The claimed subject matter provides a system and/or a method that facilitates distributed backup and versioning techniques for network-based backup architectures. When data is committed to a backup system, a backup version can be created. The generated backup version can be a complete rendition of data or the generated backup version can be a differential between the data and backup information previously committed to the system. The generated backup version can be replicated across a distributed network of peers. Replication can be based upon a desire level of redundancy of versions, location of peers, network topology of peers, storage capacity of peers, health of peers, type of peer, and the like. In addition, versions can be retained based upon a configurable number of desired versions.

Distributed backup and versioning

The claimed subject matter relates to systems and/or methodologies that facilitate intelligent distribution of backup information across storage locations in network-based backup architectures. A virtual layering of backup information across storage locations in the backup architecture can be implemented. Statistical models are utilized to dynamically re-allocate backup information among storage locations and/or layers to ensure availability of data, minimum latency upon restore, and minimum bandwidth utilization upon restore. In addition, heuristics or machine learning techniques can be applied to proactively detect failures or other changes in storage locations such that backup information can be reallocated accordingly prior to a failure.

Intelligent tiers of backup data

The claimed subject matter relates to systems and/or methodologies that facilitate distributed storage of data. A distributed file system can be implemented on storage nodes such that the system places multiple copies of data (e.g., replicas) on a variety of disparate storage nodes to guarantee availability of the data and minimize loss of the data. Storage nodes are dynamically evaluated to identify respective characteristics. In one example, the characteristics can include availability of a storage node, capacity of a storage node, data storage cost associated with a storage node, data transfer costs associated with a storage node, locality of a storage node, network topology, or user preferences associated with a storage node. The characteristics can be employed to generate optimal placements decisions.

Dynamic placement of replica data

Embodiments relate to systems and methods for automatically generating a system restoration order for network recovery. A set of managed machines, such as personal computers or servers, can be managed by a network management platform communicating with the machines via a secure channel. The network management platform can access a dependency map indicating a required order for restoration of machines or nodes on a network. The network management platform likewise access a reverse kickstart file for each machine to be automatically restored in order to ensure proper functioning of the network, and extract a current configuration of that machine for purposes of restoring the overall network.

Systems and methods for automatically generating system restoration order for network recovery

A data overlay is built as a data structure on a logical space defined by a distributed hash table (DHT) in a peer-to-peer network. The data overlay includes a tree having tree nodes that each have a zone mapped to a corresponding DHT node in the logical space of the DHT. The logical space of the DHT is mapped to machines, each of which corresponds to one or more of more of the tree node zones. The tree nodes are hierarchically situated by tree node zone size and my available resource so that tasks are performed by machines in the peer-to-peer network according to the respective abilities of the machines to supply the tasks' demand. The tree, which self-organizes and self-heals on the same scale as the underlying DHT, is used together and disseminate information from and to the DHT nodes using the hierarchy of the tree nodes.

Data overlay, self-organized metadata overlay, and application level multicasting

Disclosed is a method and apparatus for providing range based queries over distributed network nodes. Each of a plurality of distributed network nodes stores at least a portion of a logical index tree. The nodes of the logical index tree are mapped to the network nodes based on a hash function. Load balancing is addressed by replicating the logical index tree nodes in the distributed physical nodes in the network. In one embodiment the logical index tree comprises a plurality of logical nodes for indexing available resources in a grid computing system. The distributed network nodes are broker nodes for assigning grid computing resources to requesting users. Each of the distributed broker nodes stores at least a portion of the logical index tree.

Method and apparatus for distributed indexing

A resource directory is a critical component of a grid architecture. P2P computing paradigm could address some of the scalability issues that make distributed resource discovery services challenging. Unfortunately, most existing distributed hash table (DHT) based P2P overlays have difficulty in treating attribute range queries that are common in resource discovery lookups. This paper proposes a general framework for range-based resource discovery. In particular, the proposed framework maps tree-structured logical data (i.e., range indexing) onto a DHT-based physical node space (i.e., resource brokers). In this paper, we consider three mapping schemes from the logical space onto the physical space. Each mapping scheme uses a different replication mechanism to reduce range search time and to achieve load balance. We analytically and experimentally compare the performance characteristics (query/update costs and workload distributions) of these schemes and discuss their applicability under different resource discovery service scenarios.

/pdf/on-overlay-schemes-to-support-point-in-range-queries-for-4pa3ic10pl.pdf

On overlay schemes to support point-in-range queries for scalable grid resource discovery

Web Services Quality Prediction has become a popular research theme in Cloud Computing and the Internet of Things. Graph Convolutional Network (GCN)-based methods are more efficient by aggregating feature information from the local graph neighborhood. Despite the fact that these prior works have demonstrated better prediction performance, they are still challenged as follows: (1) first, the user-service bipartite graph is essentially a heterogeneous graph that contains four kinds of relationships. Previous GCN-based models have only focused on using some of these relationships. Therefore, how to fully mine and use the above relationships is critical to improving the prediction accuracy. (2) After the embedding is obtained from the GCNs, the commonly used similarity calculation methods for downstream prediction need to traverse the data one by one, which is time-consuming. To address these challenges, this work proposes a novel relationship discovery and hierarchical embedding method based on GCNs (named as RDHE), which designs a dual mechanism to represent services and users, respectively, designs a new community discovery method and a fast similarity calculation process, which can fully mine and utilize the relationships in the graph. The results of the experiment on the real data set show that this method greatly improved the accuracy of the web service quality prediction.

/pdf/relationship-discovery-and-hierarchical-embedding-for-web-8775vwpq.pdf

Relationship Discovery and Hierarchical Embedding for Web Service Quality Prediction

The information service is a critical component of a grid infrastructure for resource discovery. Although P2P computing paradigm could address some of the scalability issues that plagued grid resource discovery, most existing distributed hash table (DHT) based P2P overlays have difficulty in treating attribute range queries that are common in resource discovery lookups due to the inherent randomness of hash functions. Recently, there have been various attempts to solve the range search problem over DHT networks [Aberer, et al. (2003), Gao and Steenkiste (2004), Ratnasamy, et al. (2003), and Tanin, et al. (2005)]. Central to all of these is a mapping scheme which maps the tree-structured logical index space to some DHT-based physical node space. In this paper, we propose a general framework to put all these under the same umbrella based on how mapping of the tree-structured index that identifies a physical node responsible of a particular range is done through replication. We identify three schemes which should cover the spectrum of all meaningful replication schemes: the tree replication scheme (TRS) replicates the entire tree; the path caching scheme (PCS) replicates paths from root to leaves; and the node replication scheme (NRS) replicates individual logical nodes.

/pdf/dht-overlay-schemes-for-scalable-p-range-resource-discovery-mn25elddkv.pdf

Liping Chen

Papers

Method and apparatus for distributed indexing

On overlay schemes to support point-in-range queries for scalable grid resource discovery

Relationship Discovery and Hierarchical Embedding for Web Service Quality Prediction

DHT overlay schemes for scalable p-range resource discovery