R. J. Fowler. Decentralized Object Finding Using Forwarding Addresses. PhD thesis, University of Washington, Seattle, Washington, December 1985. (Department of Computer Science Technical Report TR85-12-1).  Home/Search   Document Not in Database   Summary   Related Articles   Check

....servers in domains at levels higher than k are not informed about the object s migration. This approach effectively introduces a chain of forwarding pointers between servers in different low level domains and is comparable to approaches for locating mobile objects in local area distributed systems [6, 7, 11, 21]. Additional techniques are needed to reduce the length of chains. Orthogonal to introducing additional pointers is to distribute the load among servers in highlevel domains by introducing a fat tree [13] In this approach, the set of object identifiers is divided into equally sized subsets, ....

R. Fowler. Decentralized Object Finding Using Forwarding Addresses. Ph.D. thesis, University of Washington, Seattle, 1985.

....of messages should be performed transparently, while respecting the ordering property of dialogues. To inform users about the object s new physical location, a signpost object is installed at the old location. This signpost points to the new object location, and forwards incoming requests [7]. As such, the signpost takes care of the rerouting of requests. Furthermore, the dialogue ordering property must be maintained. To guarantee ordered handling of requests, a dialogue maintains information at both the source and destination side (typically in the form of sequence numbers) To ....

R.J. Fowler, Decentralized Object Finding Using Forwarding Addresses, Ph.D. Thesis, University of Washington, Dept of Computer Science (December 1985).

....the use of caching a resource s location (not content) at sites where the resource is frequently accessed [41] This early work typically dealt with a single instance of the resource. The case of a mobile resource was addressed through interesting techniques such as the use of forwarding addresses [18]. Beginning with initial services like ftp, archie, and gopher and culminating more recently with the World Wide Web, the Internet has experienced a dramatic growth in the use and provision of information services. This has resulted in heavy demands placed on servers and thus the desire to ....

R. Fowler. Decentralized Object Finding Using Forwarding Addresses. PhD thesis, University of Washington, 1985.

....the use of caching a resource s location (not content) at sites where the resource is frequently accessed [32] This early work typically dealt with a single instance of the resource. The case of a mobile resource was addressed through interesting techniques such as the use of forwarding addresses [14]. Beginning with initial services like ftp, archie, and gopher and culminating more recently with the World Wide Web, the Internet has experienced a dramatic growth in the use and provision of information services. This has resulted in heavy demands placed on servers and thus the desire to ....

R. Fowler. Decentralized Object Finding Using Forwarding Addresses. PhD thesis, University of Washington, 1985.

....by allocating space for the object, and copying the state of the object into that space. The template is used in locating the pointers in the state information, which are then replaced with the new addresses. Finally, the processes are resumed on the destination. Emerald uses a forwarding address [6] concept to locate objects transparently. The main drawback with Emerald is that it is designed for homogeneous environments and local area networks, and is not scalable to large distributed systems. Only immutable objects can be replicated in Emerald, and the system does not support sharing of ....

R.J. Fowler, Decentralized Object Finding Using Forwarding Addresses. PhD thesis, University of Washington, Seattle, Dec. 1985.

....Objects can be found either locally, mapped in the address space of other applications or in the persistent store. We wanted to solve efficiently the common cases, and use simple solutions for the rare ones. We also wanted to avoid the complexity of a complete decentralized algorithms like in [1,26], and third party dependencies of centralized solutions. Based on early applications and IK releases we found out that: ffl An object is mostly found where it was invoked before. ffl If a reference for an object was obtained as a parameter in a cross context invocation, then the location hint ....

....location hint. Location hints are updated when new ones are received in incoming remote invocations. In the current implementation, if different hints exist for the same object, we merely chose the last one received. We could improve hint hit ratio by associating counters to them, as Fowler did [26]. If the location hint does not exist, or it is wrong, the GRT asks the storage server bound to the SSid tag in the global name for the object. In the common case, the object is in the persistent store and the storage server retrieves it, keeping a descriptor of the application that received the ....

R. J. Fowler. Decentralized Object Finding Using Forwarding Addresses. PhD thesis, University of Washington, Seatle, USA, December 1985.

....ownership moves to another processor, a probable owner field is left behind; probable owner fields of other processors are updated when they receive copies of the item or when their copies are invalidated or updated. The probable owner fields form a chain that leads to the item s current owner [43]. Another approach distributes the copyset data structure itself. The copyset is organized as a tree. Each processor is a node with edges leading down to every node in its local subset of the copyset and up to its probable owner. Page requests flow up the tree to the current owner, which is at the ....

R. J. Fowler. Decentralized Object Finding Using Forwarding Addresses. Phd disertation, Department of Computer Science and Engineering, University of Washington, December 1985. Technical Report 85-12-1.

....to be forwarded to the new node. The inability to distinguish between two forwarding pointer the newest one, increases the length of the forwarding path of highly traveled objects. A obvious solution was to place a time stamp on each forwarding pointer. An interest proposal presented by Fowler [Fowler 85] and implemented in Emerald, says that forwarding pointers should include a counter, maintaining the number of times the object has moved. Then whenever references were sent across the network, forwarding pointers are also sent. This simple scheme allows easy detection of the newest address, ....

