Lampson, B.W., "Atomic Transactions", Distributed Systems -- Architecture and Implementation: An Advanced Course. Chapter 11.Lecture Notes in Computer Science #105, Springer-Verlag, 1981.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....Obj85 Obj86 Oops86 Stef85 Stoy84] Transactions: The notion of transactions and atomic actions, particularly nested transactions, is relevant to environments with many concurrent active objects. Nested transactions are available in the Argus system. See also papers on concurrency and reliability. [Gray81 Lamp81 Moss81 Moss82 Moss83] Types: Various papers dealing with data types. See also papers on data abstraction. Alle86 Bruc86 Card85b Grie77 Inga86 Wall80 Zdon86b] UIMS: How object oriented ideas can be applied in the development of user interfaces. Fium83] 5 Unreferenced. The following list includes systems and work ....

B.W. Lampson, "Atomic Transactions", in Distributed Systems -- Architecture and Implementation, ed. B.W. Lampson, M. Paul and H.J. Siegert, Lecture Notes in Computer Science 150, pp. 246-265, Springer-Verlag, 1981.

....not assumed; instead, state information is replicated at other processors. However, this turns out to be just an approximation of the Stable Storage Property. The only work we know of that does not involve fail stop or stronger assumptions about processor failures is described in [6] 7] and [9]. ACM Transactions on Computer Systems, Vol. 2, No. 2, May 1984. Real processors do not satisfy the Halt on Failure, Failure Status, or Stable Storage properties. In fact, most real processors are not even good approximations of fail stop processors. This is disappointing in light of the number ....

LAMPSON, B. Atomic transactions. In Distributed Systems--Architecture and Implementation. Lecture Notes in Computer Science, vol. 105, Springer-Verlag, New York (1981), pp. 246-265.

....such Web service, it will either commit or abort, and if it commits, its effects cannot be undone. The second level consists of Web services that can be aborted or compensated. There are two traditional ways to abort or compensate a previous executed service. One way, named two phase commit (2PC) [8], is based on the idea that no constituent transaction is allowed to commit unless they are all able to commit. Another way, called compensation, is based on the idea that a constituent transaction is always allowed to commit, but its effect can be cancelled after it has committed by executing a ....

Lampson, B. W., "Atomic Transactions". In: Goos, G., Hartmanis, J. (eds.), Distributed Systems - Architecture and Implementation: An Advanced course, Spring-Verlag, pp. 246265, 1981

....mechanism is its commit p otool, which in a distributed system is a protocol that ensures that the nodes participating in an action either all commit or all abort. To commit top level actions, our models use a variation of a standard two phase commit protocol [Gray 1979, Mohan ; Lindsay 1983, Lampson 1981, Lindsay et al. 1979, Lindsay et al. 1984] Non blocking and three phase commit protocols have also been studied [Skeen 1981, Dwork i; Skeen 1983] Two phase commit and the other protocols mentioned above ensure that all partici In a timestamping approach, an action s serialization order is ....

....out of order. We assume that failures are eventually repaired: nodes eventually recover from crashes, and partitions are eventually mended. Nodes have access to both volatile and stable storage. Volatile storage is lost in a crash, while stable storage is intact upon the recovery of a node [Lampson 1981]. 2.2 High level System At a higher level of abstraction, we view a distributed system as a collection of active communicating entities, where each entity resides at a single node in the network. Entitles are resilient; with high probabi]ity, they survive crashes of their nodes (this is done, ....

[Article contains additional citation context not shown here]

Butler Lampson. Atomic transactions. In B. W. Lampson, M. Paul, and H. J. Siegert, editors, Distributed Systems -- Architecture and Implementation, chapter 11, pages 247--265, Springer-Verlag, 198]. Lecture Notes in Computer Science series, mmber 10,5.

....errors originated in the disk drive and controller are partly handled by the hardware itself; modern storage systems are equipped with powerful coding mechanisms to detect and correct decay errors. To tolerate soft errors during disk access we employ the careful disk operations, as described in [14]. A careful read operation repeatedly performs a normal disk read until it gets a good status or a predefined limit of retries is exceeded. This eliminates soft read errors. A careful write operation repeatedly performs a normal disk write followed by a read until it returns a good status with the ....

