Barbara Liskov, Miguel Castro, Liuba Shrira, and Atul Adya. Providing persistent objects in distributed systems. In European Conf. on Object-Oriented Programming, June 1999.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....Where InterWeave supports extensive cache reuse, Smart RPC invalidates the cache after each RPC session. Krishnaswamy and Haumacher [21] describe a fast implementation of Java RMI capable of caching objects to avoid serialization and retransmission. Object oriented databases (OODBs) such as Thor [23] allow objects to be cached at client front ends, but they usually neither address heterogeneity problems nor attempt to provide a shared memory programming model. PerDiS [14] is a persistent distributed object system featuring object caching, transactions, security, and distributed garbage ....

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing Persistent Objects in Distributed Systems. In Proc. of the 13th European Conf. on Object-Oriented Programming, pages 230--257, Lisbon, Portugal, June 1999.

....of CORBA s reference model: because everything is an object with no exported data members, every use of a reference parameter incurs a callback. They also suffer from the lack of diffs: small changes to large objects still require large messages. Object Oriented Databases (OODBs) such as Thor [9] allow objects to be cached at client front ends, but they usually neither address heterogeneity nor attempt to support a shared memory programming model. InterAct [10] is an object based system that uses relaxed coherence to support distributed sharing, but requires shared data to be accessed ....

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing Persistent Objects in Distributed Systems. In Proc. of the 13th European Conf. on Object-Oriented Programming, pages 230--257, Lisbon, Portugal, June 1999.

....still require very large messages. LBFS [14] is a low bandwidth network file system that saves bandwidth by taking advantage of commonality between files; InterWeave saves bandwidth by taking advantage of commonality among versions of a segment. Object Oriented Databases (OODBs) such as Thor [12] allow objects to be cached at client front ends, but they usually neither address heterogeneity nor attempt to support a shared memory programming model. InterAct [15] is an object based system that uses relaxed coherence to support distributed sharing, but requires shared data to be accessed ....

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing persistent objects in distributed systems. In R. Guerraoui, editor, ECOOP '99 --- Object-Oriented Programming 13th European Conference, Lisbon Portugal, volume 1628, pages 230--257. Springer-Verlag, New York, NY, 1999.

....Where InterWeave supports extensive cache reuse, Smart RPC invalidates the cache after each RPC session. Krishnaswamy and Haumacher [21] describe a fast implementation of Java RMI capable of caching objects to avoid serialization and retransmission. Object oriented databases (OODBs) such as Thor [23] allow objects to be cached at client front ends, but they usually neither address heterogeneity problems nor attempt to provide a shared memory programming model. PerDiS [14] is a persistent distributed object system featuring object caching, transactions, security, and distributed garbage ....

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing Persistent Objects in Distributed Systems. In Proc. of the 13th European Conf. on Object-Oriented Programming, pages 230--257, Lisbon, Portugal, June 1999.

....of the object implies that the object outlives the client who created it (and other clients) i.e. it will be available to new clients that arrive after its creator has vanished. Implementing persistent shared objects for distributed clients has been the goal of both research efforts (e.g. [LCSA99]) and products alike (e.g. JavaSpaces; see java.sun.com products javaspaces ) This paper appears in DARPA s second DARPA Information Survivability Conference and Exposition (DISCEX II) California, June 2001. School of Computer Science and Engineering, Hebrew University of Jerusalem, ....

B. Liskov, M. Castro, L. Shrira and A. Adya. Providing persistent objects in distributed systems. In Proceedings of the 13th European Conference on Object-Oriented Programming, June 1999.

....area in its attention to scale, and to arbitrary failures, whereas database replication methods traditionally focused on smaller systems with benign failures only. An example of a persistent object replication system that provide transactional guarantees in the face of Byzantine failures is Thor [LCSA99]. Due to its strong guarantee, however, the methods in Thor do not scale well. A weaker guarantee used for replicated data is linearizability [HW90] Linearizability is a condition that realizes atomic (indivisible) operation semantics on replicated objects as de ned by Lamport [Lam86] It is ....

B. Liskov, M. Castro, L. Shrira and A. Adya. Providing persistent objects in distributed systems. In Proceedings of the 13th European Conference on Object-Oriented Programming, June 1999.

....of every normal fetch 30 since an application requests only a single speci c object, but the fetch request to the OR will actually return an entire page of objects. This page will likely have some locality within it and therefore other objects on the page are likely to be useful to the client [15]. For the FE to provide a greater control over the locality of objects, it is possible for the FE to infer the locality of references made by the application prior to disconnect from usage information stored in the cache. Previous work has also been done on this in Thor with hybrid adaptive ....

