HARCHOL-BALTER, M., LEIGHTON, T., AND LEWIN, D. Resource Discovery in Distributed Networks. In ACM Symposium on Principles of Distributed Computing (May 1999).  Home/Search   Document Not in Database   Summary   Related Articles   Check

....algorithms [6] In [1, 6] gossiping is used to maintain database consistency, while in [17] gossiping is used as a mechanism to achieve fault tolerance. A theoretical analysis of gossiping is presented in [8] Recently, such techniques have also been used for resource discovery in networks [7]. Perhaps closest in philosophy to the negotiation based approach of SPIN is the popular Network News Transfer Protocol (NNTP) for Usenet news distribution on the Internet [2] Here, news servers form neighborhoods and disseminate new information between each other, using names and timestamps as ....

HARCHOL-BALTER, M., LEIGHTON, T., AND LEWIN, D. Resource Discovery in Distributed Networks. In ACM Symposium on Principles of Distributed Computing (May 1999).

....are canceled by forwarding the newest update to their targets and letting them be rolled back. Overriding the older updates requires the coordinator of the newer update to discover the older updates targets. This node discovery problem is similar to the distributed resource discovery problem [8], and the solution is also similar: the 2 nodes that receive the update send back to the coordinator the sets of nodes they know and let the coordinator expand its target node set transitively. We apply at most once messaging technique using synchronized clocks [12] to retire updates. After the ....

Mor Harchol-Balter, Tom Leighton, and Daniel Lewin. Resource Discovery in Distributed Networks. In 18th ACM Symp. on Princ. of Distr. Computing (PODC), pages 229--237, April 1999.

....contents) The problem becomes tricky otherwise, because the coordinator of the newer update sometimes must discover the targets of the older update, as demonstrated in Figure 6.2. This problem, which we call the node discovery problem, is similar to the distributed resource discovery problem [72]. The solution is also similar: a node that receives a newer, conflicting update sends the target nodes of the older update back to the coordinator and lets the coordinator expand its targets transitively. 81 Suppose a node receives an update U while it still stores gU. First, if U s timestamp is ....

Mor Harchol-Balter, Tom Leighton, and Daniel Lewin. Resource Discovery in Distributed Networks. In 18th ACM Symp. on Princ. of Distr. Computing (PODC), pages 229--237, April 1999. 6.4.4

....new members (GA3 and GA4) do not know of each other s existence. What is needed here is group membership maintenance after the initial broadcast of the membership changes so that all members will eventually obtain the complete membership information. We use a variant of the Name Dropper algorithm [9] for this purpose. Name Dropper is a distributed, randomized node discovery algorithm. Each member periodically selects a member randomly from its list, and sends its membership list to that member, which will merge that list with its own. The Name Dropper meets our goals nicely: it quickly ....

Mor Harchol-Balter, T. Leighton, and D. Lewin. Resource discovery in distributed networks. In 8th Annual ACM-SIGACT/SIGOPS Symposium on Principles of Distributed Computing, Atlanta, May 1999.

....Science and Applied Mathematics, The Weizmann Institute, Rehovot 76100, Israel. E mail: peleg wisdom.weizmann.ac.il. Supported in part by a grant from the Israel Ministry of Science and Art. 1 Introduction The resource discovery problem was introduced by Harchol Balter, Leighton and Lewin in [HLL99], as a part of their work on web caching. They developed a randomized algorithm for the problem in the weakly connected directed graph model, that was implemented within the Laboratory of Computer Science at MIT, as part of a project to build a large scale distributed cache, and then licensed to ....

....cache, and then licensed to Akamai Technologies. The motivation is to build an Internet wide content distribution system that speeds up the access of users to web pages of major content supplier web sites. In order for the machines of that system to cooperate they must rst locate each other [HLL99]. The system consists of a set V of n machines. Each machine v 2 V has a distinct identity label ID(v) which is a number of length logarithmic in n, representing its address. The machines are logically connected by a directed graph G(V; E) A directed arc hu; vi pointing from a node u to ....

[Article contains additional citation context not shown here]

M. Harchol-Balter, T. Leighton, and D. Lewin. Resource discovery in distributed networks. In Proc. 15th ACM Symp. on Principles of Distributed Computing, May 1999, pp. 229-237.

....data dissemination. A swampingbased algorithm that increases the neighborhood set as it discovers new nodes will be able to overcome this problem. To reduce the high message overhead of the swamping approach, it is possible to use a random node based approach such as the Name dropper algorithm [3]. Using the random node based approach instead of the flooding approach avoids the masking problem. Consider the example situation where a powerful node connects a network of less powerful nodes to the rest of the network. As part of their update messages each node will advertise their immediate ....

.... in distributed computer systems it has been examined in a variety of distributed systems including: mobile computing, wireless sensor networks [4] high throughput computing [9] naming systems [1] Several data dissemination algorithms based on the universal awareness scheme are examined in [3]. Their paper presents a new algorithm called the Name Dropper that is proved to have a better communication complexity when compared with three other algorithms based on flooding, swamping, and random pointer jumping, respectively. Our study is different from [3] because we examine the trade offs ....