B. Lampson, "Atomic transactions," in Distributed Systems-Architecture and Im- plementation, vol. 105 of Lecture Notes in Computer Science, pp. 246-265, New York, NY: Springer-Verlag, 1981.

....to represent replicated data. We make use of timestamp generation schemes due to Lamport [Lainport 78] Dubourdieu [Dubourdieu 82] and Chan et al. Chan 82] Logs have been used to facilitate recovery in databases [Gray 78] and as a technique for achieving reliably coordinated updates [Lampson 79] In the Swallow .repository [Reed 80] an object is represented by a sequence of versions, a mechanism with many of the same properties as a log. Replication methods for files have been proposed that relax the requirement that the value read from a file should be the last value wdtten. ....

Lampson, B., "Atomic transactions", Distributed Systems: Architecture and Implementatio n, Lecture Notes in Computer Science 105, Goos and Hartmanis editors, Springer. Verlag, Bedin, 1981, 246.265.

....limitations. However, the advent of machines such as the DEC Alpha [41] and the MIPS R4000 [18] which (logically) support a 64 bit address has created renewed interest in this approach. A number of research groups have suggested that this is an appropriate direction for modern operating systems [22]. Such an approach is tempting since it fits in well with the goals of orthogonal persistence, i.e. to abstract over all physical attributes of data. However, there are some difficulties: Most persistent systems rely upon a checkpointing mechanism to establish a consistent state on stable storage ....

....entities stored in the system so that dependency information can be maintained. ii. If a single address space is shared by multiple processes, the ability to protect separate areas of the address space must be provided. Whilst protection systems have been designed for single address spaces [22], they do not provide any support for distribution. iii. The resulting store would be huge and the management of large stores is difficult. In particular allocation of free space, garbage collection, unique naming of new objects and the construction of appropriate navigation tools are all more ....

[Article contains additional citation context not shown here]

Lampson, B. "Atomic Transactions", Distributed Systems - Architecture and Implementation, Lecture Notes in Computer Science, vol 105, Springer-Verlag, Berlin, pp. 246-265, 1981.

....atomic transactions despite the effects of concurrency and distribution, as usual [2] and partial failures of its components, notably disks and communication media. Disks may: corrupt the block written, fail to write a value, respond with a corrupted read value or corrupt the block which is read [10]. Media may: loose, replicate, corrupt or create messages [17] Before we apply such devices in the design, we add enough layers of software to mask the corresponding faults. This is done incrementally i.e. one fault at a time; there are four steps for each device. Such steps which strengthen ....

B.W. Lampson. Atomic transactions. In M. Paul B.W. Lampson and H.J. Siegert, editors, Distributed Systems - Architecture and Implementation, pages 246--265. Springer-Verlag, 1981.

....10 . Fault tolerance is (also) a recursive concept; the mechanisms designed to tolerate faults must be protected against the faults likely to affect them. Examples are given by the replication of voters, self checking controllers [Carter Schneider 1968] through the notion of stable memory [Lampson 1981] in recovery data and programs. Fault tolerance is not limited to accidental faults. Protection against intrusions has long relied on cryptography (e.g. see [Denning 1982] for an early overview) In particular, encryption can be viewed as a form of tolerance in that ciphered information can be ....

B. W. Lampson, "Atomic Transactions", in Distributed Systems --- Architecture and Implementation, (B. W. Lampson, Ed.), Lecture Notes in Computer Science, 105, pp.246-65, Springer-Verlag, Berlin, Germany, 1981.

....delay. Stable storage [21, Section 3.3] ensures that read and write operations on the storage are always successful, even if the underlying storage hardware components fail. This is useful for fault tolerance methods requiring that some state of the system be available after a failure occurs. [27] denes the problem of the stable storage and gives a protocol that works with one disk in the presence of arbitrary faults. Other classical protocols are disk shadowing and Redundant Arrays of Inexpensive Disks (RAID) Fail stop processors [21, Section 3.4] is the assumption made on the behavior ....

