Xuezheng Liu

Gridmarket: A Practical, Efficient Market Balancing Resource for Grid and P2P Computing

The emergency of computational Grid and Peer-to-Peer (P2P) computing system is promising to us. It challenges us to build a system to maximize collective utilities through presumed participants’ rational behavior. Although economic theories sound reasonable, many existent or proposed solutions based on that face problem of feasibility in practice. This paper proposes Gridmarket: an infrastructure relying on resource standardization, continuous double auction, and straightforward pricing algorithms which are based on price elasticity inherent in consumers and suppliers. Gridmarket efficiently equates resource’s demand with supply through continuous double auction and price tracing mechanism in the required price ranges. Software agent employing Gridmarket’s schedule is easy to write. To demonstrate its efficacy and efficiency, we have designed, built a simulation prototype and found the experiments promising.

DSI: distributed service integration for service grid

This paper presents DSI, a distributed service discovery and integration utility for service grid. The goal of DSI is efficiently to improve service publishing, deletion and lookup, decentralized control, scalability, and availability. DSI comprises a logically global metadata pool and a double-layered DHT (Distributed Hash Table). Efficient service publishing and lookup are achieved by adopting DHT. Clean service deletion is guaranteed by time-stamping. And system availability is improved by replication combined with automatic indexer selection. Experiments show that DSI can achieve short response time, low processing cost and high availability.

Efficiently Rationing Resources for Grid and P2P Computing

As Grid and P2P computing become more and more popular, many schedule algorithms based on economics rather than traditional pure computing theory have been proposed. Such algorithms mainly concern balancing resource supply and resource demand via economic measures. As we know, fairness and efficiency are two conflicting goals. In this paper, we argue that overbooking resources can greatly improve usage rates of resources and simultaneously reduce responsive time of tasks by shortening schedule time especially under extreme overload, while maintaining schedule principles. This is accomplished by scheduling more eligible tasks above resource capacity in advance, therefore overbooking the resource. We verify our claim on Grid Market[1].

Latent Semantic Indexing in Peer-to-Peer Networks

Searching in decentralized peer-to-peer networks is a challenging problem. In common applications such as Gnutella, searching is performed by randomly forwarding queries to all peers, which is very inefficient. Recent researches utilize metadata or correlations of data and peers to steer search process, in order to make searching more purposeful and efficient. These efforts can be regarded as primitively taking advantage of Latent Semantics inhering in association of peers and data. In this paper, we introduce latent semantics analysis to peer-to-peer networks and demonstrate how it can improve searching efficiency. We characterize peers and data with latent semantic indexing (LSI) defined as K-dimensional vectors, which indicates the similarities and latent correlations in peers and data. We propose an efficient decentralized algorithm derived from maximizing-likelihood to automatically learn LSI from existing associations of peers and data (i.e. from (peer, data) pairs). In our simulations, searching efficiency can be greatly improved based on LSI, even with the simplest greedy search preference. Our approach is a framework to exploit inherent associations and semantics in peer-to-peer networks, which can be combined fundamentally with existing searching strategies and be utilized in most peer-to-peer applications.

Stationary and adaptive replication approach to data availability in structured peer-to-peer overlay networks

Structured peer-to-peer overlay networks offer a novel infrastructure for large-scale applications. However, replications aimed for high data availability often have too much maintenance cost. In this paper we propose a simple and efficient replication strategy over low available peers, which has stationary replica locations and variation-tolerant data recovery mechanism, so that it can greatly reduce the costs for maintaining replicas, esp. in the circumstance of poor peer availabilities. We also present an analytic model for data availability and replication maintenance costs. Based on our model, we make adaptive mechanism for replicas checking, which can further reduce the cost and guarantee data availability in variable circumstance and conditions for living replicas. The simulation results proved the contribution of our approach and model.

B3A: Prompting Cooperation Among Rational Participants in P2P Bulk File Sharing Environments

Abstract Peer-to-peer (P2P) file sharing systems assume that their users download files in proportion to their sharing. Unfortunately, users are unlikely to do so without an incentive mechanism. Previous solutions to the problem required some types of centralized control, which strays from the spirit of P2P to provide long-term incentives. This paper presents the balance-based bandwidth allocation, B 3 A , an incentive framework to inspire participants to share files. In the B 3 A framework, each peer keeps the differences between the amount of data it has received from each peer and the amount it has sent to that one. When receiving simultaneous requests from various peers, the peer prefers to allocate its limited upload bandwidth to those peers with larger differences. Therefore, the downloading bandwidth perceived by a peer is positively related to its contributions to others, consequently stimulating it to share more files. The B 3 A framework is fully decentralized, scalable, and secure.

Lookup-Ring: Building Efficient Lookups for High Dynamic Peer-to-Peer Overlays

This paper is motivated by the problem of poor searching efficiency in decentralized peer-to-peer file-sharing systems. We solve the searching problem by considering and modeling the basic trade-off between forwarding queries among peers and maintaining lookup tables in peers, so that we can utilize optimized lookup table scale to minimize bandwidth consumption, and to greatly improve the searching performance under arbitrary system parameters and resource constraints (mainly the available bandwidth). Based on the model, we design a decentralized peer-to-peer searching strategy, namely the Lookup-ring, which provides very efficient keyword searching in high dynamic peer-to-peer environments. The simulation results show that Lookup-ring can easily support a large-scale system with more than 106 participating peers at a very small cost in each peer.

Paramecium: Assembling Raw Nodes into Composite Cells

In conventional DHTs, each node is assigned an exclusive slice of identifier space. Simple it is, such arrangement may be rough. In this paper we propose a generic component structure: several independent nodes constitute a cell; a slice of identifier space is under nodes’ condominium; part of nodes in the same cell cooperatively and transparently shield the internal dynamism and structure of the cell from outsiders; this type of structure can be recursively repeated. Cells act like raw nodes in conventional DHTs and cell components can be used as bricks to construct any DHT-like systems. This approach provides encapsulation, scalable hierarchy, and enhanced security with bare incurred complexity.

Efficient Search Using Adaptive Metadata Spreading in Peer-to-Peer Networks *

Search is until now a difficult problem in peer-to-peer (P2P) file-sharing systems. In this paper, we propose to use adaptive metadata spreading to make search in P2P networks efficient and scalable to large-scale systems. We model the search process in unstructured P2P networks obtain the optimized metadata populations for performance optimization. Based on the model, we propose adaptive metadata spreading approach which can adapt metadata populations to variational environment and achieve the optimized search performance. To implement our approach in fully decentralized P2P system, we employ self-organized fault-tolerant overlay trees, through which peers can easily cooperate with each other to perform adaptive metadata spreading with minor overhead.

Scalable Resource Management and Load Assignment for Grid and Peer-to-Peer Services

In this paper we propose DRMI, a fully decentralized, highly scalable and efficient resource management infrastructure, which is built on top of prefix-based DHT. DRMI provides large-scale resource monitoring and management for arbitrary metadata in variational Grid and peer-to-peer environment. DRMI monitors resources with their natural organizations and can trace them in a wide scope from large aggregations to a single resource. Based on DRMI we also propose “gradual-decision” principle and algorithm, which is consistent with requirement of self-organizing or autonomy in distributed system management. Using gradual-decision we can accomplish variety of and programmable resource management, e.g. the balanced load assignment.