Copyright Nany Leveson, Sept. 1999 A Little History... At first, struggling

More information

Pipelined Multipliers for Reconfigurable Hardware

Pipelined Multipliers for Reconfigurable Hardware Pipelined Multipliers for Reonfigurable Hardware Mithell J. Myjak and José G. Delgado-Frias Shool of Eletrial Engineering and Computer Siene, Washington State University Pullman, WA 99164-2752 USA {mmyjak,

More information

Automatic Physical Design Tuning: Workload as a Sequence Sanjay Agrawal Microsoft Research One Microsoft Way Redmond, WA, USA +1-(425)

Automatic Physical Design Tuning: Workload as a Sequence Sanjay Agrawal Microsoft Research One Microsoft Way Redmond, WA, USA +1-(425) Automati Physial Design Tuning: Workload as a Sequene Sanjay Agrawal Mirosoft Researh One Mirosoft Way Redmond, WA, USA +1-(425) 75-357 sagrawal@mirosoft.om Eri Chu * Computer Sienes Department University

More information

A DYNAMIC ACCESS CONTROL WITH BINARY KEY-PAIR

A DYNAMIC ACCESS CONTROL WITH BINARY KEY-PAIR Malaysian Journal of Computer Siene, Vol 10 No 1, June 1997, pp 36-41 A DYNAMIC ACCESS CONTROL WITH BINARY KEY-PAIR Md Rafiqul Islam, Harihodin Selamat and Mohd Noor Md Sap Faulty of Computer Siene and

More information

CleanUp: Improving Quadrilateral Finite Element Meshes

CleanUp: Improving Quadrilateral Finite Element Meshes CleanUp: Improving Quadrilateral Finite Element Meshes Paul Kinney MD-10 ECC P.O. Box 203 Ford Motor Company Dearborn, MI. 8121 (313) 28-1228 pkinney@ford.om Abstrat: Unless an all quadrilateral (quad)

More information

Algorithms for External Memory Lecture 6 Graph Algorithms - Weighted List Ranking

Algorithms for External Memory Lecture 6 Graph Algorithms - Weighted List Ranking Algorithms for External Memory Leture 6 Graph Algorithms - Weighted List Ranking Leturer: Nodari Sithinava Sribe: Andi Hellmund, Simon Ohsenreither 1 Introdution & Motivation After talking about I/O-effiient

More information

Algorithms, Mechanisms and Procedures for the Computer-aided Project Generation System

Algorithms, Mechanisms and Procedures for the Computer-aided Project Generation System Algorithms, Mehanisms and Proedures for the Computer-aided Projet Generation System Anton O. Butko 1*, Aleksandr P. Briukhovetskii 2, Dmitry E. Grigoriev 2# and Konstantin S. Kalashnikov 3 1 Department

More information

What are Cycle-Stealing Systems Good For? A Detailed Performance Model Case Study

What are Cycle-Stealing Systems Good For? A Detailed Performance Model Case Study What are Cyle-Stealing Systems Good For? A Detailed Performane Model Case Study Wayne Kelly and Jiro Sumitomo Queensland University of Tehnology, Australia {w.kelly, j.sumitomo}@qut.edu.au Abstrat The

More information

A Dual-Hamiltonian-Path-Based Multicasting Strategy for Wormhole-Routed Star Graph Interconnection Networks

A Dual-Hamiltonian-Path-Based Multicasting Strategy for Wormhole-Routed Star Graph Interconnection Networks A Dual-Hamiltonian-Path-Based Multiasting Strategy for Wormhole-Routed Star Graph Interonnetion Networks Nen-Chung Wang Department of Information and Communiation Engineering Chaoyang University of Tehnology,

More information

Gray Codes for Reflectable Languages

Gray Codes for Reflectable Languages Gray Codes for Refletable Languages Yue Li Joe Sawada Marh 8, 2008 Abstrat We lassify a type of language alled a refletable language. We then develop a generi algorithm that an be used to list all strings

More information

Exploring the Commonality in Feature Modeling Notations

Exploring the Commonality in Feature Modeling Notations Exploring the Commonality in Feature Modeling Notations Miloslav ŠÍPKA Slovak University of Tehnology Faulty of Informatis and Information Tehnologies Ilkovičova 3, 842 16 Bratislava, Slovakia miloslav.sipka@gmail.om

More information

HEXA: Compact Data Structures for Faster Packet Processing

HEXA: Compact Data Structures for Faster Packet Processing Washington University in St. Louis Washington University Open Sholarship All Computer Siene and Engineering Researh Computer Siene and Engineering Report Number: 27-26 27 HEXA: Compat Data Strutures for

More information

A Load-Balanced Clustering Protocol for Hierarchical Wireless Sensor Networks

A Load-Balanced Clustering Protocol for Hierarchical Wireless Sensor Networks International Journal of Advanes in Computer Networks and Its Seurity IJCNS A Load-Balaned Clustering Protool for Hierarhial Wireless Sensor Networks Mehdi Tarhani, Yousef S. Kavian, Saman Siavoshi, Ali

More information

- 1 - S 21. Directory-based Administration of Virtual Private Networks: Policy & Configuration. Charles A Kunzinger.