B.W. Lampson. Atomic transactions. In B.W. Lampson, M. Paul, and H.J. Siegert, editors, Distributed Systems-Architecture and Implementation, volume 105 of LNCS, pages 246265. Springer Verlag, 1981.

....9 the possibility of a collection of operations either succeeding or failing atomically. This layering of failure guarantees is in turn essential in handling the complexity of building fault tolerant applications. Various mechanisms have been proposed for achieving atomic failure. Transactions [30] have been used very successfully in databases to make compound database updates atomic. Atomicity here is used in two respects: an update either succeeds or fails atomically, and it is not possible to observe the interleaving of two updates. Transaction correctness is normally based on ....

Butler Lampson. Atomic transactions. In B. Lampson, M. Paul, and H. Siegert, editors, Distributed Systems--Architecture and Implementation, volume 205 of Lecture Notes in Computer Science, pages 246--285. Springer-Verlag, 1981.

....network failures are more common than client crashes so it is valuable to arrange the le system operation in a way that means that the integrity of the le system is maintained, even when a client operation fails. To do this the block based network le system uses a modi ed atomic update scheme[Lampson, 1981] that we call Unstable Atomic Update . In this algorithm the server delays all write operations until the end of a le system transaction. When the transaction completes the server writes all the modi ed blocks to the disk (and releases any leases associated with the blocks) If the network ....

....executed and the le system might be left in an inconsistent state. It is this violation of the atomic nature of a transaction when a server failure occurs that gives Unstable Atomic Update its name. It would be possible to protect the le system against server failure by using stable storage[Lampson, 1981]. However stable storage require that at least three extra write operations occur for each write operation placed in the intentions list and is based on an assumption that is dicult to meet without writing the intentions list to two di erent disk drives concurrently. Full atomic update is not ....

Lampson, B. W. (1981). Atomic transactions. Lecture Notes in Computer Science. Distributed Systems { Architecture and Implementation, 105:246-265.

....to propose some programming guidelines. These are discussed in the following. 7.1.3. 1 General principles for non standard servers If each service of a server is programmed as a restartable operation O, servicing a request despite a processor failure can be obtained in the following way [Lampson 81a] 1) Save the server s context in stable storage, 2) Perform O,and (3) Erase the server s context from stable storage. If a processor failure occurs while performing O, the process will resume after (1) and will perform O again, the resulting execution sequence is equivalent to a single ....

....processes need to obey a particular protocol. In order to facilitate the provision of fault tolerance measures within a non standard server, the server might require that the client commits its request before further action by the server. In other words, the client s request is an intention [Lampson 81a] that has to be performed by the server. The underlying reason is that in general it will be easier for a non standard server to restart the processing of a request than to be possibly obliged to cancel the processing of a request (an orphan execution) retracted by a rolled back client. In the ....

LAMPSON, B. Atomic Transactions. In Distributed Systems and Architecture and Implementation : an Advanced Course, Vol.105 of Lecture Notes in Computer Science, pp.246--265, Springer Verlag, 1981.

....10 . Fault tolerance is (also) a recursive concept; the mechanisms designed to tolerate faults must be protected against the faults likely to affect them. Examples are given by the replication of voters, self checking controllers [Carter Schneider 1968] through the notion of stable memory [Lampson 1981] in recovery data and programs. Fault tolerance is not limited to accidental faults. Protecting against intrusions has long relied on cryptography [Denning 1982] which can be viewed as a form of tolerance in that ciphered information can be inspected by an intruder without compromising its ....

B. W. Lampson, "Atomic Transactions", in Distributed Systems --- Architecture and Implementation, (B. W. Lampson, Ed.), Lecture Notes in Computer Science, 105, pp.246-65, Springer-Verlag, Berlin, Germany, 1981. Malicious- and Accidental- Fault Tolerance for Internet Applications 109

....loosely synchronized clocks at each process and a bounded delay for messages in transit. To reclaim cycles this variant may require a process to traverse its local pointer graph multiple times. This is not a good solution because of its cost in terms 2 A stable store, as defined by Lampson [75], is a set of objects that move atomically from one consistent state to another. CHAPTER 2. OVERVIEW OF GC ALGORITHMS 47 of performance. Therefore, to reclaim distributed cycles a solution similar to the previous one (tracing with timestamps) must be used. Finally, the central service can become ....

B. Lampson. Atomic transactions. In Distributed Systems---Architecture and Implementation, pages 246--265. Spinger-Verlag, New York, 1981. Lecture Notes in Computer Science vol. 105.

....[PA95] 1.2.1.2 Transaction logs Transactions are a collection of events (typically disk reads and writes) that are viewed as an atomic operation by the system. That is, if one read or write occurs, then the system guarantees that all of them occur. This is accomplished with a commit protocol [EGLT76, Gra80, Lam81]. Transaction logs are event logs (or traces) that record the completion of each transaction. Transaction logs are used for various purposes. They were originally developed to reconstruct index structures after a crash. For instance, if a crash occurs in the middle of a linked list update, the ....

Lampson, B.W. Atomic Transactions. In Distributed Systems - Architecture and Implementation: An Advanced Course. G. Goos and J. Hartmanis (Eds.), Springer-Verlag, New York, 1981, Ch. 11, pp. 246-265.

....and volatile. Information recorded on stable storage survives (with high probability) the crash of a computer and any guardians running on it. However, stable storage is more expensive than volatile storage, since duplicate copies of information are necessary to ensure that it survives crashes [Lam81] It is therefore important to decide what information will be kept in stable storage, and what will be kept in volatile storage. Since the coordinator s purpose is to do bookkeeping, most of its information must be kept in stable storage. This includes the following essential information. 1. ....

B. W. Lampson. Atomic Transactions. In Distributed Systems--Architecture and Implementation, volume 105 of Lecture Notes in Computer Science, pages 246-- 265. Springer-Verlag, New York, N.Y., 1981. This is a revised version of Lampson and Sturgis's unpublished Crash Recovery in a Distributed Data Storage System.

....Failed modules can also be easily replaced by other modules implementing the same abstraction. Some commonly used abstractions for fault tolerance have been formalized and are often provided to programmers as standardized modules. An example of such an abstrac 26 tion is that of stable storage [Lam81] Stable storage is an idealized storage device that suffers no failures and is unaffected by failures of other components of the system. Reads and writes to a stable storage complete successfully or do not happen at all, and values written to a stable storage survive all failures. Examples of ....

....must recover to complete successfully. These restartable actions, in turn, depend on two other abstractions: idempotent actions and replicated state machines. Idempotent actions are used to implement schemes like intentions lists to install a set of changes to data on a stable storage device [Lam81] The stable storage itself may be implemented using the replicated state machine paradigm, as was shown in Chapter 4. Finally, the mechanisms and runtime modules in FT SR are at the base of the hierarchy. The horizontal line in Figure 6.1 divides functionality implemented within the FT SR ....

Butler W. Lampson. Atomic transactions. In B.W. Lampson, M. Paul, and H.J. Seigert, editors, Distributed Systems---Architecture and Implementation, chapter 11, pages 246--265. Springer-Verlag, 1981. Originally vol. 105 of Lecture Notes in Computer Science.

....distributed or because they are located at different places to improve availability, data management systems are usually distributed. A fundamental issue in these systems is to ensure that data always remain consistent. This is the aim of the transaction concept introduced in the 1970s ([13, 20]) A transaction is a data manipulation unit (1) that, when executed alone, keeps data consistent and (2) to which two atomicity properties are attached. The first one, concurrency atomicity ensures that a parallel execution of a set of transactions cannot make data inconsistent: a set of ....

B. Lampson. Atomic Transactions. In Distributed Systems Architecture and Implementation : An Advanced Course, Springer-Verlag LNCS 105 ( B. Lampson, M. Paul and H. Siegert Eds), 1981, pp. 246-265.