R. J. Fowler. Decentralized object finding using forwarding addresses. PhD thesis, University of Washington, Seatle, December 1985.

.... to lookup the location of the resource (e.g. 12, 13, 14] 3) the use of caching of a resource s location (not content) at sites where the resource is frequently accessed [15] The case of a mobile resource was addressed through interesting techniques such as the use of forwarding addresses [16]. More recently, the Service Location Working Group of the IETF is considering the design of the Service Location Protocol which allows a user to specify a set of service attributes which can be bound to a server s network address in a dynamic fashion [17] The Internet has experienced a dramatic ....

R. Fowler, Decentralized Object Finding Using Forwarding Addresses. PhD thesis, University of Washington, 1985.

....le cas de la migration de plusieurs processus en parall ele. Dans les syst emes d exploitation bas es sur l echange de messages tels que Demos mp, Accent et Emerald, pour supporter la propri et e de transparence, il suffit d avoir un m ecanisme de liens de poursuite ( forwarding ) de messages [Fow85] Dans le syst eme Demos mp, les processus sont d esign es par une adresse locale. L adresse d un processus poss ede deux composants (fig.11) Le premier composant est un identificateur global et unique du processus dans tout le syst eme. Il consiste en un identificateur du processeur sur lequel ....

R. J. Fowler. Decentralized object finding using forwarding addresses. Master's thesis, Univ. of Washington, Dec 1985.

....is caught by the Amber runtime system. The runtime system handles the failure by querying the NodeSet object for the new binding, updating its local cache of bindings, and retrying the call. 5. 3 Locating Remote Objects Amber s scheme for finding remote objects is based on forwarding addresses [Fowler 85] Each time an object moves off of a node it leaves behind the node number of its destination as a forwarding address. The forwarding address may be out of date if the object moves frequently. In this case the object s location can be determined by following a chain of forwarding addresses, since ....

Fowler, R. J. Decentralized Object Finding Using Forwarding Addresses. PhD dissertation, University of Washington, December 1985. Department of Computer Science technical report 85-12-1.

....Implementing and evaluating the scalability of this approach, and improving upon it, are topics for future research. 6.3.3 Object Mobility A Prospero link continues to work even if the object it references has moved. Prospero supports object mobility through the use of forwarding pointers [Fowler 85] A user attempting to access an object at its old location is given the new location of the object and the Prospero library automatically retries the request using the new information. When a link is refreshed, any forwarding pointers are followed and the link is updated to reflect the new ....

R. J. Fowler. Decentralized Object Finding Using Forwarding Addresses. Ph.D. Dissertation, University of Washington, December 1985. Department of Computer Science Technical Report 85-12-01.

....the most frequent user can always acquire it with light load synchronization delay 2. Our algorithm can make it possible for the application to request the migration of the master. During the transition period, the old master will forward requests to the new one using the technique described in [5]. Or, again, the decision to migrate can be made by the master itself based on the prior behavior of the system. Future work is required to refine the ideas outlined here. What this subsection shows is that a centralized approach opens up many possibilities for performance improvement. 5.2 ....

R. J. Fowler. Decentralized Object Finding Using Forwarding Addresses. Ph.D. dissertation. University of Washington, December 1985. UW CS Tech. Report 85-12-1.

....location (not content) at sites where the resource is frequently accessed [15] This early work dealt with situations where there is typically a single instance of the resource. The case of a mobile resource was addressed through interesting techniques such as the use of forwarding addresses [16]. More recently, the Service Location Working Group of the IETF is considering the design of the Service Location Protocol which allows a user to specify a set of service attributes which can be bound to a server s network address in a dynamic fashion [17] Beginning with initial services like ....

R. Fowler, Decentralized Object Finding Using Forwarding Addresses. PhD thesis, University of Washington, 1985.

....VOM keeps location hints about the object s last known position. These hints are passed along with the LLIs during remote invocations. In the current implementation, if different hints exist for the same object, we merely choose the last received one. Better decisions can be made based on counters [7]. Recycling Distributed Objects. Distributed objects are not recycled by the local garbage collector, and therefore the VOM must decide what to do with them, in particular when the context terminates: discard them, migrate them to referring contexts or turn them persistent. Each VOM maintains a ....

....apposite to persistent objects which are normally found either in the corresponding storage system or in other context. The location of distributed objects can be easily controlled and passed along with the references during cross context invocation, as in techniques based on forwarding pointers [7]. However, our distributed objects migrate mostly when the context holding them wants to exit, and so no forwarding pointer can be left in it. As already mentioned, currently we keep location information in the Storage System indicated in the LLI, for both distributed and persistent objects. This ....

R. J. Fowler. Decentralized Object finding using Forwarding Addresses. PhD thesis, University of Washington, Seatle, December 1985.

