Edgar Knapp. "Deadlock Detection in Distributed Databases". ACM Computing Surveys, 19(4), 1987.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....locking rule, no lock on behalf of a transaction should be set once a lock previously held by the transaction is released , then it is possible to ensure the serializability. The side e ect is that they cause dead locks. Distributed deadlock detection algorithms are presented by [RSL78, CHM83] Kna87] surveys the distributed deadlock detection literature. Distributed reliability protocols. Distributed database system are potentially more reliable since there are multiple copies of each system component, which eliminates single points of failure. Data is replicated to ensure that data is ....

E. Knapp. Deadlock detection in distributed database. ACM Computing Surveys, 19(4), 1987.

....Distributed Operating Systems (DOSs) which can be attributed to faster processors and better communication technologies [1,2] A distributed operating system requires distributed algorithms to provide basic operating system functionality like mutual exclusion, deadlock detection, etc. References [3 5] are surveys for algorithms on distributed deadlock detection, distributed mutual exclusion and distributed termination detection. Study of these algorithms is a key topic for students in courses on advanced operating # Correspondence to: Sol M. Shatz, Department of Electrical Engineering and ....

....that preserves the happened before relationship. Of course, this one figure does not prove the correctness of the LLC algorithm, but multiple experiments of this type can be performed to provide more insights and confidence in understanding the behavior of the (1) 2) 3) 1) 1] 4] 5] 6] 7] [3] [4] 6] 7] 2] 3] Terminals ID s Virtual Time (Events) Figure 8. Time sequence graph displayed by Xplot for LLC: a) original plot; b) expanded part (1) protocol. Note that one advantage of the Xplot tool is its ability to let the user isolate segments of a plot and create expanded ....

[Article contains additional citation context not shown here]

Knapp E. Deadlock detection in distributed databases ACM Computing Surveys 1987; 19:303--328.

....in the distributed database systems. Gra78] gives a definition of a distributed wait for graph for the detection of deadlocks, and proposes a centralized protocol to detect the global deadlocks. A detailed study of deadlock detection algorithms for distributed database systems is presented in [Kna87] Timestamp based concurrency control protocols for SDD 1 are discussed in [BRGP78] Sto79] describes the concurrency control protocol used in distributed INGRES. A locking protocol based on limited wait depth is proposed and evaluated in [FHRT93] A detailed discussion on the issues in the ....

Edgar Knapp. Deadlock Detection in Distributed Databases. ACM Computing Surveys, 19(4), 1987.

....i and T j are said to be in conflict, if at least one of the above conflicts exists. It is assumed that two phase locking algorithm is employed for concurrency control and 2PC protocol with centralized communication paradigm is employed for commit processing. Also, deadlock detection and removal [15, 16] is carried out whenever a deadlock occurs in the system. i f s i e c i Figure 1: Processing of a T i in the DDBSs The processing of a T i , in DDBSs is depicted in Figure 1. In this Figure, s i denotes the starting of execution and e i denotes the completion of execution. The data values ....

E.Knapp. Deadlock detection in distributed databases, ACM Computing Surveys, 19(4), December 1987, pp.303-328.

....T j are said to be in conflict, if at least one of the above conflicts exists. It is assumed that two phase locking algorithm is employed for concurrency control and the 2PC protocol with centralized communication paradigm is employed for commit processing. Also, deadlock detection and removal [8] is carried out whenever a deadlock occurs in the system. f s i e c i Figure 1: Processing of a T i in the DDBSs The processing of a T i , in the DDBSs is depicted in Figure 1. In this Figure, s i denotes the starting of execution and e i denotes the completion of execution. The data values ....

E.Knapp, Deadlock detection in distributed databases. ACM Computing Surveys, vol.19, no.4, pp.303-328, December 1987.

....reordering were employed as heuristics to bypass the software bugs. This paper shows that it is possible to guarantee error recovery for more specific types of errors such as deadlocks. Much of the literature on deadlocks has focused on the detection problem for different resource request models [11 13]. After a deadlock is detected, resolution (or recovery) is usually performed by the detecting process simply aborting itself to release the resources it has held. A static priority can be assigned to each process so that the lowest priority process involved in a deadlock cycle becomes the victim ....

....c 1;2 and c 2;3 . All the reachable nodes are marked to indicate that they must be rolled back, and the last unmarked node of each process (fc 0;1 ; c 1;1 ; c 2;2 ; c 3;1 g) forms the desired global checkpoint. 2. 2 Deadlocks and wait for graphs In this paper, we consider the one resource model [11] in which each process can have at most one outstanding resource request at a time, and blocks its execution until the resource is granted. Figure 1(a) shows a resource access pattern in the form of message exchanges, that results in a deadlock. Notations are defined as follows: P j stands for ....