....structures, access and concurrency control, reliability, heterogeneity, efficiency techniques, and real time distributed databases. 3. 1 Transaction Structures A transaction is an abstraction that allows programmers to group a sequence of actions on the database into a logical execution unit [17, 58, 65, 107]. Transactions either commit or abort. If the transaction successfully completes all its work, the transaction commits. A transaction which aborts does so without completing its operations and any effect of executed actions must be undone. Transactions have four properties, known as the ACID ....

B. Lampson, "Atomic Transactions," Lecture Notes in Computer Science, Vol. 105, Springer-Verlag, 1980, pp. 365-370.

....how the primary backup replication scheme can be optimized through the stable cache paradigm. 3. 2 Conciliating Fault tolerance and Performance with a Stable Cache Our basic idea is to use the underlying replication of main memory in the primary and backup servers to build a fast stable memory [18, 19] and to manage it as a cache of files. The stable memory presents properties of resiliency and consistency of data against a single machine failure. These properties are implemented by atomically modifying data structures in stable memory with a two phase commit protocol. Old Old Old New New ....

B. Lampson. Atomic transactions. In Distributed Systems and Architecture and Implementation : an Advanced Course, volume 105 of Lecture Notes in Computer Science, pages 246--265. Springer Verlag, 1981.

....to an exception handler) later of the final state (commit or abort) We u se a Partially Committed Transaction (PCT) table, which essentially is a checkpointing log, to record the events of partially committed transactions in a local site. The PCT table of each site is kept in stable storage [8]. Therefore, information stored in the PCT table will not be affected by system failures. The PCT table contains information about partially committed transactions, the recovery information and the address of an asynchronous procedure used to notify the client of a transaction about the final ....

B. W. Lampson. Atomic transactions. In Lecture Notes in Computer Science, volume 105, pages 246-- 265. Springer-Verlag, Berlin, Germany, 1981.

....in shared memory communication as in MEMSY. The checkpoint data is transferred to a stable storage. As part of the fault tolerant design effort for MEMSY the checkpointing scheme as well as a stable memory module has been designed [21] The idea of a stable storage was originally proposed in [28] with a disk based solution. Paper [13] describes a memory mapped stable storage and [5] a RAM based stable storage. Our stable storage is also RAM based and accessible by a group of processing nodes (sharing principle) Like the stable memory of the Sequoia [8] multiprocessor it uses dual memory ....

B.W. Lampson, Atomic Transactions, in: M. Paul and H.J. Siegert (eds.), Distributed Systems - Architecture and Implementation, Springer LNCS, Vol. 105 (1988) 246-265.

....stream components that concern changes in internal component states, as exempli ed by a split in which half the outputs from the previous component should be sent to one vs. the other split task. The resulting need for multi component adaptations is met by ACDS adaptation transactions [GR93, Lam81] ACDS adaptation transactions constitute a distributed version of the multiprocessor mechanism rst described in [BS91] However, ACDS adaptation transactions do not attempt to guarantee their completion times. Furthermore, our current implementation assumes a no fault error model due to our ....

B. W. Lampson. Atomic transactions. In B. W. Lampson, M. Paul, and H. J. Siegert, editors, Distributed Systems | Architecture and Implementation. Springer Verlag, 1981.

....simple protocol described above to reduce the probability of agents to be blocked due to failures. 3. Models and Protocol Overview In our model, agents are performed by nodes, that are interconnected by means of a communication network. Each node is assumed to have volatile and stable storage [Lam81], where the latter is never lost, independent of failures. Moreover, nodes are assumed to suffer from crash failures [Jal94] only. Network partitioning may occur due to communication failures. Both, the network and nodes are assumed to eventually recover from failures The network provides for ....

Lampson, B.: "Atomic Transactions". In: Lampson, B. et al (eds): "Distributed Systems - Architecture and Implementation", Springer-Verlag, 1981 Page 15

....which essentially is a checkpointing log, to record the events of partially committed RPCs. Then during a recovery,we let replicas exchange their NTD table entries in order to reach a consensus for actions on the transactions with PC returns. The NTD table of each replica is kept in stable storage [22]. Therefore, information stored in the NTD table will not be affected by system failures. The NTD table structure is listed in Listing 3. Listing 3: The Need To Do Table typedef struct ntd char rpc# name of the partially committed RPC char data# data object name used in this RPC ....