[Article contains additional citation context not shown here]

M. Harchol-Balter, T. Leighton, and D. Lewin, "Resource discovery in distributed networks," ACM Symposium on Principles of Distributed Computing, May 1999, pp. 229-237.

....scalable distributed systems. The information disseminated by a gossip protocol usually consists of a set of distinct entries, each entry comprising a discrete piece of information about a system. Examples of information disseminated by gossip protocols include: addresses of participating hosts [5, 7, 11, 21]; locations of resources [20] bibliographic data [5] and broadcast messages [1, 5, 6, 9] When a pair of hosts exchange information, they must reconcile their respective data sets, so that each ends up knowing the other s information. What makes this reconciliation dicult is that the hosts do ....

Harchol-Balter, M., Leighton, T., and Lewin, D. Resource discovery in distributed networks. In 18th Annual ACM-SIGACT/SIGOPS Symposium on Principles of Distributed Computing (Atlanta, GA, May 1999).

.... could contain a variety of different kinds of information, such as entries in a bibliographic database [6] USENET postings [2] broadcast messages and information about the stability of those messages [6, 7, 10] the names of hosts participating in the system and the names of failed hosts [6, 22, 8], or information about the location of resources [21] The number of data entries that need to be exchanged in a given information exchange is generally small compared to the size of the entire database known to the participating hosts. It is therefore natural to ask whether it is possible to ....

....is assumed to be a subset of the other. We provide two interactive algorithms based on hash functions to solve this problem. In Section 4 we examine some extensions of our reconciliation algorithms, an in Section 5 we demonstrate an application of our results to the resource discovery problem [8]. Finally, we present conclusions and future research directions in Section 6. 2 Client Server Reconciliation From an information theoretic perspective, we can construct a lower bound on the communication complexity of the set reconciliation problem by considering the case in which each host ....

[Article contains additional citation context not shown here]

Harchol-Balter, M., Leighton, T., and Lewin, D. Resource discovery in distributed networks. In 18th Annual ACM-SIGACT/SIGOPS Symposium on Principles of Distributed Computing (Atlanta, GA, May 1999).

....algorithms [6] In [1, 6] gossiping is used to maintain database consistency, while in [18] gossiping is used as a mechanism to achieve fault tolerance. A theoretical analysis of gossiping is presented in [9] Recently, such techniques have also been used for resource discovery in networks [8]. Perhaps closest in philosophy to the negotiation based approach of SPIN is the popular Network News Transfer Protocol (NNTP) for Usenet news distribution on the Internet [2] Here, news servers form neighborhoods and disseminate new information between each other, using names and timestamps as ....

Harchol-Balter, M., Leighton, T., and Lewin, D. Resource Discovery in Distributed Networks. In ACM Symposium on Principles of Distributed Computing (May 1999).

....We also describe a variation of the algorithm, which runs in O(log 2 n) expected time but has O(n) expected message complexity. 1 Introduction The resource discovery problem of networked machines seeking and learning about each others is introduced by Harchol Balter, Leighton, and Lewin in [1]. In a distributed network, if machine u knows the address of machines v, machine u can connect machine v and inform it of other machines known to u. A resource discovery algorithm specifies how the machines should communicate with each other, with the goal that each machine will be aware of all ....

....other, with the goal that each machine will be aware of all other machines. There are three performance measures for a resource discovery algorithm: 1. time complexity number of time steps taken; 2. message complexity number of messages sent (called connection communication complexity in [1]) and This work was supported in part by the MIT Auto ID Center. Please forward any future correspondence to K. Y. Siu at siu list.mit.edu. 3. pointer complexity number of pointers (machine addresses) passed (called pointer communication complexity in [1] Prior algorithms for resource ....

[Article contains additional citation context not shown here]

Mor Harchol-Balter, Tom Leighton, and Daniel Lewin. Resource discovery in distributed networks. In 18th Annual ACMSIGACT /SIGOPS Symposium on Principles of Distributed Computing, May 1999.

....the Grid. This paper focuses on approaches for performing the data dissemination that is necessary to keep the status databases consistent. Recently, data dissemination for resource discovery has been an active research area with applications such as web content management motivating the research [3, 10]. Several research initiatives have examined this issue in the context of Grid computing as well [8, 12] Resource discovery is defined as the process by which a node or machine in a distributed system becomes aware of the attributes or capabilities of the other nodes that are part of the system. ....

....before forwarding them. This is necessary because the network condition can change dynamically and the sender of the message may have been unaware of the change when it sent the message. Universal algorithm: This is a well studied algorithm that has been used as a baseline in several studies [3, 4, 12]. The pseudo code in Figure 4 describes the major steps in this algorithm. In this algorithm, a node sends status messages to every node it is knows about, i.e. in its neighbor list. The also relays the messages without any filtering and when it receives a message updates the local database. The ....

[Article contains additional citation context not shown here]

M. Harchol-Balter, T. Leighton, and D. Lewin. Resource discovery in distributed networks. In 18th ACM Symposium on Principles of Distributed Computing, pages 229-- 237, May 1999.

