Peer-to-Peer Data Sharing Among LabVIEW NodesSadia MalikCS 395T- Grid ComputingProject Report12/13/03INTRODUCTIONGlobal NamespaceSharing of Sensor Data Among PeersDynamic Network Topology and Fault ToleranceHardwareP2P Data Grid ImplementationPeer-to-Peer Data Sharing Among LabVIEW Nodes Sadia Malik CS 395T- Grid Computing Project Report 12/13/03INTRODUCTION The goal of this project is to impose a peer-to-peer (P2P) data grid on a set of LabVIEW (LV) nodes. LV nodes, in this context, are data acquisition sensors that access live data. This data is published over the network via a publish-subscribe protocol used by these sensors, and any client who knows the physical address of these sensors can directly get the sensor data from the node that owns the sensor. These nodes are also capable of running embedded LabVIEW Real-Time (LVRT) applications, which allows us to create a P2P network with these nodes by creating a global namespace for these sensors [1] such that a client can use a logical name of the sensor and query the P2P grid to locate the sensor. This report is organized as follows: First, I give a brief background of P2P computing model, sensor networks, and relevant applications [1]. Second, I describe the experimental setup and hardware I used to create the P2P netwok. The third section gives a detailed description of the P2P network design and the LV based implementation. The final section gives expected results and conclusion. BACKGROUND This project is based on two broad research areas in grid computing: P2P networks and sensor networks [1]. Both of these technologies have been around for some time [2], but the broad use of Internet technology has created several new opportunities for these technologies. Because these technologies are still being defined, I give a brief overview of each as it applies to this project. P2P Computing Model The P2P computing model allows different nodes on the network to communicate directly with each other to share data and/or resources. For this project, P2P is defined as follows [3]: • Nodes have awareness of other nodes • Peers create virtual network that abstracts the complexity of interconnecting peers • Each node can act as both a client and a server • Peers form working communities of data and application • Overall performance of the system increases as more nodes are added Sensor Network Sensor network covers the distribution of small and relatively cheap sensors that are capable of wireless communication and significant computation [4]. Sensor network technology allows us to place sensors closer to the phenomena being monitored so that the data can be processed and filtered closer to the data source. It also allows multiple sensors to collect the same information and then collaborate with neighboring nodes to resolve ambiguities. Typical sensors in this network have embedded processor, wireless communication circuitry, data storage capacity, etc.The sensor network technology is revolutionary for many situations where information gathering and processing is needed for inhospitable physical environments and/or less accessible environments, such as remote geographic regions or large industrial plants [4]. Application This project will be a prototype of a real life application that can benefit from a P2P system which allows sharing of data in a standard data grid format. The real life application can be a subset of emergency control response system to a natural disaster, for example, earthquake or a tornado. In a situation like that, data needs to be acquired from different, spatially distributed sensors, and analyzed for different behaviors; for example, to predict the next course for the disaster and damage caused by it. Currently most of these systems use central data storage and supercomputers to store and process this data [5]. As distributed and networking technology becomes faster and more reliable, more and more applications are moving away from central repository for live sensor data. In this project, I’ll present a solution that stores the data locally, in a distributed data grid, by combining the storage devices on all sensors and sharing that storage so that different sensors can put their data in a virtual global data space. This not only allows sensors to share unused storage space, but by abstracting the physical to logical memory mapping in the data grid, it also allows clients to access that data as if is located in a central global memory1. EXPERIMENT The goal of this project is to create a P2P data sharing network using LV nodes that are capable of acquiring sensor data and publishing it on the network [6]. Currently these nodes work in a client server mode. However, because these nodes can run LVRT applications [7], we can create a P2P system where each node in the grid is aware of every other node in the P2P system, and can query the node to access the sensor data. By adding a P2P data grid, we get the following benefits: Global Namespace The major benefit of having a P2P data grid is that is allows internal and external nodes to access data without having any knowledge of where that data is physically located. This is done by providing a physical to logical mapping of data to a global namespace that all peers in the grid can have equal access to. External nodes can then query any peer node in the grid to locate the actual data point in the grid. Sharing of Sensor Data Among Peers Since all peers in the grid are aware of the data and resources available in the grid, they can use this information to minimize network traffic and share resources on need basis. For example, if one node needs more disk space it can request other nodes to share any free disk space that they are not using. 1 There are many open issues with P2P and sensor networks, like network security, wireless communication, power consumption, etc. These issues are not addressed in this report.Dynamic Network Topology and Fault Tolerance A P2P data grid allows dynamic insertion and removal of sensors in the data grid. This way even if one sensor or node goes offline, the overall data grid can continue to function. By adding some redundancy in the system, we can also make the grid fault tolerant, such that if one sensor dies another sensor can take over its responsibilities. Hardware For this project, I used National Instruments’ FieldPoint RT controllers ([c]FP-20xx serial product line)
View Full Document
Unlocking...