B. W. Lampson. Atomic transactions. In Lecture Notes in Computer Science, volume 105, pages 246-- 265. Springer-Verlag, Berlin, Germany, 1981.

....for the RHODOS system we must take into consideration the recoverability of the data. To achieve this mirror images of the file location map, file indices, and all the important information of RHODOS file facility are kept in a separate disk drive (provided it exists) known as stable storage [Lampson 1981], Sturgis et al. 1980] Note in the absence of a stable storage, the mirror images will be recorded duplicately on the same disk drive. 4.4 File Attributes Each of the types of the RHODOS objects is characterized by different properties (attributes) and different number of attributes. In ....

....the server. 6 Transaction Service The main objective of the transaction service in the RHODOS system is to achieve the following goals: ffl Atomicity: The transaction service should allow grouping of file operations into a transaction and allow them to execute atomically. An atomic transaction [Lampson 1981] is an all or nothing computation either it installs a complete collection of changes or none, even if interrupted by a failure. ffl Recoverability: If the transaction aborts then the transaction service should present a scenario as if the transaction had not started at all. Note that a ....

[Article contains additional citation context not shown here]

Lampson, B.W. Atomic Transactions, in Distributed Systems -- Architecture and Implementation. An Advanced Course Lampson, B.W. Paul, M. and Siegart , H.J. (editors), Lecture Notes in Computer Science, 105, pp. 246-264

....current version. Using this version the comatose objects can recover and can become available again. Note that a write quorum must contain at least one stable container, that is, a container in which the copies survive crashes, and the contents and version number remain consistent with each other [6]. Otherwise it is possible that data be lost if all containers crash. WV, in contrast, requires that all containers are stable, since there is no recovery involved after a crash. 2.3. Boot Service Container crashes are detected by the boot service . There is a boot service on each segment that ....

....of nonghosts, the intersection consists of current, non ghost copies. It is a copy from this set that the read operation uses. Serial consistency is guaranteed if reads and writes are atomic. This can be achieved by using a two phase commit protocol [7] Two phase commit requires stable storage [6]. However, in a write operation, some (not all) participants can be ghosts, and they do not have storage at all. Fortunately, ghosts can still cooperate in an atomic action since they only pretend to execute the operation. They do so by always sending positive responses on messages from the ....

B. W. Lampson, "Atomic Transactions," in Distributed Systems --- Architecture and Implementation, pp. 246-265, Springer Verlag, West Berlin, 1981.

....Error Recovery (BER) is a fault tolerant strategy which attempts to restore a correct system state after a failure has been detected. To achieve this goal, a consistent system state, called a recovery point (made up of a set of recovery data) has to be periodically saved on a stable storage [14]. BER seems to be the most attractive solution to provide fault tolerance in SSMM since it limits the hardware development and allows the use of all the processors for a computation. In this paper, we advocate that a COMA is a good candidate to build a faulttolerant SSMM. We show that COMA ....

Lampson, B. Atomic transactions. In Distributed Systems and Architecture and Implementation : an Advanced Course, vol. 105 of Lecture Notes in Computer Science. Springer Verlag, 1981, pp. 246--265.

....one requires that the components respect some protocol for the generation of recovery points. This protocol for the generation should guarantee that each generated recovery point can be included in at least one recovery line, and could be deduced from the specific internal activity of the system [6]: however, such a deduction may be difficult or extremely critical. For this, we shall discuss a generalized rule. A very simplified one is based on the consideration that, if the state of the system is consistent at a certain time, then the set of the states of the components at that time is a ....

B.W. Lampson. Atomic transactions. In B.W. Lampson, M. Paul, and H.J. Siegert, editors, Distributed Systems: Architecture and Implementation, volume 105 of Lecture Notes in Computer Science, pages 246--264. SpringerVerlag, year?

....manner, whether the data is short or long lived. Persistent systems are usually implemented above a stable store which supports the long term storage of persistent data. Such a store is stable in the sense that, following a crash or system failure, it can always return to a consistent state [lam81]. This may be implemented either by a checkpointing mechanism [lor77] or by a low level transaction facility [HR83] Several stable stores have been designed and constructed and are being used to support experimental persistent languages [bro89, mos89, ros91] In this paper we consider the issue ....

Lampson, B.W. "Atomic Transactions", Lecture Notes in Computer Science, 105, pp. 246--265 (Springer--Verlag, New York, 1981).

....of programming paradigms, which simplify the development of certain types of fault tolerant distributed programs by providing canonical organization techniques and abstractions. Examples of popular programming paradigms include the object action model [Gra86] the restartable action paradigm [Lam81] and the replicated state machine approach [Sch90] In this paper, we focus on a third area related to simplifying the construction of fault tolerant distributed programs, that of providing adequate programming language support. Specifically, we describe the design and implementation of FT SR, a ....

....by composition. When invoked, an operation exported by an FS atomic object normally executes as an atomic action that is both unitary all or nothing despite failures and serializable executed relative to other atomic actions such that the result is equivalent to some serial schedule [Lam81] However, as is always the case with fault tolerance, these properties can only be approximated by an implementation; that is, they can be only guaranteed relative to some set of assumptions concerning the number and type of failures. For example, algorithms to realize the unitary property often ....

[Article contains additional citation context not shown here]

Butler W. Lampson. Atomic transactions. In B.W. Lampson, M. Paul, and H.J. Seigert, editors, Distributed Systems---Architecture and Implementation, chapter 11, pages 246--265. Springer-Verlag, 1981. Originally vol. 105 of Lecture Notes in Computer Science.

....programming fault tolerant parallel applications. To do this, FT Linda includes two major enhancements: the ability to have stable tuple spaces and support for atomic execution of TS operations. The former is a type of stable storage in which the contents are guaranteed to persist across failures [26]; the latter allows collections of tuple operations to be executed in an all or nothing fashion despite failures and concurrency. The specific design of these enhancements has been based on examining common program structuring paradigms, both those used for Linda and those found in fault tolerant ....

....concurrent access to TS while an update is in progress. A number of schemes would satisfy these requirements. For example, techniques based on the two phase commit 2 This ordering can be relaxed in some cases; see [37] protocol for implementing general database transactions could be used [18, 26]. While sufficient, these techniques are expensive, requiring multiple rounds of message passing between the processors hosting replicas. At least part of the reason for the heavyweight nature of the technique is that it supports atomic execution of essentially arbitrary computations. While ....

B. Lampson. Atomic transactions. In Distributed Systems---Architectureand Implementation, pages246--265. Springer-Verlag, Berlin, 1981.

....corrupted. Furthermore, machines are assumed to suffer fail silent semantics [PSB 88] i.e. they fail by crashing without making any incorrect state transitions. Finally, Consul assumes that stable storage is available to each machine, and that data written to stable storage survives crashes [Lam81] In keeping with our emphasis on modularity, the mapping from fault tolerant services to protocols is primarily 1 to 1 or 1 to few; that is, the services are implemented independently of one another as individual protocols or a small set of protocols, rather than together in one monolithic ....

B. Lampson. Atomic transactions. In Distributed Systems---Architecture and Implementation, pages 246--265. Springer-Verlag, Berlin, 1981.

....service provides functionality similar to standard hardware or operating systems, but with improved semantics in the presence of failures. This category includes stable storage, atomic actions, and resilient processes. The contents of a stable storage are guaranteed to survive failures [Lam81] Atomic actions allow a number of computations to be seen as an indivisible unit by other processes despite concurrency and failure [Lam81] A resilient process can be restarted and then continue to correctly execute even if the processor on which it is executing fails; techniques to construct ....

....failures. This category includes stable storage, atomic actions, and resilient processes. The contents of a stable storage are guaranteed to survive failures [Lam81] Atomic actions allow a number of computations to be seen as an indivisible unit by other processes despite concurrency and failure [Lam81] A resilient process can be restarted and then continue to correctly execute even if the processor on which it is executing fails; techniques to construct resilient processes for example, checkpointing are discussed in [MS92] The other category of fault tolerant service provides ....

[Article contains additional citation context not shown here]

Butler Lampson. Atomic transactions. In Distributed Systems--- Architecture and Implementation, pages 246--265. Springer-Verlag, Berlin, 1981.

....is detected by these built in mechanisms, the operating system halts the machine. 2.2 RAM Stable Storage: Architectural Support for Efficient Checkpointing The implementation of checkpointing requires that checkpoints be stored in a memory which is unaffected by crashes. Stable storage disks [Lampson 81] are usually used for this purpose. Such a solution offers acceptable performance for file meta data but is inefficient when using small structures such as those of the low level system (e.g. lists, graphs) To permit efficient use of persistent memory within the operating system, we retained ....

B. Lampson. Atomic transactions. In Distributed Systems and Architecture and Implementation : an Advanced Course, volume 105 of Lecture Notes in Computer Science, pages 246--265. Springer Verlag, 1981.

No context found.

Lampson, B.W., "Atomic Transactions", Distributed Systems -- Architecture and Implementation: An Advanced Course. Chapter 11.Lecture Notes in Computer Science #105, Springer-Verlag, 1981.

No context found.

Lampson, B.W., Sturgis, H., "Atomic Transactions", Distributed Systems Architecture and Implementation; An Advanced Course, Lecture Notes in Computer Science, V. 105, Springer Verlag, 1981.

No context found.

Butler Lampson. Atomic transactions. In B. Lampson et. al., editor, Distributed Systems -- Architecture and Implementation: An Advanced Course, volume 105 of Lecture Notes on Computer Science, pages 246--265. Springer-Verlag, 1981.

No context found.

Lampson, B. W. (1981) Atomic transactions. In Lecture Notes in Computer Science, vol. 105, pp. 246--265. Springer-Verlag, Berlin.

No context found.

B. W. Lampson. Atomic transactions. In Distributed Systems: Architecture and Implementation, Lecture Notes in Computer Science, pages 254-259. Springer-Verlag, Berlin, 1981.

No context found.

B. Lampson, "Atomic transactions," in Distributed Systems--Architecture and Implementation, ser. Lecture Notes in Computer Science, B. Lampson, M. Paul, and H. Siegert, Eds. Springer-Verlag, 1981, vol. 205, pp. 246--285.

No context found.

Lampson, B.W. Atomic Transactions, in Distributed Systems: Architecture and Implementation - AnAdvanced Course, B.W. Lampson (Ed.), LNCS, Vol. 105, Springer-Verlag, pp. 246-265, 1981.

No context found.

B. W. Lampson and H. E. Sturgis. Atomic transactions. In B. W. Lampson, M. Paul, and H. J. Siegert, editors, Distributed Systems---Architecture and Implementation: an Advanced Course, volume 105 of Lecture Notes in Computer Science, pages 246--65. Springer-Verlag, New York, 1981.

No context found.

B. Lampson, `Atomic transactions', in Distributed Systems---Architecture and Implementation, SpringerVerlag, Berlin, 1981, pp. 246--265.

No context found.

Lampson, B.W., Atomic Transactions, in the same book as [Hol 81], pp.246-265. Ch. II.6.

No context found.

B. W. Lampson. Atomic Transactions. In B. W. Lampson, editor, Distributed Systems: Architecture and Implemementation, pages 246--285. Spinger-Verlag, LNCS 105, 1981.

No context found.

B.W. Lampson. Atomic Transactions. In B.W. Lampson, H.Siegert, and M. Paul, editors, Distributed Systems --- Architecture and Implementation, volume 105, pages 246--265. Springer-Verlag, 1981. Lecture Notes in Computer Science.

No context found.

Lampson, B.W. "Atomic Transactions," in Distributed Systems - Architecture and Implementation , Berlin: Springer-Verlag, pp. 246-265, 1981

No context found.

Lampson, B.W. "Atomic Transactions" in "Distributed Systems Architecture and Implementation", Springer-Verlag 81, pp. 246-264

First 50 documents  Next 50

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