D.P. Reed. Naming and Synchronization in a Decentralized Computer System. PhD thesis, MIT, Dept. of Electrical Engineering, 1978.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... are in many cases constructed not as part of any particular algorithm but for direct use as the building blocks of various storage and retrieval mechanisms, such as distributed dictionaries, name servers in communication networks, bulletin boards, resource allocation managers and the like [108,95,49,85,116,90,65]. The function common to all of these mechanisms is supplying facilities for storing accumulated information in the system and making it available to potential users throughout the system. The topic of distributed data structures is rather wide, and the distributed setting raises several issues ....

D.P. Reed. Naming and Synchronization in a Decentralized Computer System. PhD thesis, MIT, Dept. of Electrical Engineering, 1978.

....CFL 82] prevent update transactions from conflicting with read only transactions by providing the latter with a consistent but somewhat out of date view of the database. In order to provide this view, multiple versions of recently updated data items are retained. Early multi version schemes [Ree78] used timestamps for readers and writers, but more recent multi version locking schemes [CFL 82, AS89, BC92a, MPL92] use timestamps with read only transactions, allowing them to use old versions without locking, while requiring updaters to perform locking. However, none of the above techniques ....

....respectively. We present the design of our logical version manager in Section 4. In Section 5, we discuss related work, and in Section 6, we give our conclusions and directions for future work. 2 Logical Versioning The idea of maintaining multiple versions of an item was first proposed by [Ree78] and is known as multi versioning or just versioning. In this paper, we refer to this as logical versioning to differentiate it from physical versioning, which is described further in Section 3. In this section, we describe the basic structure of our multi version locking scheme [CFL 82, AS89, ....

D. P. Reed. Naming and synchronization in a decentralized computer system. Technical Report MIT-LCS-TR-205, Massachusetts Institute of Technology, Cambridge, September 1978.

....of [9] ensure serial correctness for non orphans. We are also currently working on generalizing the results in [20] to nested transaction systems. This will permit us to show that many other kinds of objects ensure serial correctness, including objects that use timestamps for concurrency control [16], and objects that use more general approaches to locking f5, 17, 20] The results in this paper indicate that the orphan elimination algorithms analyzed here can be combined wth any of these objects. 7The only difference is that the CREATE input operation ha nother name in 15 4. Information ....

Reed., D.P. Naming and synchronization in a decentralized computer system. PhD thesis, Massachusetts Institute of Technology, 1978. Available as Technical Report MIT/LCS/TR-205.

....generated when transactions commit, so the order of items in the queue can be determined for subsequent dequeues. In this paper, we show how Herlihy and Weihl s algorithm can be extended to accommodate nested transactions. Nested transactions have been explored in a number of projects (e.g. [12, 11,9, 3, 1]) for building reliable distributed systems. In a nested transaction system, a transaction can have subtransactions, each of which appears to run atomically within the transaction. Thus, concurrent subtransactions are serializablethey appear to run in some serial order and recoverable they appear ....

D.P. Reed, Naming and synchronization in a decentralized computer system, Ph.D. Thesis, Massachusetts Institute Technology, 1978.

....5, the characterization theorem is stated and proved. Section 6 contains discussion of the possible uses of the characterization theorem for concurrency control design. Section 7 contains discussion of the relationship of multilevel atomicity to the nested transaction model of [8, 9, 11] and [13]. Much work remains to be done in designing and evaluating concurrency control algorithms for multilevel atomicity. It remains to be seen whether new concurrency control algorithms which achieve multilevel atomicity can be made to operate much more efficiently than existing concurrency control ....

....and B of ti l with B e a This means that it is quite plausible that a rollback of steps of ti= can cause a rollback of steps of ti, and so on. 7. DISCUSSION 7.1. Nested Transactions It is interesting to compare multilevel atomicity to the atomicity achieved by the nested transaction model [9, 11, 13]. The latter model permits transactions to be nested, and then requires serializability of transactions at every level, including the top level. The remainder of this subsection assumes familiarity with some of the work on nested transactions. At first glance, it appears that the nested ....

[Article contains additional citation context not shown here]

REED, D. P. Naming and synchronization in a decentralized computer system. PhD Dissertation, Tech. Rep. MIT/LCS/TR-205, MIT, Laboratory for Computer Science, Cambridge, Mass., 1978.

.... method [Lausen 1983, Herlihy 1986] studied alternative validation methods [Harder 1984] and compared optimistic and pessimistic techniques [Agrawal 1983, Badal 1981, Carey 1983, Franaszek Robinson 1985, Menasce Nakanishi 1982, Tay et al. 1984] Nested actions are proposed in [Davies 1978, Reed 1978] Several approaches to us ing pessimistic concurrency control for nested actions have been studied. Read write locking for nested actions is described in [Moss 1981] and has been implemented by systems such as Argus [Liskov 1984, Liskov et al. 1987a] and Camelot [Spector et al. 1987, Spector i: ....

D. P. Reed. Naming and Synchronization in a Decentralized Computer System. Technical Report 205, MIT Laboratory for Computer Science, Cambridge, MA, 1978.

....distributed computing systems is maintaining data integrity in the presence of concurrency and failures. Atomic transactions, or actions, are a widely accepted solution to these two problems; they are serializable and recoverable, thus hiding concurrency and failures. Nested transactions ([Reed 1978, Moss 1981, Liskov Scheifler 1983] are a generalization of the model of atomic transactions. Nested transactions, or subactions, provide a uniform mechanism for coping with failures and obtaining concurrency within an action. Execution of actions in a nested action model is synchronized in a ....

....records from an up to date replica and write those records to the archaic one. 1. 2 Related work The idea of nested atomic actions was initially proposed, as spheres of control, in [Davies 1973, Bjork 1973] The first detailed design for a model that uses nested atomic actions was developed by Reed ( Reed 1978]) Reed proposed a multi version, timestamp based algorithm to ensure serialization of concurrent actions. A locking based model of a nested atomic action system was developed by Moss ( Moss 1981] Several research projects have designed and implemented a nested atomic action system: Argus ....

[Article contains additional citation context not shown here]

D.P. Reed. Naming and synchronization in a decentralized computer system. Ph.D. thesis, Massachusetts Institute of Technology, 1978. Available as Technical Report MIT/LCS/TR-205.

....a new transaction without danger of inconsistency arising from concurrent access to the same data object by two or more of the composed transactions. 1.2 Points of Novelty. Our nested transaction system uses locking for synchronization, and is (to our knowledge) the first design to do so. Reed treed78] presented the first comprehensive design of a nested transaction system, but his design uses timestamps for synchronization. Our System and his deal with deadlock in different ways, too. The idea of nested transactions seems to have originated with Davies IDavies73] some time ago, but we know ....

.... restoration (transaction undo) algorithm is also based on methods used in many systems (e.g. Gray78, IMS78, SMB79, Paxton79, LeLann81] The distributed transaction management algorithm is original, except in its incorporation of a two phase commit protocol [Gray78, Gray80a, Paxton79, IMS78, Reed78, LeLann81, Lindsay79, HS80] The distributed deadlock detection algorithm is an extension of work reported in [Goldman77] and [Obermarck80] Some of the methods of presentation that we have used derive from other specific works. In particular, the system model and failure model of Chapter 2 was ....

[Article contains additional citation context not shown here]

David P. Reed, "Naming and Synchronization in a Decentralized Computer System", Ph.D. Thesis, M.I.T. Dept. of Elec. Eng. and Comp. Sci., available as M.I.T. Lab. for Comp. Sci. Technical Report 205, Sept. 1978.

....protocol augmented with adaptive broadcast as they do with an update protocol. 4 Related Work The idea of using version numbers for consistency and concurrency control is fairly old. The NAMOS system used a version based scheme for concurrency control and consistency in a distributed database[4]. The Sprite file system also used a consistency scheme based on version numbers[3] Both of these systems, like the Jade implementation, present the abstraction of mutable data and use version numbers as an internal implementation mechanism. SAM [7] and VDOM [1] two more recent systems in the ....

D. Reed. Naming and Synchronization in a Decentralized Computer System. PhD thesis, Dept. of Electrical Engineering and Computer Science, Massachusetts Institute of Technology, September 1978.

....to handle. Much of the TMF [Borr 81] implementation is within the operating system, although there is no support for nesting. Transactions which are composed of other transactions, or nested transactions, have been the subject of much current literature [Moss 81] Liskov 82] Reed 78] Svob 81] A transaction invoked from within a transaction, or subtransaction, appears atomic to its caller. That is, the operations it performs take place indivisibly with respect to both failures and concurrent computations, just as for traditional transactions. Thus a nested transaction ....

Reed, D. P., "Naming and Synchronization in a Decentralized Computer System", Technical Report MIT/LCS/TR-205, Laboratory for Computer Science, M.I.T., 1978.

....by allowing multiple transactions to read and write different versions of the same data item, as long as each transaction sees a consistent set of versions for all the data items that it accesses. This is the basic idea of the first multiversion timestamp ordering scheme introduced by Reed [Reed 78] In Reed s mechanism, each transaction is assigned a unique timestamp when it starts; all operations of the transaction are assigned the same timestamp. In addition, each data item x has a set of transient versions, each a write timestamp, value pair, and a set of read timestamps. If a ....