[Article contains additional citation context not shown here]

E. Knapp, "Deadlock detection in distributed databases," ACM Computing Surveys, Vol. 19, No. 4, pp. 303--328, Dec. 1987.

....that is, that a directed cycle exist in G. For the OR model, a necessary and sufficient condition is the existence of a knot in G. A knot is a subset K C N with IKI I having the property that, for all ni K, Di = K. For details on these conditions and related material, the reader is referred to [6, 10] and the references therein. In spite of the existence of several approaches to the detection of deadlocks in distributed computations under the AND OR model (the approaches in [3, 7, 9] are representative recent examples) no graph structure appears to have been identified that accounts for ....

E. Knapp, "Deadlock detection in distributed databases," A CM Computing Surveys 19 (1987), 303-328.

....lock information is managed by a centralized GLM, the maintenance of the WFG is straightforward. The distribution of the GLM or the introduction of authorizations, however, spreads the lock information over multiple nodes, and global or distributed deadlock detection algorithms must be applied [Kna87] 2.4 Buffer Management We introduce buffer management at this point because the design of the buffer manager has great influence on the requirements for coherency control and recovery described in Sections 2.5 and 2.6. The buffer management policy is chosen so that it introduces as little disk ....

E. Knapp. Deadlock detection in distributed databases. ACM Computing Surveys, 19(4):303--328, 1987.

....is, that a directed cycle exist in G. For the OR model, a necessary and sufficient condition is the existence of a knot in G. A knot is a subset K N with jKj 1 having the property that, for all n i 2 K, D i = K. For details on these conditions and related material, the reader is referred to [6, 10] and the references therein. In spite of the existence of several approaches to the detection of deadlocks in distributed computations under the AND OR model (the approaches in [3, 7, 9] are representative recent examples) no graph structure appears to have been identified that accounts for ....

E. Knapp, "Deadlock detection in distributed databases," ACM Computing Surveys 19 (1987), 303--328.

....are all caused by task interaction and once they occur will remain until they are detected and resolved. Therefore, techniques such as diffusing computations[13] and global state detection[6] can be extended and applied to solve these problems. 4 A Hierarchy of Deadlock Models and Ada Knapp[25] classified the deadlock problem into a hierarchy of six models to reflect the complexity of a particular deadlock problem. Each model is characterized by the restrictions that are imposed upon the form resource requests can assume. For example, a task might need to acquire a combination of ....

E. Knapp, "Deadlock detection in distributed databases," ACM Computing Surveys, vol. 19, pp. 303--328, Dec. 1987.

....in the context of distributed database systems is given by Elmagarmid [Elm86] However, many researchers feel that the performance and scalability requirements imposed by large scale distributed systems require to revisit the implementation methods developed so far. In a survey article Knapp [Kna87] classifies distributed deadlock detection algorithms in the following categories: 1) centralized approach [Gra78] by maintaining the global wait for graph, 2) path pushing algorithms [Obe82] by sending parts of the WFG to neighboring sites, 3) edge chasing detection by sending probes, 4) ....

E. Knapp. Deadlock detection in distributed databases. ACM Computing Surveys, 19(4):303--328, December 1987.

....In centralized database systems deadlock detection and resolution has been thoroughly investigated, e.g. in [ACM87] Deadlocks have also been studied in other areas, such as operating systems. Surveys of earlier work on distributed deadlock detection in distributed database systems are given in [Kna87, Elm86, Sin89] The surveys described different algorithms, but no quantitative analysis, in terms of benchmarking, has been carried out. First we describe the two predominant deadlock models underlying locking based database transaction synchronization. Then, the four different distributed ....

....made by transactions, different deadlock models apply. In distributed DBMS the single resource and the AND model are prevailing, so only those will be discussed here. Descriptions of the other models such as the OR model and the general model, which are much less common, are given in [BO81,MC82,Kna87,KS94a,BHRS95] The simplest, and most widely used model in DBMSs is the single resource model. In this model a transaction has only one outstanding request at a time, i.e. it requests a lock on one object, waits until it is granted and only afterwards 4 requests a lock on the next object. ....

[Article contains additional citation context not shown here]

E. Knapp. Deadlock detection in distributed databases. ACM Computing Surveys, 19(4):303--328, December 1987.

....provide. The approaches here vary greatly, and may be grouped into two broad categories. On the optimistic side, one may opt for a somewhat loose priority scheme and risk the loss of those guarantees. In such cases, the loss of safety leads to the need for the capability of detecting deadlocks [11, 17, 23]. The opposing, more conservative side is the side of those strategies which by design guarantee safety and liveness, thereby preventing their loss beforehand. As we demonstrate in the remainder of the paper, both categories give rise to interesting combinatorial structures and properties, ....

E. Knapp, "Deadlock detection in distributed databases," ACM Computing Surveys 19 (1987), 303--328.

....an approach may provide an efficient solution. Such a generalization attempt is not without precedent. Another important problem in distributed systems is distributed detection of global properties. It has been studied in specialized forms such as termination detection [4] and deadlock detection [10]. However, studies in predicate detection [7] have generalized the problem to the detection of arbitrary predicates, provided an understanding of its limitations, and defined classes of predicates for which an efficient general solution is possible. We are aware of two previous studies of the ....