Barbara Liskov, Miguel Castro, Liuba Shrira, and Atul Adya. Providing persistent objects in distributed systems. In Rachid Guerraoui, editor, ECOOP '99 | Object-Oriented Programming 13th European Conference, Lisbon Portugal, volume 1628, pages 230-257. Springer-Verlag, New York, NY, 1999. 66

....programmers in writing distributed systems. However, from the viewpoint of one process that fails and recovers, those that have continued exhibit persistence. Hence Argus explores persistence and provides a transactional model of persistent state change. Similarly, the later work by Liskov on Thor [140] developing a distributed highly reliable object store, further develops transactional persistence in their programming language Theta [141] The sustained research at MIT on this theme has clearly developed in depth experience with transactional persistent languages, but we are not aware of any ....

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing persistent objects in distributed systems. In Rachid Guerraoui, editor, ECOOP '99 --- Object-Oriented Programming 13th European Conference, Lisbon Portugal, volume 1628 of Lecture Notes in Computer Science, pages 230--257. SpringerVerlag, New York, NY, June 1999. 91

....storage abstraction for applications. Palm Computing s Palm OS [6] does not distinguish between main memory and persistent storage. Its record storage is limited to providing a possibly sorted list of records and thus represents an even lower level of abstraction. Object stores, such as Thor [29], provide a persistent heap of objects. By preserving the structure of application objects, they let applications safely share data. By using transactions, they provide both concurrency control and reliability across failures. However, object stores are optimized for storing heterogeneous objects ....

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing persistent objects in distributed systems. In R. Guerraoui, editor, Proceedings of the 13th European Conference on Object-Oriented Programming, volume 1628 of Lecture Notes in Computer Science, pages 230-257, Lisbon, Portugal, June 1999. Springer-Verlag.

....control policies, they proved difficult to implement efficiently. Of the many store architectures in the literature that do not directly target orthogonal persistence, most seek to address specific optimizations and improvements with respect to OODB design. The THOR architecture from MIT [Liskov et al. 1999] has been the context for OODB research in the areas of buffer management, recovery algorithms, cache coherency and garbage collection. Other projects (e.g. Voruganti et al. 1999] have supported a similar bread of work in the context of an OODB architecture. The system that comes closest to the ....

....has focused on efficient use of buffers in a client server context. Kemper and Kossmann [1994] explored adaptive use of both object and page grain buffering object buffering incurs a copy cost, but may be substantially more space efficient when objects are not well clustered. The THOR project [Liskov et al. 1999] has included a lot of work on buffer management in client server databases, mostly focusing on efficiently managing client, server, and peer memory in the face of object updates at the clients. ANU Store s buffer cache is in most respects orthodox (implementing a STEAL NO FORCE [Franklin 1997] ....

LISKOV, B., CASTRO, M., SHRIRA, L., AND ADYA, A. 1999. Providing persistent objects in distributed systems. In R. GUERRAOUI Ed., ECCOP '99 - Object-Oriented Programming, 13th European Conference, Lisbon, Portugal, June 14-18, 1999, Proceedings, Volume 1628 of Lecture Notes in Computer Science (1999), pp. 230--257.

....specific enough. The authors conclude that the direct reuse of the other object services is impossible given the current object persistence specification and that enhancements are require to the other specifications, e.g. the Compound Externalization Service. 13.4. 2 Research Systems Thor Thor [LCSA99] is a transactional persistent object store that provides access to objects over a wide area, large scale, distributed environment. Thor provides a universe of persistent objects, so the existence of objects is not tied to the running of particular programs. The universe has a persistent root and ....

Barbara Liskov, Miguel Castro, Liuba Shrira, and Atul Adya. Providing Persistent Objects in Distributed Systems. In Rachid Guerraoui, editor, Proceedings of the European Conference on Object-Oriented Programming (ECOOP '99), volume 1628 of LNCS, pages 230--257, Lisbon, Portugal, jun 1999. Springer.

....storage abstraction for applications. Palm Computing s Palm OS [6] does not distinguish between main memory and persistent storage. Its record storage is limited to providing a possibly sorted list of records and thus represents an even lower level of abstraction. Object stores, such as Thor [29], provide a persistent heap of objects. By preserving the structure of application objects, they let applications safely share data. By using transactions, they provide both concurrency control and reliability across failures. However, object stores are optimized for storing heterogeneous objects ....

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing persistent objects in distributed systems. In R. Guerraoui, editor, Proceedings of the 13th European Conference on Object-Oriented Programming, volume 1628 of Lecture Notes in Computer Science, pages 230--257, Lisbon, Portugal, June 1999. Springer-Verlag.

No context found.

Barbara Liskov, Miguel Castro, Liuba Shrira, and Atul Adya. Providing persistent objects in distributed systems. In European Conf. on Object-Oriented Programming, June 1999.

No context found.