.... of nested spheres of control, which is the origin of the nested transactions concept, was first introduced by Davies [Davies 73] and expanded by Bjork [Bjork 73] Reed presented a comprehensive solution to the problem of composing transactions by formulating the concept of nested transactions [Reed 78] A nested transaction is a composition of a set of subtransactions; each subtransaction can itself be a nested transaction. To other transactions, only the top level nested transaction is visible and appears as a normal atomic transaction. Internally, however, subtransactions are run ....

Reed, R. Naming and Synchronization in a Decentralized Computer System. Ph.D. Thesis, M.I.T. Laboratory of Computer Science, September 1978. MIT Technical Report 205

....mechanism for user defined lock types. 6. RELATED WORK I know of only one integrated environment that supports a transaction model, the Cosmos Eclipse environment [6] at the University of Lancaster. The transient versions and time domain addressing used for Reed s multiple version implementation [7] of serializable transactions is replaced in Cosmos by immutable versions and domain relative addressing on configurations and configuration histories. The primary disadvantage of this scheme is the nonserializability of the committed transactions. Commit serializability avoids this disadvantage, ....

Reed, D.P. Naming and Synchronization in a Decentralized Computer System. PhD thesis, MIT, September, 1978. MIT LCS TR-205.

....in terms of nested transaction protocols. Locking is the most used technique for concurrency control protocols. The other two techniques are timestamps and graph testing. Such techniques are also employed to derive protocols that ensure conflict serializability of nested transactions. The work of [Ree78] describes a protocol for nested transactions that uses timestamp technique. The work of [Mos81] describes a protocol for nested transactions that uses lock techniques. Hadzilacos and Hadzilacos [HH91] show proofs of correctness for these two protocols. The work of [Jag90] describes 27 a ....

D. P. Reed. Naming and Synchronization in a Decentralized Computer System. PhD thesis, Massachusetts Institute of Technology, 1978.

....its modifications do not conflict with any uncommitted transaction s reads writes. With a backward scheme, T 2 commits successfully since it does not conflict with an already committed transaction; similarly, T 1 is also allowed to commit later. 2.4. 3 Multi version schemes Multi version schemes [Ree78, BHG87, BBG 95, Wei87] allow multiple versions of the same object to exist in the database state. Thus, in a history such as H 2 , when transaction T 2 tries to read object y, it can be provided with an old version of y (and not the latest version) resulting in a serializable history. ....

....where a read only transaction reads object versions according to its timestamp requires access times to be stored with an object, thereby converting every read into a write. This information is needed so that later update transactions can be assigned a higher timestamp. The scheme suggested in [Ree78] also has a similar requirement. In another scheme suggested by Weihl, each transaction T i maintains dependency and antidependency information and stores it with the versions that T i creates. A read only transaction T r uses this information to determine which object versions to read. However, ....

D. P. Reed. Naming and Synchronization in a Decentralized Computer System. Technical Report MIT/LCS/TR-205, Laboratory for Computer Science, MIT, Cambridge, MA, 1978.

.... Manager Buffer Manager Data Access Item B T CT BT AI AT PC CT AI AT BT AT BT Begin Trans CT Commit Trans AT Abort Trans AI Access Item PC Prepare to Commit Figure 1: System Architecture Example 1: Assume that there are two CCMs, A and B, implementing Basic Timestamp Ordering (BTO) [Ree78] and Optimistic Concurrency Control, respectively, and that the MCC implements 2PL. Consider the following schedule, where T 1 A and T 2 A are transactions submitted to CCM A and TB is a transaction submitted to CCM B: T 1 A T 2 A TB Begin R(X) Begin R(X) W(X) R(Y) W(Y) Begin R(Y) ....

D. Reed. Naming and Synchronization in a Decentralized Computer System. Ph.D. Thesis , Massachusetts Institute of Technology, 1978.

....database is the serializability of transactions [Pap79, BHG87] This calls for the property of as if sequential execution of transactions. For most updates in an ATIS database, serializability can be enforced by either the two phase locking rule [EGLT76] or the time stamp ordering rule [Ree78] In the context of the above query update situation, where we do not require an absolutely shortest alternative route, it is possible to relax the correctness condition of serializability to bounded ignorance [KB91] or bounded inconsistency [WA92] This allows a greater degree of concurrency and ....

D. P. Reed. Naming and Synchronization in a Decentralized Computer System. Technical Report MIT-LCS-TR-205, Massachusetts Institute of Technology, Cambridge, Massachusetts, September 1978.

....Fuji90a, Fuji90b] Time Warp, in particular, is one of the more successful optimistic protocols used for PDES. A similar protocol, which has been found to have good performance [Mill86, Mill92, Mill94] enforcing serializability in parallel systems, is the Multiversion Timestamp Ordering protocol [Reed78, Reed83]. Protocols such as Multiversion Timestamp Ordering and Time Warp have the following advantages: 1) high effective concurrency levels, 2) deadlock free operation (because there is no or limited blocking) and (3) efficient operation unless there is a conflict. These types of protocols resolve ....

D.P. Reed, Naming and Synchronization in a Decentralized Computer System, Ph.D. Thesis, MIT, September 1978.

....is adopted in this paper. Serializability guarantees that an execution of transactions must be equivalent to a serial execution of the transactions. Serializability can be enforced by employing a concurrency control protocol such as the two phase locking protocol [8] or timestamp ordering protocol [15]. Since a data item may be replicated on several servers, these copies of a data item must appear as a single logical data item to transactions. This property is called one copy equivalence and is enforced by a replica control protocol such as the quorum protocol [10] By using a quorum protocol, ....

D. P. Reed. Naming and Synchronization in a Decentralized Computer System. Technical Report MIT-LCS-TR-205, Massachusetts Institute of Technology, Cambridge, Massachusetts, September 1978.

....proven to perform correctly under the assumption that the MCC employs 2PL and that CCMA and CCMB are either 2PL or OPT. However, as we will next show, the algorithm may not work under all situations. For example, the algorithm fails if either CCMA or CCMB uses basic timestamp ordering (BTO) Ree78] Consider the following schedule T1A T1B T2A R(X) R(X) W(X) R(Y) W(Y) Commit R(Y) W(Z) Commit R(Z) Commit It is clear that the above schedule is not serializable (due to the T2A T1A T1B T2A cycle) However, it would be allowed by a combination of ....

D. Reed. Naming and Synchronization in a Decentralized Computer System. PhD thesis, Massachusetts Institute of Technology, 1978.

....[Gra78] the results of all its operations. A commonly accepted correctness criterion in databases is the serializable execution of transactions [EGLT76] Serializable executions are guaranteed by employing a concurrency control mechanism, e.g. locking [EGLT76, SK80, AE90] timestamp ordering [Ree78] or optimistic concurrency control [KR81] protocols. Since strict twophase locking 1 is widely used, we assume in this paper that concurrency control is locally enforced by strict two phase locking at all database sites. In a replicated database, copies of an object may be stored at several ....

D. P. Reed. Naming and Synchronization in a Decentralized Computer System. Technical Report MIT-LCS-TR-205, Massachusetts Institute of Technology, Cambridge, Massachusetts, September 1978.

....and defines its key concepts. Section 4 presents an architecture for our model, and details its components. Finally, some concluding remarks and future work is described in Section 5. 2 Related Work Using multiple versions of data items for transaction synchronization was first proposed by Reed [Ree78] and Bayer, et al.: BEHR80] Multiversioning allows for enhanced concurrency, simplifies recoverability, and supports temporal data management. The rest of this section reviews multiversion and object base concurrency control literature. 2.1 Multiversion Concurrency Control Two phase locking and ....

D. Reed. Naming and synchronization in a decentralized computer system. Technical Report MIT/LCS/TR-205, MIT Laboratory for Computer Science, 1978.

....at the disks. These policies utilize the priority ordering established by the mappings described in the previous section in a straightforward manner. For implementing concurrency control, however, several different mechanisms are available, including locking (e.g. Gray79] timestamps (e.g. [Reed78]) and optimistic concurrency control (e.g. Kung81] In this section, we describe the concurrency control algorithms that were chosen for evaluation in this study. These algorithms are a subset of those that were investigated in our earlier studies on the performance of concurrency control ....

Reed, D., "Naming and Synchronization in a Decentralized Computer System," Ph.D. Thesis, Dept. of Computer Science, Massachusetts Institute of Technology, 1978.

No context found.

Reed, D.P., [1978], "Naming and Synchronization in a Decentralized Computer System", Technical Report 205 (Ph.D. Thesis), M.I.T. Dept. of Electrical Engineering and Computer Science, September.

No context found.

REED78 Reed D. P., Naming and Synchronization in a Decentralized Computer System, PhD dissertation, Department of Electrical Engineering, Massachusetts Institute of Technology, Sep 1978.

No context found.

Reed D., "Naming and Synchronization in a decentralized computer systems", Ph.D. Dissertation, Dept. Elec. Eng. Comp. Sc., MIT Sept. 1978

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