E. Knapp. Deadlock detection in distributed databases. ACM Computing Surveys, 19(4):303 -- 328, December 1987.

....Figure 4. Types of AGVs While packets in communication networks can be removed, vehicles cannot be removed. Problem is how to treat vehicles if the vehicles cannot move due to events like accidents and deadlock. For example, no deadlocked vehicle can be removed although deadlocked transactions [10] can be removed in database systems. In order to resolve the deadlock or make some vehicle release the objects, the vehicle has to move to another object from the current object. Another point is concerned with whether or not and how vehicles can stop. For example, low acceleration AGVs cannot ....

Knapp, E., "Deadlock Detection in Distributed Databases," ACM Computing Surveys, Vol.19, No.4, 1987, pp.303-328.

....relations among the operations. Since the CAD transactions require more objects and take longer time than the conventional ones, there is higher possibility that deadlock occurs, and more information has to be stored in the log. Many researchers have discussed how to detect deadlock as surveyed in [8]. In this paper, we discuss how to resolve deadlock occurring among nested transactions. When the deadlock is detected by using a waitfor graph [8] one transaction in the deadlock cycle is aborted by restoring the old state from the state log [2, 6] It is important to reduce cost for aborting ....

....that deadlock occurs, and more information has to be stored in the log. Many researchers have discussed how to detect deadlock as surveyed in [8] In this paper, we discuss how to resolve deadlock occurring among nested transactions. When the deadlock is detected by using a waitfor graph [8], one transaction in the deadlock cycle is aborted by restoring the old state from the state log [2, 6] It is important to reduce cost for aborting the transaction and storing recovery data in the log. In our method, only a part of the transaction T which is necessary to resolve the deadlock is ....

[Article contains additional citation context not shown here]

Knapp, E., "Deadlock Detection in Distributed Databases," ACM Computing Surveys , Vol.19, No.4, 1987, pp.303-328.

....way by deciding a new path from Kawasaki . This is the re planning. 2 If v is partially open, higher objects still hold the objects although the lower ones are released. 5 Deadlock There might occur deadlock among vehicle transactions. The deadlock is represented by a well known wait for graph [3, 10, 11, 18] where each node denotes a vehicle transaction and each edge v 1 v 2 represents that v 1 waits on an object held by v 2 . Vehicle transactions in a directed cycle of the wait for graph are deadlocked. Example 5.1] For example, there are four vehicles v 1 , v 2 , v 3 , and v 4 which are in ....

Knapp, E., "Deadlock Detection in Distributed Databases," ACM Computing Surveys, Vol.19, No.4, 1987, pp.303-328.

....to be satisfied. Because distributed systems are vulnerable to deadlocks, the problems of deadlock detection and resolution have long been considered important problems in such systems. Existing deadlock detection algorithms can be classified in terms of their underlying resource request models [8], such as AND [2, 4, 5, 13, 14, 15] OR [2, 7, 11, 12] AND OR [6] and p out of q This work was supported in part by Rome Laboratory, Air Force under Contract No. F30602 93 C 0247, by NASA under Grant No. NAGW 4080, and by NSF under Grant No. CDA 9313624. y Department of Computer Science, ....

E. Knapp. Deadlock detection in distributed database. ACM Computing Surveys, 19(4):303 -- 328, December 1987.

.... operational arguments [4, 5, 6] Intuitive operational arguments are prone to errors, and many of the published algorithms have been found to be incorrect [4, 5, 1, 3] Only rigorous proofs, using as few operational arguments as possible, suffice to show the correctness of such algorithms [7]. In a distributed system with no global memory, deadlock cycles can form in innumerable ways, message delays are unpredictable, and deadlock is very sensitive to the timing of resource requests besides. These attributes make a formal correctness proof of deadlock detection resolution algorithms ....

E. Knapp, "Deadlock detection in distributed databases," ACM Comput. Surveys, vol. 19, pp. 303--328, Dec. 1987.

No context found.

Edgar Knapp. "Deadlock Detection in Distributed Databases". ACM Computing Surveys, 19(4), 1987.

No context found.

Edgar Knapp. "Deadlock Detection in Distributed Databases". ACM Computing Surveys, 19(4), 1987.

No context found.

E. Knapp. Deadlock Detection in Distributed Database. Computing Surveys, 19(4):303--328, December 1987.

No context found.

Knapp, E., "Deadlock Detection in Distributed Databases," ACM Computing Surveys , Vol.19, No.4, 1987, pp.303-328.

No context found.

Knapp, E., "Deadlock Detection in Distributed Databases," ACM Computing Surveys, Vol.19, No.4, 1987, pp.303-328.

No context found.

E. Knapp. Deadlock detection in distributed databases. ACM Computing Surveys, 19(4):303--328, December 1987.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC