Deborah J. Hwang. Constructing a Highly-Available Location Service for a Distributed Environment. Master's thesis, M.I.T. Laboratory for Computer Science, November 1987. Master's thesis.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....server that 83 stores the mapping; workers would maintain a cache containing the mapping for recently used tuples and consult the server only when there is a cache miss or when the information in the cache is found to be out of date. Implementations of location servers are discussed in [12] 15][17][21] The portion of a worker s code that interacts with replicas would need to take the multiple sets into account. Operations concerning the same set would be done in order just as described in Chapter 3. Operations that make use of different sets can be done in the background in parallel, ....

Deborah J. Hwang. Constructing a Highly-Available Location Service for a Distributed Environment. Master's thesis, M.I.T. Laboratory for Computer Science, November 1987. Master's thesis.

....the interpretation of the operations. In addition, the categories of the updates are defined when the service is instantiated. We have applied the method to a number of applications, including detecting orphans in a distributed system [21, 22] locating movable objects in a distributed system [15], garbage collection of a distributed heap [17] deletion of unused versions in a hybrid concurrency control scheme [33] and deadlock detection in a distributed transaction system [9] We begin in Section 2 by describing our assumptions about the environment. Section 3 describes the method that ....

....3.3. Service Architecture A service is implemented by a number of replicas. We assume there is a fixed number of replicas residing at fixed locations and that clients and replicas know how to find replicas; a technique for reconfiguring services (i.e. adding or removing replicas) is described in [15]. We also assume that replicas eventually recover from crashes, and that the state of each replica is kept on stable storage [20] and restored after a crash; this assumption is discussed in Section 3.6.2. Each client runs a front end at its node. When the client calls a service operation, its ....

[Article contains additional citation context not shown here]

Hwang, D. Constructing a Highly-Available Location Service for a Distributed Environment. Technical Report M1T/LCSfI'R-410, M.I.T. Laboratory for Computer Science, Cambridge, MA, January, 1988.

....service continues to provide service in spite of node failures and network partitions. We have applied the method to a number of applications, including distributed garbage collection [18] deadlock detection [8] and orphan detection [23] locating movable objects in a distributed system [15], and deletion of unused versions in a hybrid concurrency control scheme [35] Another system that can benefit from causallyordered operations is the familiar electronic mail system. Normally, the delivery order of mail messages sent by different clients to different recipients, or even to the ....

....the response from a replica, and then returns the result to the client. We assume there is a fixed number of replicas residing at fixed locations and that front ends and replicas know how to find replicas; a technique for reconfiguring services (i.e. adding or removing replicas) is described in [15]. We also assume that replicas eventually recover from crashes; Section 3.2 discusses how this happens. In this section we describe how the front end and service replicas together implement the three types of operations. Section 2.1 describes the implementation of causal operations; Section 2.2 ....

Hwang, D. Constructing a Highly-Available Location Service for a Distributed Environment. Technical Report MIT/LCS/TR-410, M.I.T. Laboratory for Computer Science, Cambridge, MA, January, 1988.

....calls, so weaker orderings can be specified and the service is responsible for scheduling operation execution to respect that order. This method is suitable for applications such as garbage collection in a distributed heap [LL86b, Lad89] and locating movable objects in distributed systems [Hwa88] However, lazy replication does not take full advantage of the information available in its timestamps and sends a complete copy of the message log in a gossip message which increases the communication and processing overhead. Also, it does not scale very well as every node has to communicate ....

D. Hwang. Constructing a highly-available location service for a distributed environment. Technical Report MIT/LCS/TR-410, M.I.T. Laboratory for Computer Science, Cambridge, MA, January 1988.

....purpose services in a straightforward manner. Details of how a new guardian takes over from an old one need to be worked out. In 6.2. EXTENSIONS 71 particular, some form of a view change algorithm is required to replace the original guardian after it crashes. In addition, a location service [Hwa87] is needed to route requests to the new guardian. It would also be interesting to see if the voting scheme used for implementing the replicated stable storage services should be replaced by a primary copy replication scheme. The relative merits of the two schemes in this particular instance are ....

D. J. Hwang. Constructing a highly-available location service for a distributed environment. Technical Report MIT/LCS/TR-410, M.I.T. Laboratory for Computer Science, Cambridge, MA, November 1987. Master's thesis.

....R sites were not highly available, so the crash of a locator could make an migrated object unavailable even though the object s site was available. This problem does not exist in Thor because ORs are highly available. 3 Mechanisms for implementing such highly available services are described in [6] and [9] unless the hint is usually accurate. This can be accomplished with the tracking mechanism described in Section 4.3.1. Using the birth site as the locator seems preferable to using the name service because it is simpler and has at least as good performance. Therefore we consider only ....

Hwang, D. J.-H. Constructing a highly-available location service for a distributed environment. Tech. Rep. MIT/LCS/TR-410, MIT Laboratory for Computer Science, January 1988.

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