....that garbage collection happens reasonably often. Propagation of the new xref means that surrogates and location table entries are eventually unneeded. The garbage collector can recognize this and discard surrogates and table entries. Our use of forwarding addresses is similar to Fowler s work[3, 7]. Frequently moved objects may generate chains of surrogates. In Fowler s scheme, clients perform path compression by sending updates to members of a chain of forwarding addresses. In Thor, path compression is performed by the ORs instead. 4.3.2 Xrefs vs. Orefs as references With ....

Fowler, R. J. Decentralized object finding using forwarding addresses. Tech. Rep. 85-12-1, Department of Computer Science, University of Washington, December 1985.

....location (not content) at sites where the resource is frequently accessed [10] This early work dealt with situations where there is typically a single instance of the resource. The case of a mobile resource was addressed through interesting techniques such as the use of forwarding addresses [11]. More recently, the Service Location Working Group of the IETF is considering the design of the Service Location Protocol [3] We describe the relationship of our work to the Server Location Protocol efforts in Section 3.1. Beginning with initial services like ftp, archie, and gopher and ....

R. Fowler, Decentralized Object Finding Using Forwarding Addresses. PhD thesis, University of Washington, 1985.

....If it is, the invocation is a little more complex in order to handle concurrent sharing, a topic addressed in section 5. The location algorithm first asks the node given by the Storage System Identifier. This is the primary location of the object. Whenever the object moves, a forwarding pointer[4] is left on that node. 2 Measured in our current network environment. Dynamic Linking An implementation object is dynamically linked to a context as a consequence of an object fault the first time it is used. The corresponding file is read to virtual memory, internal references are relocated ....

R. J. Fowler. Decentralized Object Finding Using Forwarding Addresses. PhD thesis, University of Washington, Seatle, USA, December 1985.

....heap blocks which were previously deallocated and reused, but this mechanism is not described here. When the kernel handles a trap on an invocation of a remote object it retrieves a forwarding address for the object from the object s descriptor. The use of forwarding addresses is described in [Fowler 85] The forwarding address may be out of date if the object moves frequently. In this case the object s location can be determined by following a chain of forwarding addresses, since the object leaves a new forwarding address on each node through which it passes. It is costly to locate an object ....

R. J. Fowler. Decentralized Object Finding Using Forwarding Addresses. PhD dissertation, University of Washington, December 1985. Department of Computer Science technical report 85-12-1.

....evolved over time in a way that suggests the unconscious use of the preference hierarchy. 1.2 Related Work Previous studies of naming focus on the operational aspects of naming systems, describing their architecture and the implementation of the architecture s base elements. For example, Fowler [Fowl85], Lampson [Lamp86] Mann [Mann87] Oppen and Dalal [Oppe81] and Terry [Terr87] each describe techniques for managing a decentralized naming service. In addition, studies by Comer and Peterson [Come89] and Watson [Wats81] give general characterizations of the resolution process. In contrast, this ....

Fowler, R. J. Decentralized Object Finding Using Forwarding Addresses. PhD thesis, University of Washington, December 1985.

....location algorithm finds objects given their global name. Objects can be found either locally, mapped in the address space of other applications or in the persistent store. We wanted to solve efficiently the common cases, but also avoid the complexity of a complete decentralized algorithms as in (Fowler, 1985; Black Artsy, 1989) and third party dependencies of centralized solutions. To determine the common cases, we trace early applications and IK releases and found out that: ffl An object is mostly found where it was invoked before. ffl If a reference for an object was obtained as a parameter ....

....the cached location hint. Location hints are updated when new ones are received in incoming remote invocations. In the current implementation, if different hints exist for the same object, we merely chose the last one received. We could improve hint hit ratio by associating counters to them, as in (Fowler, 1985). If the location hint does not exist, or it is obsolete, the GRT asks the storage server bound to the SSid tag in the global name for the object. In the common case, the object is in the persistent store and the storage server retrieves it, keeping a handle to the application that received the ....

Fowler, R. J. 1985 (December). Decentralized Object Finding Using Forwarding Addresses. Ph.D. thesis, University of Washington, Seatle, USA.

....last known position as a location hint. These hints are passed along with the LLIs during remote invocations. In the current implementation, if different hints exist for the same object, we merely choose the last one received. We may improve hint quality by associating hints with counters as in [Fowler 85] In the current version of IK, the SS must be informed when an object moves from one context to another. We are improving the location algorithm to allow migration of distributed objects without involving the SS [Black 89] Object Naming. The Name Service associates human readable names with ....

R. J. Fowler. Decentralized Object Finding Using Forwarding Addresses. PhD thesis, University of Washington, Seatle, December 1985.

No context found.

R. J. Fowler. Decentralized Object Finding Using Forwarding Addresses. PhD thesis, University of Washington, Seattle, Washington, December 1985. (Department of Computer Science Technical Report TR85-12-1).

No context found.

R. J. Fowler, `Decentralized object finding using forwarding addresses', Ph.D. Thesis, University of Washington, December 1985. Department of Computer Science Technical Report 85-12-1.

No context found.

R. J. Fowler. Decentralized object finding using forwarding addresses. Tech. Rep. 85-12-1, Dept. of Computer Science, Univ. of Washington, Dec. 1985.

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