- 1 - S 21. Directory-based Administration of Virtual Private Networks: Policy & Configuration. Charles A Kunzinger. - 1 - S 21 Diretory-based Administration of Virtual Private Networks: Poliy & Configuration Charles A Kunzinger kunzinge@us.ibm.om - 2 - Clik here Agenda to type page title What is a VPN? What is VPN Poliy?

More information

CA Privileged Identity Manager r12.x (CA ControlMinder) Implementation Proven Professional Exam (CAT-480) Study Guide Version 1.5

CA Privileged Identity Manager r12.x (CA ControlMinder) Implementation Proven Professional Exam (CAT-480) Study Guide Version 1.5 Proven Professional Exam (CAT-480) Study Guide Version 1.5 PROPRIETARY AND CONFIDENTIAL INFORMATION 2016 CA. All rights reserved. CA onfidential & proprietary information. For CA, CA Partner and CA Customer

More information

Partial Character Decoding for Improved Regular Expression Matching in FPGAs

Partial Character Decoding for Improved Regular Expression Matching in FPGAs Partial Charater Deoding for Improved Regular Expression Mathing in FPGAs Peter Sutton Shool of Information Tehnology and Eletrial Engineering The University of Queensland Brisbane, Queensland, 4072, Australia

More information

Performance Improvement of TCP on Wireless Cellular Networks by Adaptive FEC Combined with Explicit Loss Notification

Performance Improvement of TCP on Wireless Cellular Networks by Adaptive FEC Combined with Explicit Loss Notification erformane Improvement of TC on Wireless Cellular Networks by Adaptive Combined with Expliit Loss tifiation Masahiro Miyoshi, Masashi Sugano, Masayuki Murata Department of Infomatis and Mathematial Siene,

More information

Chapter 2: Introduction to Maple V

Chapter 2: Introduction to Maple V Chapter 2: Introdution to Maple V 2-1 Working with Maple Worksheets Try It! (p. 15) Start a Maple session with an empty worksheet. The name of the worksheet should be Untitled (1). Use one of the standard

More information

C 2 C 3 C 1 M S. f e. e f (3,0) (0,1) (2,0) (-1,1) (1,0) (-1,0) (1,-1) (0,-1) (-2,0) (-3,0) (0,-2)

C 2 C 3 C 1 M S. f e. e f (3,0) (0,1) (2,0) (-1,1) (1,0) (-1,0) (1,-1) (0,-1) (-2,0) (-3,0) (0,-2) SPECIAL ISSUE OF IEEE TRANSACTIONS ON ROBOTICS AND AUTOMATION: MULTI-ROBOT SSTEMS, 00 Distributed reonfiguration of hexagonal metamorphi robots Jennifer E. Walter, Jennifer L. Welh, and Nany M. Amato Abstrat

More information

Trajectory Tracking Control for A Wheeled Mobile Robot Using Fuzzy Logic Controller

Trajectory Tracking Control for A Wheeled Mobile Robot Using Fuzzy Logic Controller Trajetory Traking Control for A Wheeled Mobile Robot Using Fuzzy Logi Controller K N FARESS 1 M T EL HAGRY 1 A A EL KOSY 2 1 Eletronis researh institute, Cairo, Egypt 2 Faulty of Engineering, Cairo University,

More information

Learning Convention Propagation in BeerAdvocate Reviews from a etwork Perspective. Abstract

Learning Convention Propagation in BeerAdvocate Reviews from a etwork Perspective. Abstract CS 9 Projet Final Report: Learning Convention Propagation in BeerAdvoate Reviews from a etwork Perspetive Abstrat We look at the way onventions propagate between reviews on the BeerAdvoate dataset, and

More information

Query Optimization for Structured Documents Based on Knowledge. on the Document Type Denition. Institute of Inf. Systems. ETH Zentrum Dolivostr.

Query Optimization for Structured Documents Based on Knowledge. on the Document Type Denition. Institute of Inf. Systems. ETH Zentrum Dolivostr. Query Optimization for Strutured Douments Based on Knowledge on the Doument Type Denition Klemens Bohm Karl Aberer Institute of Inf. Systems GMD{IPSI ETH Zentrum Dolivostr. 15 8092 Zurih, Switzerland 64293

More information

mahines. HBSP enhanes the appliability of the BSP model by inorporating parameters that reet the relative speeds of the heterogeneous omputing omponen

mahines. HBSP enhanes the appliability of the BSP model by inorporating parameters that reet the relative speeds of the heterogeneous omputing omponen The Heterogeneous Bulk Synhronous Parallel Model Tiani L. Williams and Rebea J. Parsons Shool of Computer Siene University of Central Florida Orlando, FL 32816-2362 fwilliams,rebeag@s.uf.edu Abstrat. Trends

More information

1. The collection of the vowels in the word probability. 2. The collection of real numbers that satisfy the equation x 9 = 0.

1. The collection of the vowels in the word probability. 2. The collection of real numbers that satisfy the equation x 9 = 0. C HPTER 1 SETS I. DEFINITION OF SET We begin our study of probability with the disussion of the basi onept of set. We assume that there is a ommon understanding of what is meant by the notion of a olletion

More information

A Partial Sorting Algorithm in Multi-Hop Wireless Sensor Networks

A Partial Sorting Algorithm in Multi-Hop Wireless Sensor Networks A Partial Sorting Algorithm in Multi-Hop Wireless Sensor Networks Abouberine Ould Cheikhna Department of Computer Siene University of Piardie Jules Verne 80039 Amiens Frane Ould.heikhna.abouberine @u-piardie.fr

More information

Bayesian Belief Networks for Data Mining. Harald Steck and Volker Tresp. Siemens AG, Corporate Technology. Information and Communications

Bayesian Belief Networks for Data Mining. Harald Steck and Volker Tresp. Siemens AG, Corporate Technology. Information and Communications Bayesian Belief Networks for Data Mining Harald Stek and Volker Tresp Siemens AG, Corporate Tehnology Information and Communiations 81730 Munih, Germany fharald.stek, Volker.Trespg@mhp.siemens.de Abstrat

More information

RAC 2 E: Novel Rendezvous Protocol for Asynchronous Cognitive Radios in Cooperative Environments

RAC 2 E: Novel Rendezvous Protocol for Asynchronous Cognitive Radios in Cooperative Environments 21st Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communiations 1 RAC 2 E: Novel Rendezvous Protool for Asynhronous Cognitive Radios in Cooperative Environments Valentina Pavlovska,

More information

This fact makes it difficult to evaluate the cost function to be minimized

This fact makes it difficult to evaluate the cost function to be minimized RSOURC LLOCTION N SSINMNT In the resoure alloation step the amount of resoures required to exeute the different types of proesses is determined. We will refer to the time interval during whih a proess

More information

The recursive decoupling method for solving tridiagonal linear systems

The recursive decoupling method for solving tridiagonal linear systems Loughborough University Institutional Repository The reursive deoupling method for solving tridiagonal linear systems This item was submitted to Loughborough University's Institutional Repository by the/an

More information

Batch Auditing for Multiclient Data in Multicloud Storage

Batch Auditing for Multiclient Data in Multicloud Storage Advaned Siene and Tehnology Letters, pp.67-73 http://dx.doi.org/0.4257/astl.204.50. Bath Auditing for Multilient Data in Multiloud Storage Zhihua Xia, Xinhui Wang, Xingming Sun, Yafeng Zhu, Peng Ji and

More information

The Minimum Redundancy Maximum Relevance Approach to Building Sparse Support Vector Machines

The Minimum Redundancy Maximum Relevance Approach to Building Sparse Support Vector Machines The Minimum Redundany Maximum Relevane Approah to Building Sparse Support Vetor Mahines Xiaoxing Yang, Ke Tang, and Xin Yao, Nature Inspired Computation and Appliations Laboratory (NICAL), Shool of Computer

More information

A Novel Bit Level Time Series Representation with Implication of Similarity Search and Clustering

A Novel Bit Level Time Series Representation with Implication of Similarity Search and Clustering A Novel Bit Level Time Series Representation with Impliation of Similarity Searh and lustering hotirat Ratanamahatana, Eamonn Keogh, Anthony J. Bagnall 2, and Stefano Lonardi Dept. of omputer Siene & Engineering,

More information

Colouring contact graphs of squares and rectilinear polygons de Berg, M.T.; Markovic, A.; Woeginger, G.

Colouring contact graphs of squares and rectilinear polygons de Berg, M.T.; Markovic, A.; Woeginger, G. Colouring ontat graphs of squares and retilinear polygons de Berg, M.T.; Markovi, A.; Woeginger, G. Published in: nd European Workshop on Computational Geometry (EuroCG 06), 0 Marh - April, Lugano, Switzerland

More information

Flow Demands Oriented Node Placement in Multi-Hop Wireless Networks

Flow Demands Oriented Node Placement in Multi-Hop Wireless Networks Flow Demands Oriented Node Plaement in Multi-Hop Wireless Networks Zimu Yuan Institute of Computing Tehnology, CAS, China {zimu.yuan}@gmail.om arxiv:153.8396v1 [s.ni] 29 Mar 215 Abstrat In multi-hop wireless

More information

A Novel Validity Index for Determination of the Optimal Number of Clusters

A Novel Validity Index for Determination of the Optimal Number of Clusters IEICE TRANS. INF. & SYST., VOL.E84 D, NO.2 FEBRUARY 2001 281 LETTER A Novel Validity Index for Determination of the Optimal Number of Clusters Do-Jong KIM, Yong-Woon PARK, and Dong-Jo PARK, Nonmembers

More information

Multi-Channel Wireless Networks: Capacity and Protocols

Multi-Channel Wireless Networks: Capacity and Protocols Multi-Channel Wireless Networks: Capaity and Protools Tehnial Report April 2005 Pradeep Kyasanur Dept. of Computer Siene, and Coordinated Siene Laboratory, University of Illinois at Urbana-Champaign Email:

More information

Detection and Recognition of Non-Occluded Objects using Signature Map

Detection and Recognition of Non-Occluded Objects using Signature Map 6th WSEAS International Conferene on CIRCUITS, SYSTEMS, ELECTRONICS,CONTROL & SIGNAL PROCESSING, Cairo, Egypt, De 9-31, 007 65 Detetion and Reognition of Non-Oluded Objets using Signature Map Sangbum Park,

More information

Self-Adaptive Parent to Mean-Centric Recombination for Real-Parameter Optimization

Self-Adaptive Parent to Mean-Centric Recombination for Real-Parameter Optimization Self-Adaptive Parent to Mean-Centri Reombination for Real-Parameter Optimization Kalyanmoy Deb and Himanshu Jain Department of Mehanial Engineering Indian Institute of Tehnology Kanpur Kanpur, PIN 86 {deb,hjain}@iitk.a.in

More information

Space- and Time-Efficient BDD Construction via Working Set Control

Space- and Time-Efficient BDD Construction via Working Set Control Spae- and Time-Effiient BDD Constrution via Working Set Control Bwolen Yang Yirng-An Chen Randal E. Bryant David R. O Hallaron Computer Siene Department Carnegie Mellon University Pittsburgh, PA 15213.

More information

Analysis of input and output configurations for use in four-valued CCD programmable logic arrays

Analysis of input and output configurations for use in four-valued CCD programmable logic arrays nalysis of input and output onfigurations for use in four-valued D programmable logi arrays J.T. utler H.G. Kerkhoff ndexing terms: Logi, iruit theory and design, harge-oupled devies bstrat: s in binary,

More information

Accommodations of QoS DiffServ Over IP and MPLS Networks

Accommodations of QoS DiffServ Over IP and MPLS Networks Aommodations of QoS DiffServ Over IP and MPLS Networks Abdullah AlWehaibi, Anjali Agarwal, Mihael Kadoh and Ahmed ElHakeem Department of Eletrial and Computer Department de Genie Eletrique Engineering

More information

Implementing Load-Balanced Switches With Fat-Tree Networks

Implementing Load-Balanced Switches With Fat-Tree Networks Implementing Load-Balaned Swithes With Fat-Tree Networks Hung-Shih Chueh, Ching-Min Lien, Cheng-Shang Chang, Jay Cheng, and Duan-Shin Lee Department of Eletrial Engineering & Institute of Communiations

More information

A {k, n}-secret Sharing Scheme for Color Images

A {k, n}-secret Sharing Scheme for Color Images A {k, n}-seret Sharing Sheme for Color Images Rastislav Luka, Konstantinos N. Plataniotis, and Anastasios N. Venetsanopoulos The Edward S. Rogers Sr. Dept. of Eletrial and Computer Engineering, University

More information

References. December 1992, pp. 71 { 81. pp.457{467. Magazine, June for very large high throughput database systems,"

References. December 1992, pp. 71 { 81. pp.457{467. Magazine, June for very large high throughput database systems, the overall working time for other appliations. In ase, data ltering was the only appliation being run, then using distributed indexing, we an serve 00 times as many requests. 6 Conlusion We have explored

More information

Abstract. Key Words: Image Filters, Fuzzy Filters, Order Statistics Filters, Rank Ordered Mean Filters, Channel Noise. 1.

Abstract. Key Words: Image Filters, Fuzzy Filters, Order Statistics Filters, Rank Ordered Mean Filters, Channel Noise. 1. Fuzzy Weighted Rank Ordered Mean (FWROM) Filters for Mixed Noise Suppression from Images S. Meher, G. Panda, B. Majhi 3, M.R. Meher 4,,4 Department of Eletronis and I.E., National Institute of Tehnology,

More information

Improved flooding of broadcast messages using extended multipoint relaying

Improved flooding of broadcast messages using extended multipoint relaying Improved flooding of broadast messages using extended multipoint relaying Pere Montolio Aranda a, Joaquin Garia-Alfaro a,b, David Megías a a Universitat Oberta de Catalunya, Estudis d Informàtia, Mulimèdia

More information

A Comparison of Hard-state and Soft-state Signaling Protocols

A Comparison of Hard-state and Soft-state Signaling Protocols University of Massahusetts Amherst SholarWorks@UMass Amherst Computer Siene Department Faulty Publiation Series Computer Siene 2003 A Comparison of Hard-state and Soft-state Signaling Protools Ping Ji

More information

35 th Design Automation Conference Copyright 1998 ACM

35 th Design Automation Conference Copyright 1998 ACM Using Reongurable Computing Tehniques to Aelerate Problems in the CAD Domain: A Case Study with Boolean Satisability Peixin Zhong, Pranav Ashar, Sharad Malik and Margaret Martonosi Prineton University

More information

Parametric Abstract Domains for Shape Analysis

Parametric Abstract Domains for Shape Analysis Parametri Abstrat Domains for Shape Analysis Xavier RIVAL (INRIA & Éole Normale Supérieure) Joint work with Bor-Yuh Evan CHANG (University of Maryland U University of Colorado) and George NECULA (University

More information

1. Introduction. 2. The Probable Stope Algorithm

1. Introduction. 2. The Probable Stope Algorithm 1. Introdution Optimization in underground mine design has reeived less attention than that in open pit mines. This is mostly due to the diversity o underground mining methods and omplexity o underground

More information

An Alternative Approach to the Fuzzifier in Fuzzy Clustering to Obtain Better Clustering Results

An Alternative Approach to the Fuzzifier in Fuzzy Clustering to Obtain Better Clustering Results An Alternative Approah to the Fuzziier in Fuzzy Clustering to Obtain Better Clustering Results Frank Klawonn Department o Computer Siene University o Applied Sienes BS/WF Salzdahlumer Str. 46/48 D-38302

More information

The Happy Ending Problem

The Happy Ending Problem The Happy Ending Problem Neeldhara Misra STATUTORY WARNING This doument is a draft version 1 Introdution The Happy Ending problem first manifested itself on a typial wintery evening in 1933 These evenings

More information

PBFT: A Byzantine Renaissance. The Setup. What could possibly go wrong? The General Idea. Practical Byzantine Fault-Tolerance (CL99, CL00)

PBFT: A Byzantine Renaissance. The Setup. What could possibly go wrong? The General Idea. Practical Byzantine Fault-Tolerance (CL99, CL00) PBFT: A Byzantine Renaissane Pratial Byzantine Fault-Tolerane (CL99, CL00) first to be safe in asynhronous systems live under weak synhrony assumptions -Byzantine Paos! The Setup Crypto System Model Asynhronous

More information

Facility Location: Distributed Approximation

Facility Location: Distributed Approximation Faility Loation: Distributed Approximation Thomas Mosibroda Roger Wattenhofer Distributed Computing Group PODC 2005 Where to plae ahes in the Internet? A distributed appliation that has to dynamially plae

More information

DETECTION METHOD FOR NETWORK PENETRATING BEHAVIOR BASED ON COMMUNICATION FINGERPRINT

DETECTION METHOD FOR NETWORK PENETRATING BEHAVIOR BASED ON COMMUNICATION FINGERPRINT DETECTION METHOD FOR NETWORK PENETRATING BEHAVIOR BASED ON COMMUNICATION FINGERPRINT 1 ZHANGGUO TANG, 2 HUANZHOU LI, 3 MINGQUAN ZHONG, 4 JIAN ZHANG 1 Institute of Computer Network and Communiation Tehnology,

More information

Why then another BFT protocol? Zyzzyva. Simplify, simplify. Simplify, simplify. Complex decision tree hampers BFT adoption. H.D. Thoreau. H.D.

Why then another BFT protocol? Zyzzyva. Simplify, simplify. Simplify, simplify. Complex decision tree hampers BFT adoption. H.D. Thoreau. H.D. Why then another BFT protool? Yes No Zyzzyva Yes No Yes No Comple deision tree hampers BFT adoption Simplify, simplify H.D. Thoreau Simplify, simplify H.D. Thoreau Yes No Yes No Yes Yes No One protool

More information

Evaluation of Benchmark Performance Estimation for Parallel. Fortran Programs on Massively Parallel SIMD and MIMD. Computers.

Evaluation of Benchmark Performance Estimation for Parallel. Fortran Programs on Massively Parallel SIMD and MIMD. Computers. Evaluation of Benhmark Performane Estimation for Parallel Fortran Programs on Massively Parallel SIMD and MIMD Computers Thomas Fahringer Dept of Software Tehnology and Parallel Systems University of Vienna

More information

splitting tehniques that partition live ranges have been proposed to solve both the spilling problem[5][8] and the assignment problem[8][9]. The parti

splitting tehniques that partition live ranges have been proposed to solve both the spilling problem[5][8] and the assignment problem[8][9]. The parti Load/Store Range Analysis for Global Register Alloation Priyadarshan Kolte and Mary Jean Harrold Department of Computer Siene Clemson University Abstrat Live range splitting tehniques divide the live ranges

More information

CA Release Automation 5.x Implementation Proven Professional Exam (CAT-600) Study Guide Version 1.1

CA Release Automation 5.x Implementation Proven Professional Exam (CAT-600) Study Guide Version 1.1 Exam (CAT-600) Study Guide Version 1.1 PROPRIETARY AND CONFIDENTIAL INFORMATION 2016 CA. All rights reserved. CA onfidential & proprietary information. For CA, CA Partner and CA Customer use only. No unauthorized

More information

A Unified Subdivision Scheme for Polygonal Modeling

A Unified Subdivision Scheme for Polygonal Modeling EUROGRAPHICS 2 / A. Chalmers and T.-M. Rhyne (Guest Editors) Volume 2 (2), Number 3 A Unified Subdivision Sheme for Polygonal Modeling Jérôme Maillot Jos Stam Alias Wavefront Alias Wavefront 2 King St.

More information

Acoustic Links. Maximizing Channel Utilization for Underwater

Acoustic Links. Maximizing Channel Utilization for Underwater Maximizing Channel Utilization for Underwater Aousti Links Albert F Hairris III Davide G. B. Meneghetti Adihele Zorzi Department of Information Engineering University of Padova, Italy Email: {harris,davide.meneghetti,zorzi}@dei.unipd.it

More information

Sparse Certificates for 2-Connectivity in Directed Graphs

Sparse Certificates for 2-Connectivity in Directed Graphs Sparse Certifiates for 2-Connetivity in Direted Graphs Loukas Georgiadis Giuseppe F. Italiano Aikaterini Karanasiou Charis Papadopoulos Nikos Parotsidis Abstrat Motivated by the emergene of large-sale

More information

Verification of Parameterized Concurrent Programs By Modular Reasoning about Data and Control

Verification of Parameterized Concurrent Programs By Modular Reasoning about Data and Control Verifiation of Parameterized Conurrent Programs By Modular Reasoning about Data and Control Azadeh Farzan Zahary Kinaid University of Toronto azadeh,zkinaid@s.toronto.edu Abstrat In this paper, we onsider

More information

Multi-Piece Mold Design Based on Linear Mixed-Integer Program Toward Guaranteed Optimality

Multi-Piece Mold Design Based on Linear Mixed-Integer Program Toward Guaranteed Optimality INTERNATIONAL CONFERENCE ON MANUFACTURING AUTOMATION (ICMA200) Multi-Piee Mold Design Based on Linear Mixed-Integer Program Toward Guaranteed Optimality Stephen Stoyan, Yong Chen* Epstein Department of

More information

SAND Unlimited Release Printed November 1995 Updated November 29, :26 PM EXODUS II: A Finite Element Data Model

SAND Unlimited Release Printed November 1995 Updated November 29, :26 PM EXODUS II: A Finite Element Data Model SAND92-2137 Unlimited Release Printed November 1995 Updated November 29, 2006 12:26 PM EXODUS II: A Finite Element Data Model Gregory D. Sjaardema (updated version) Larry A. Shoof, Vitor R. Yarberry Computational

More information

Test Case Generation from UML State Machines

Test Case Generation from UML State Machines Test Case Generation from UML State Mahines Dirk Seifert To ite this version: Dirk Seifert. Test Case Generation from UML State Mahines. [Researh Report] 2008. HAL Id: inria-00268864

More information

Speeding up Consensus by Chasing Fast Decisions

Speeding up Consensus by Chasing Fast Decisions Speeding up Consensus by Chasing Fast Deisions Balaji Arun, Sebastiano Peluso, Roberto Palmieri, Giuliano Losa, Binoy Ravindran ECE, Virginia Teh, USA {balajia,peluso,robertop,giuliano.losa,binoy}@vt.edu

More information

Folding. Hardware Mapped vs. Time multiplexed. Folding by N (N=folding factor) Node A. Unfolding by J A 1 A J-1. Time multiplexed/microcoded

Folding. Hardware Mapped vs. Time multiplexed. Folding by N (N=folding factor) Node A. Unfolding by J A 1 A J-1. Time multiplexed/microcoded Folding is verse of Unfolding Node A A Folding by N (N=folding fator) Folding A Unfolding by J A A J- Hardware Mapped vs. Time multiplexed l Hardware Mapped vs. Time multiplexed/mirooded FI : y x(n) h

More information

UCSB Math TI-85 Tutorials: Basics

UCSB Math TI-85 Tutorials: Basics 3 UCSB Math TI-85 Tutorials: Basis If your alulator sreen doesn t show anything, try adjusting the ontrast aording to the instrutions on page 3, or page I-3, of the alulator manual You should read the

More information

COST PERFORMANCE ASPECTS OF CCD FAST AUXILIARY MEMORY

COST PERFORMANCE ASPECTS OF CCD FAST AUXILIARY MEMORY COST PERFORMANCE ASPECTS OF CCD FAST AUXILIARY MEMORY Dileep P, Bhondarkor Texas Instruments Inorporated Dallas, Texas ABSTRACT Charge oupled devies (CCD's) hove been mentioned as potential fast auxiliary

More information

Verifying Interaction Protocol Compliance of Service Orchestrations

Verifying Interaction Protocol Compliance of Service Orchestrations Verifying Interation Protool Compliane of Servie Orhestrations Andreas Shroeder and Philip Mayer Ludwig-Maximilians-Universität Münhen, Germany {shroeda, mayer}@pst.ifi.lmu.de Abstrat. An important aspet

More information

Multi-Level Modeling of Concurrent and Distributed Systems

Multi-Level Modeling of Concurrent and Distributed Systems Multi-Level Modeling of Conurrent and Distriuted Systems Peter Taeling Hasso-Plattner-Institute for Software Systems Engineering P.O. Box 90 04 60, 14440 Potsdam, Germany taeling@hpi.uni-potsdam.de strat

More information

Performance Benchmarks for an Interactive Video-on-Demand System

Performance Benchmarks for an Interactive Video-on-Demand System Performane Benhmarks for an Interative Video-on-Demand System. Guo,P.G.Taylor,E.W.M.Wong,S.Chan,M.Zukerman andk.s.tang ARC Speial Researh Centre for Ultra-Broadband Information Networks (CUBIN) Department

More information

A Formal Hybrid Analysis Technique for Composite Web Services Verification

A Formal Hybrid Analysis Technique for Composite Web Services Verification A Formal Hybrid Analysis Tehnique for Composite Web Servies Verifiation MAY HAIDAR 1,2, HICHAM H. HALLAL 1 1 Computer Siene Department / Department of Eletrial Engineering Fahad Bin Sultan University P.O

More information

Cross-layer Resource Allocation on Broadband Power Line Based on Novel QoS-priority Scheduling Function in MAC Layer

Cross-layer Resource Allocation on Broadband Power Line Based on Novel QoS-priority Scheduling Function in MAC Layer Communiations and Networ, 2013, 5, 69-73 http://dx.doi.org/10.4236/n.2013.53b2014 Published Online September 2013 (http://www.sirp.org/journal/n) Cross-layer Resoure Alloation on Broadband Power Line Based

More information

1 The Knuth-Morris-Pratt Algorithm

1 The Knuth-Morris-Pratt Algorithm 5-45/65: Design & Analysis of Algorithms September 26, 26 Leture #9: String Mathing last hanged: September 26, 27 There s an entire field dediated to solving problems on strings. The book Algorithms on

More information

DECT Module Installation Manual

DECT Module Installation Manual DECT Module Installation Manual Rev. 2.0 This manual desribes the DECT module registration method to the HUB and fan airflow settings. In order for the HUB to ommuniate with a ompatible fan, the DECT module

More information

Directed Rectangle-Visibility Graphs have. Abstract. Visibility representations of graphs map vertices to sets in Euclidean space and

Directed Rectangle-Visibility Graphs have. Abstract. Visibility representations of graphs map vertices to sets in Euclidean space and Direted Retangle-Visibility Graphs have Unbounded Dimension Kathleen Romanik DIMACS Center for Disrete Mathematis and Theoretial Computer Siene Rutgers, The State University of New Jersey P.O. Box 1179,

More information

Taming Decentralized POMDPs: Towards Efficient Policy Computation for Multiagent Settings

Taming Decentralized POMDPs: Towards Efficient Policy Computation for Multiagent Settings Taming Deentralized PMDPs: Towards ffiient Poliy omputation for Multiagent Settings. Nair and M. Tambe omputer Siene Dept. University of Southern alifornia Los Angeles A 90089 nair,tambe @us.edu M. Yokoo

More information

Parallelizing Frequent Web Access Pattern Mining with Partial Enumeration for High Speedup

Parallelizing Frequent Web Access Pattern Mining with Partial Enumeration for High Speedup Parallelizing Frequent Web Aess Pattern Mining with Partial Enumeration for High Peiyi Tang Markus P. Turkia Department of Computer Siene Department of Computer Siene University of Arkansas at Little Rok

More information

A Multi-Head Clustering Algorithm in Vehicular Ad Hoc Networks

A Multi-Head Clustering Algorithm in Vehicular Ad Hoc Networks International Journal of Computer Theory and Engineering, Vol. 5, No. 2, April 213 A Multi-Head Clustering Algorithm in Vehiular Ad Ho Networks Shou-Chih Lo, Yi-Jen Lin, and Jhih-Siao Gao Abstrat Clustering

More information

8 Instruction Selection

8 Instruction Selection 8 Instrution Seletion The IR ode instrutions were designed to do exatly one operation: load/store, add, subtrat, jump, et. The mahine instrutions of a real CPU often perform several of these primitive

More information

Semi-Supervised Affinity Propagation with Instance-Level Constraints

Semi-Supervised Affinity Propagation with Instance-Level Constraints Semi-Supervised Affinity Propagation with Instane-Level Constraints Inmar E. Givoni, Brendan J. Frey Probabilisti and Statistial Inferene Group University of Toronto 10 King s College Road, Toronto, Ontario,

More information

Detection of RF interference to GPS using day-to-day C/No differences

Detection of RF interference to GPS using day-to-day C/No differences 1 International Symposium on GPS/GSS Otober 6-8, 1. Detetion of RF interferene to GPS using day-to-day /o differenes Ryan J. R. Thompson 1#, Jinghui Wu #, Asghar Tabatabaei Balaei 3^, and Andrew G. Dempster

More information

Zyzzyva: Speculative Byzantine Fault Tolerance By Ramakrishna Kotla,* Allen Clement, Edmund Wong, Lorenzo Alvisi, and Mike Dahlin

Zyzzyva: Speculative Byzantine Fault Tolerance By Ramakrishna Kotla,* Allen Clement, Edmund Wong, Lorenzo Alvisi, and Mike Dahlin : Speulative Byzantine Fault Tolerane By Ramakrishna Kotla,* Allen Clement, Edmund Wong, Lorenzo Alvisi, and Mike Dahlin doi:10.1145/1400214.1400236 Abstrat A longstanding vision in distributed systems

More information

Mining effective design solutions based on a model-driven approach

Mining effective design solutions based on a model-driven approach ata Mining VI 463 Mining effetive design solutions based on a model-driven approah T. Katsimpa 2, S. Sirmakessis 1,. Tsakalidis 1,2 & G. Tzimas 1,2 1 Researh ademi omputer Tehnology Institute, Hellas 2

More information

Architecture and Performance of the Hitachi SR2201 Massively Parallel Processor System

Architecture and Performance of the Hitachi SR2201 Massively Parallel Processor System Arhiteture and Performane of the Hitahi SR221 Massively Parallel Proessor System Hiroaki Fujii, Yoshiko Yasuda, Hideya Akashi, Yasuhiro Inagami, Makoto Koga*, Osamu Ishihara*, Masamori Kashiyama*, Hideo

More information

Reducing Runtime Complexity of Long-Running Application Services via Dynamic Profiling and Dynamic Bytecode Adaptation for Improved Quality of Service

Reducing Runtime Complexity of Long-Running Application Services via Dynamic Profiling and Dynamic Bytecode Adaptation for Improved Quality of Service Reduing Runtime Complexity of Long-Running Appliation Servies via Dynami Profiling and Dynami Byteode Adaptation for Improved Quality of Servie ABSTRACT John Bergin Performane Engineering Laboratory University

More information

Automated System for the Study of Environmental Loads Applied to Production Risers Dustin M. Brandt 1, Celso K. Morooka 2, Ivan R.

Automated System for the Study of Environmental Loads Applied to Production Risers Dustin M. Brandt 1, Celso K. Morooka 2, Ivan R. EngOpt 2008 - International Conferene on Engineering Optimization Rio de Janeiro, Brazil, 01-05 June 2008. Automated System for the Study of Environmental Loads Applied to Prodution Risers Dustin M. Brandt

More information

Detecting Outliers in High-Dimensional Datasets with Mixed Attributes

Detecting Outliers in High-Dimensional Datasets with Mixed Attributes Deteting Outliers in High-Dimensional Datasets with Mixed Attributes A. Koufakou, M. Georgiopoulos, and G.C. Anagnostopoulos 2 Shool of EECS, University of Central Florida, Orlando, FL, USA 2 Dept. of

More information

Establishing Secure Ethernet LANs Using Intelligent Switching Hubs in Internet Environments

Establishing Secure Ethernet LANs Using Intelligent Switching Hubs in Internet Environments Establishing Seure Ethernet LANs Using Intelligent Swithing Hubs in Internet Environments WOEIJIUNN TSAUR AND SHIJINN HORNG Department of Eletrial Engineering, National Taiwan University of Siene and Tehnology,

More information

Make your process world

Make your process world Automation platforms Modion Quantum Safety System Make your proess world a safer plae You are faing omplex hallenges... Safety is at the heart of your proess In order to maintain and inrease your ompetitiveness,

More information

Title: Time-Based Tree Graphs for Stabilized Force Structure Representations

Title: Time-Based Tree Graphs for Stabilized Force Structure Representations Paper for the 8 th International Command & Control Researh & Tehnology Symposium Title: Time-Based Tree Graphs for Stabilized Fore Struture Representations Submitted by: Sam Chamberlain U.S. Army Researh

More information

An Optimized Approach on Applying Genetic Algorithm to Adaptive Cluster Validity Index

An Optimized Approach on Applying Genetic Algorithm to Adaptive Cluster Validity Index IJCSES International Journal of Computer Sienes and Engineering Systems, ol., No.4, Otober 2007 CSES International 2007 ISSN 0973-4406 253 An Optimized Approah on Applying Geneti Algorithm to Adaptive

More information

Cluster Centric Fuzzy Modeling

Cluster Centric Fuzzy Modeling 10.1109/TFUZZ.014.300134, IEEE Transations on Fuzzy Systems TFS-013-0379.R1 1 Cluster Centri Fuzzy Modeling Witold Pedryz, Fellow, IEEE, and Hesam Izakian, Student Member, IEEE Abstrat In this study, we

More information

Model Based Approach for Content Based Image Retrievals Based on Fusion and Relevancy Methodology

Model Based Approach for Content Based Image Retrievals Based on Fusion and Relevancy Methodology The International Arab Journal of Information Tehnology, Vol. 12, No. 6, November 15 519 Model Based Approah for Content Based Image Retrievals Based on Fusion and Relevany Methodology Telu Venkata Madhusudhanarao

More information

Concurrent Timing Optimization of. Hong-Yean Hsieh, Wentai Liu, Ralph K. Cavin. Electronics Research Laboratory. North Carolina State University

Concurrent Timing Optimization of. Hong-Yean Hsieh, Wentai Liu, Ralph K. Cavin. Electronics Research Laboratory. North Carolina State University Conurrent Timing Optimization of Lath-Based Digital Systems 1 Hong-Yean Hsieh, Wentai Liu, Ralph K. Cavin Eletronis Researh Laboratory Department of Eletrial and Computer Engineering North Carolina State

More information