....a component based agent framework. 2 Algorithms and Models for Distributed Awareness A first step in all applications that require some knowledge about the other nodes of a network is to learn about the existence of each other. We call this process distributed awareness , while other authors [1] refer to it as resource discovery. We believe that in a heterogeneous environment learning about the existence of other nodes is only the first step in a complex process and that resource discovery requires a set of progressively more intricate interactions with the newly discovered object. 2.1 ....

....pointer jump with back edge requires u to send back to v a pointer to all its neighbors. There are even strongly connected graphs that require with high probability Omega Gamma n) time to converge to a complete graph in the random pointer jump algorithm. The Name Dropper algorithm is proposed in [1]. During each round each machine v transmits Gamma(v) to one randomly chosen neighbor. A machine u that receives Gamma(v) merges Gamma(v) with Gamma(u) In this algorithm after O(log 2 n) rounds the graph evolves into a complete graph with probability greater than 1 Gamma 1= n O(1) ....

[Article contains additional citation context not shown here]

M. Harchol-Balter, T. Leighton, and D. Lewin. Resource Discovery in Distributed Networks. In Proceedings of PODC'99, pg. 229-237, Atlanta, 1999.

....algorithms [6] In [1, 6] gossiping is used to maintain database consistency, while in [19] gossiping is used as a mechanism to achieve fault tolerance. A theoretical analysis of gossiping is presented in [9] Recently, such techniques have also been used for resource discovery in networks [8]. Close in philosophy to the negotiation based approach of SPIN is the popular Network News Transfer Protocol (NNTP) for Usenet news distribution on the Internet [2] Here, news servers form neighborhoods and disseminate new information between each other, using names and timestamps as meta data ....

Harchol-Balter, M., Leighton, T., and Lewin, D. Resource Discovery in Distributed Networks. In ACM Symposium on Principles of Distributed Computing (May 1999).

....algorithms [6] In [1, 6] gossiping is used to maintain database consistency, while in [18] gossiping is used as a mechanism to achieve fault tolerance. A theoretical analysis of gossiping is presented in [9] Recently, such techniques have also been used for resource discovery in networks [8]. Perhaps closest in philosophy to the negotiation based approach of SPIN is the popular Network News Transfer Protocol (NNTP) for Usenet news distribution on the Internet [2] Here, news servers form neighborhoods and disseminate new information between each other, using names and timestamps as ....

Harchol-Balter, M., Leighton, T., and Lewin, D. Resource Discovery in Distributed Networks. In ACM Symposium on Principles of Distributed Computing (May 1999).

....may be performed on top of the Gossamer layers. We now look at the details of the Gossamer protocol starting with a brief overview of the entire protocol. 3. 2 Protocol Overview When an SCX joins a session, it uses a variant of a network resource discovery protocol proposed by Harchol et al. [20] to discover other mesh members. It uses the set of rendezvous points listed in the session announcement to bootstrap the discover process. As the SCX encounters new nodes, it selects some of them to be its neighbors in the mesh. As defined by the degree constraints, each SCX has a target number ....

....to construct source rooted reverse shortest path distribution trees. We now look at some of the details of the mesh construction and tree building algorithms. 3. 3 Node Discovery: Name Dropper When an SCX joins a session, it uses a variant of the Name Dropper protocol proposed by Harchol et al. [20] to discover other SCX nodes. We note that it is not required for SCXs to have complete and accurate information of mesh membership at all times. An SCX (say X i ) initiates the Name Dropper algorithm by discovering a small set of mesh members through some startup rendezvous mechanism. Let us ....

[Article contains additional citation context not shown here]

HARCHOL-BALTER, M., LEIGHTON, T., AND LEWIN, D. Resource Discovery in Distributed Networks. In Proceedings of the 18th Annual ACM-SIGACT/SIGOPS Symposium on Principles of Distributed Computing (Atlanta, GA, May 1999).

....are canceled by forwarding the newest update to their targets and letting them be rolled back. Overriding the older updates requires the coordinator of the newer update to discover the older updates targets. This node discovery problem is similar to the distributed resource discovery problem [8], and the solution is also similar: the nodes that receive the update send back to the coordinator the sets of nodes they know and let the coordinator expand its target node set transitively. We apply at most once messaging technique using synchronized clocks [12] to retire updates. After the ....

M. Harchol-Balter, T. Leighton, and D. Lewin. Resource discovery in distributed networks. In ACM Symp. on Princ. of Distr. Computing, pages 229-237, 1999.

....algorithms [6] In [1, 6] gossiping is used to maintain database consistency, while in [17] gossiping is used as a mechanism to achieve fault tolerance. A theoretical analysis of gossiping is presented in [8] Recently, such techniques have also been used for resource discovery in networks [7]. Perhaps closest in philosophy to the negotiation based approach of SPIN is the popular Network News Transfer Protocol (NNTP) for Usenet news distribution on the Internet [2] Here, news servers form neighborhoods and disseminate new information between each other, using names and timestamps as ....

HARCHOL-BALTER, M., LEIGHTON, T., AND LEWIN, D. Resource Discovery in Distributed Networks. In ACM Symposium on Principles of Distributed Computing (May 1999).

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