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing persistent objects in distributed systems. In Proceedings of the 13th European Conference on Object-Oriented Programming (ECOOP), Lisbon, Portugal, June 1999.

No context found.

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing persistent objects in distributed systems. In Proceedings of the 13th European Conference on Object-Oriented Programming (ECOOP), Lisbon, Portugal, June 1999.

No context found.

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing Persistent Object in Distributed Systems. In Proceedings of the 13th European Conference on Object-Oriented Programming, Lisbon, Portugal, June 1999.

No context found.

Barbara Liskov, Miguel Castro, Liuba Shrira, and Atul Adya. Providing persistent objects in distributed systems. In European Conf. on Object-Oriented Programming, June 1999.

No context found.

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing persistent objects in distributed systems. In European Conference for Object-Oriented Programming (ECOOP), June 1999.

No context found.

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing Persistent Object in Distributed Systems. In Proceedings of the 13th European Conference on Object-Oriented Programming, Lisbon, Portugal, June 1999.

No context found.

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing persistent objects in distributed systems. In Proc. ECOOP '99, Lisbon, Portugal, June 1999.

No context found.

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing persistent objects in distributed systems. In European Conference for Object-Oriented Programming (ECOOP), June 1999.

No context found.

B. Liskov, M. Castrol, L. Shrira, and A. Adya. Providing Persistent Object in Distributed Systems. In Proceedings of the 13th European Conference on ObjectOriented Programming, Lisbon, Portugal, June 1999.

No context found.

B. Liskov, M. Castrol, L. Shrira, and A. Adya. Providing Persistent Object in Distributed Systems. In Proceedings of the 13th European Conference on ObjectOriented Programming, Lisbon, Portugal, June 1999.

No context found.

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing persistent objects in distributed systems. In European June 1999.

....However, in the absence of ownership types, the system has to depend on either informal reasoning by programmers or some form of conservative formal veri cation to check that transform functions are well behaved. We have implemented a prototype lazy modular upgrade infrastructure in Thor [41, 9], a highly optimized objectoriented database. The paper describes the prototype and discusses the design trade o s we made to optimize performance in the common case. The paper presents performance results that indicate that the infrastructure has low cost. It has negligible impact on applications ....

....objects and states of new and modi ed objects to one of the servers. The server decides whether the transaction can commit (using two phase commit if the transaction used objects at more than one server) and informs the client machine of its decision. More information about Thor can be found in [41, 19, 1, 9]. 5.1 Installing Upgrades Upgrades are installed by interacting with one of the servers. This server checks the upgrade for completeness, and determines whether Condition E holds. If E doesn t hold for some TF, the server interacts with the user, which may result in a trigger being added to the ....

[Article contains additional citation context not shown here]

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing persistent objects in distributed systems. In European Conference for Object-Oriented Programming (ECOOP), June 1999.

....later and shared with other applications. The database acts as an extension of an object oriented programming language such as Java, allowing programs access to long lived objects in a manner analogous to how they manipulate ordinary objects whose lifetime is determined by that of the program [37, 24, 13, 43, 12, 5]. The objects stored in an OODB may live a long time and as a result there may be a need to upgrade them, that is, change their code and storage representation. An upgrade can improve an object s implementation, to make it run faster or to correct an error; extend the object s interface, e.g. by ....

....T we might encounter an object that has pending transforms. If the pending transform function is for T s upgrade or a later upgrade, we do not do anything. Otherwise, we interrupt T (just as we interrupted A) to run the pending transform function. We implemented this approach in the Thor OODB [37, 7]; the implementation is described in [38] 2.3 Semantics of Upgrades As we mentioned in the introduction, an upgrade system should guarantee that when a transform function runs, it encounters only interfaces that existed at the time its upgrade was installed and states that satisfy its object s ....

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing persistent objects in distributed systems. In European Conference for Object-Oriented Programming (ECOOP), June 1999.

....later and shared with other applications. The database acts as an extension of an object oriented programming language such as Java, allowing programs access to long lived objects in a manner analogous to how they manipulate ordinary objects whose lifetime is determined by that of the program [37, 24, 13, 43, 12, 5]. The objects stored in an OODB may live a long time and as a result there may be a need to upgrade them, that is, change their code and storage representation. An upgrade can improve an object s implementation, to make it run faster or to correct an error; extend the object s interface, e.g. by ....

....T we might encounter an object that has pending transforms. If the pending transform function is for T s upgrade or a later upgrade, we do not do anything. Otherwise, we interrupt T (just as we interrupted A) to run the pending transform function. We implemented this approach in the Thor OODB [37, 7]; the implementation is described in [38] 2.3 Upgrade Semantics As we mentioned in the introduction, an upgrade system should guarantee that when a transform function runs, it encounters only interfaces that existed at the time its upgrade was installed and states that satisfy its object s ....

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing persistent objects in distributed systems. In European Conference for Object-Oriented Programming (ECOOP), June 1999.

....the order they are placed in the array) Triggers are also supposed to be well behaved: they should be pure observers. 3. IMPLEMENTATION This section describes how we implement upgrades within the Thor object oriented database. More information about Thor and its implementation can be found in [19, 10, 1, 7]. Thor is a client server system. Persistent objects reside at servers; each persistent object resides at a particular server. Application transactions run at client machines on cached copies of the persistent objects. Thor uses optimistic concurrency control [1] Client machines fetch objects ....

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing Persistent Objects in Distributed Systems. In Proceedings of the 13th European Conference for Object-Oriented Programming (ECOOP), 1999.

....the order they are placed in the array) Triggers are also supposed to be well behaved: they should be pure observers. 3. IMPLEMENTATION This section describes how we implement upgrades within the Thor object oriented database. More information about Thor and its implementation can be found in [19, 10, 1, 7]. Thor is a client server system. Persistent objects reside at servers; each persistent object resides at a particular server. Application transactions run at client machines on cached copies of the persistent objects. Thor uses optimistic concurrency control [1] Client machines fetch objects ....

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing Persistent Objects in Distributed Systems. In Proceedings of the 13th European Conference for Object-Oriented Programming (ECOOP), 1999.

....introducing software errors. Since all replicas share this code this induces an undesirable correlation in replica failure. 46 Chapter 5 Example II: Object Oriented Database The methodology described in Chapter 3 was applied to the Thor object oriented database management system (OODBMS) [36, 35]. In this example, the diversity in the replicas is not obtained from using distinct implementations, but from using a single, non deterministic implementation. The first section in this chapter is an overview of the Thor system. Sections 5.2, 5.3, and 5.4 describe the three main steps in the ....

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing persistent objects in distributed systems. In Proceedings of the 13th European Conference on Object-Oriented Programming (ECOOP '99), Lisbon, Portugal, June 1999.

....fault tolerance as discussed in Section 2. Additionally, it can improvedisk localityby clustering blocks from the same le and les that are in the same directory. 3. 2 Object Oriented Database We have also applied our methodology to replicate the servers in the Thor object oriented database [18]. In this example, all the replicas run the same server implementation. The example is interesting because the service is more complex than NFS, and the server implementation is multithreaded and exhibits a signi cant degree of nondeterminism. The methodology enabled reuse of the existing server ....

....enabled reuse of the existing server code and could enable software rejuvenation through proactiverecovery. Webeginby giving a brief overview of Thor and then describe how the methodology was applied in this example. A more detailed description can be found in [32] 3.2. 1 System Overview Thor [18] provides a persistent object store that can be shared by applications running concurrently at di erent locations. It guarantees type safe sharing by ensuring that all objects are used in accordance with their types. Additionally, itprovides atomic transactions [14] to guarantee strong consistency ....

[Article contains additional citation context not shown here]

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing persistent objects in distributed systems. In Proceedings of the 13th European Conferenceon Object-OrientedProgramming (ECOOP '99), Lisbon, Portugal, June 1999.

No context found.

Barbara Liskov, Miguel Castro, Liuba Shrira, and Atul Adya. Providing persistent objects in distributed systems. In European Conference for Object-Oriented Programming (ECOOP), June 1999.

No context found.

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing Persistent Objects in Distributed Systems. In Proc. of the 13th European Conf. on Object-Oriented Programming, pages 230--257, Lisbon, Portugal, June 1999.

No context found.

Barbara Liskov, Miguel Castro, Liuba Shrira, and Atul Adya. Providing persistent objects in distributed systems. In European Conf. on Object-Oriented Programming, 1999.

No context found.

B. Liskov, M. Castro, L. Shrira, and A. Adya. Providing persistent objects in distributed systems. In R. Guerraoui, editor, ECOOP '99 --- Object-Oriented Programming 13th European Conference, Lisbon Portugal, volume 1628, pages 230--257. Springer-Verlag, New York, NY, 1999.

No context found.

Barbara Liskov, Miguel Castro, Liuba Shrira, and Atul Adya. Providing persistent objects in distributed systems. In European Conference for Object-Oriented Programming (ECOOP), June 1999.

No context found.

Barbara Liskov, Miguel Castro, Liuba Shrira, and Atul Adya. Providing persistent objects in distributed systems. In European Conference for Object-Oriented Programming (ECOOP), June 1999.

No context found.

Barbara Liskov, Miguel Castro, Liuba Shrira, and Atul Adya. Providing persistent objects in distributed systems. In European Conf. on Object-Oriented Programming, 1999. 11

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