Munin: A Peer-to-Peer Middleware for Ubiquitous Analytics and Visualization Spaces
Download as PPTX, PDF0 likes845 views
The document discusses Munin, a peer-to-peer middleware framework designed for ubiquitous analytics and visualization. It highlights the architecture, features of collaborative visualizations, and interaction with heterogeneous data during emergency situations. Additionally, it addresses potential shortcomings of the peer-to-peer architecture while emphasizing its scalability and fault tolerance.
Munin: A Peer-to-Peer Middleware for Ubiquitous Analytics and Visualization Spaces
- 1. A Peer-to-Peer Middleware for Ubiquitous Analytics and Visualization Spaces Sriram Karthik Badam, Eli Fisher, Niklas Elmqvist Munin TVCG-invited at IEEE VIS 2014 sbadam@umd.edu 1
- 2. “The most profound technologies are those that disappear. They weave themselves into the fabric of everyday life until they are indistinguishable from it.”… Mark Weiser Ubiquitous Computing 2
- 3. Co-located Collaborative Visualization Large multi-screen display Tabletop computer Interactive Tree Comparison for co-located collaborative visualization (Isenberg and Carpendale, 2007) Hugin (Kim et al., 2010) Lark (Tobiasz et al., 2011) 3
- 4. Mobile Visualization Visualizing information on mobile devices (Chittaro, 2006) 4
- 5. Ubiquitous Analytics Massive heterogeneous data Ubiquitous analytics: Interacting with big data anywhere, anytime (Elmqvist and Irani, 2013) 5
- 6. Emergency Situation Heterogeneous Information sources Multiple analysts/ stakeholders Ensemble of devices 6
- 7. Interacting with data during emergency Let’s see how the spread of this fire is different from two years ago. Aha!! that’s why it happened! Siri, capture screenshot. Sam, look at this! Wait, Linda, let me visualize the sensor data here! 7
- 8. Design Supporting CollaborationMultiple and Networked Devices Post-WIMP interaction WIMP – Windows, Icons, Menus, Pointers Supporting Visualization 8
- 9. Munin • Peer-to-peer infrastructure • Uses three-layered architecture • Shared space • Service • Visualization • Java based • Middleware framework for ubiquitous analytics 9
- 10. Munin Shared state Display service Display service Input service Input service Munin Peer Munin Peer Munin Peer IP MulticastIP MulticastIP Multicast Services Visualization Layer 10
- 11. Shared state Associative memory for sharing objects Shared event space P2P JGroups framework for setting up and communicating through a P2P infrastructure //create peer Peer peer = new Peer(); //Get access to shared space SharedSpace sharedState = peer.getSpace(); state.addObjectListener(new mediaSpaceListener()); //Create a Shared Object Listener public class mediaSpaceListener implements SharedSpaceObjectListener { public void objectCreated(SharedObject so) { add(so); so.addChangeListener(new mediaChangeListener()); so.commit(); } public void objectDeleted(SharedObject so) { } } 11
- 12. Services Display service Input service Shared state Shared memory Shared Event Display service Touch events Shared objects P2P connection (local subnet) public class DisplayService implements Service, SharedSpaceEventListener { //Register service with the shared space to receive shared events public DisplayService (Platform platform, Surface surface) { platform.getSpace().addEventListener(this); } //Shared Event Handler public void eventReceived(SharedEvent event) { } } 12
- 13. Connected over Munin Visualization Layer U V 44.968046 44.33328 33.755787 33.844843 44.92057 44.240309 44.968041 44.333304 33.755783 -94.420307 -89.132008 -116.35999 -116.54911 -93.44786 -91.493619 -94.419696 -89.132027 -116.36006 Draw map Draw circles at locations (u1, v1), (u5, v5) …. Draw image Draw heat map Draw rectangles at locations (u3, v3), (u7, v7), (u8, v8) Shared state Services Visualization layer Piccolo framework for visualization on desktop platforms 13
- 14. Visualization Layer Dataset on the shared state layer Display Service Input and Display Services 14
- 15. Application Examples Collaborative Visual Search Distributed PolyZoom Distributed Media Player 15
- 16. Collaborative Visual Search Shared state Touch input service for pan and zoom Touch input service for pan and zoom Services Shared touch events Java2D display service for the view Android display service for the view SVG rendering service for shapes Android SVG rendering service for shapes OpenStreetMap renderer Android OpenStreetMap renderer Real estate dataset Visualization layerDistributed scene graph Draw map Draw markers for locations Draw map Draw markers for locations 16
- 17. Collaborative Visual Search Public view (shared) Private view Branch-Explore-Merge: McGrath et al., 2012 17
- 18. PolyZoom (Javed et al., 2012) 18
- 19. Distributed PolyZoom Shared state Touch input service to select zoom regions Services Shared touch events JavaFX display service Android display service for the view Google map renderer for JavaScript Google map renderer for Android Visualization layer Draw map Distributed scene graph Draw markers Draw tree-layout with edges Draw maps at each node 19
- 20. Distributed Media Player Services Streaming display service (VLCJ) Player state (play/pau se) Control events (FF/rewin d) Shared state Android video player Android display service Input service 20
- 21. Shortcomings Peer-to-Peer Architecture Service-oriented model Requires: • Conflict and consistency management at each peer. • Exclusive handling logic at each peer. Requires: • Redevelopment of simple services for each platform. • Platform dependency But: • More scalable, fault-tolerant, and secure. • Data-driven and service-oriented models for easier development. 21
- 22. PolyChrome Presenting at ACM ITS 2014 • Completely web-based • Support for JS visualization frameworks • Hybrid architecture 22
- 23. Conclusion Conceptual model of Munin framework Application examples How-to-use: 23
- 24. Acknowledgment S. Karthik Badam sbadam@umd.edu Questions? IIS-1253863 Munin source https://github.com/karthikbadam/Munin 24
Editor's Notes
- #2: Name – framework for building analytical spaces that go beyond a single device – done by our group includes Eli Fisher, Niklas – Now, taking a few steps back into the past --
- #3: Mark Weiser had the vision of ubiquitous computing in which technologies evolve disappear into the background to aid us in achieving higher and bigger goals – Ubicomp inspires many fields including mobile computing, distributed computing, and HCI. However, Ubicomp has not made a lot of progress in visualization (Segue)
- #4: (start animation) We do have some brilliant research on co-located collaborative visualization including
- #5: (start animation) and we have some research exploring the Mobile visualization
- #6: In order to apply Mark Weiser’s vision and combine these efforts to interact with data in visual form anywhere and anytime for analytical processes, an approach called ubiquitous analytics was proposed. To understand how ubiquitous analytics is useful, lets consider an emergency situation (Segue)
- #7: During an emergency situation – get data form multiple sources (resource information, sensor, social media) – multiple analysts and stakeholders working together on multiple devices (segue)
- #8: They can work with multiple devices, interact with a visualization through gestures with large displays, collaborate with others, and finally we have users not confined to a single office space (without restrictions). In order to develop ubilytics applications we need
- #9: Summing it up… we have 4 requirements To support development of applications for ubiquitous analytics we created the Munin framework (segue) Cross-platform, to support heterogeneous devices. Peer-to-peer, to avoid servers and increase robustness. Graphics-agnostic, since devices have varying APIs. Multiple concurrent users with concurrent interactions. Co-located and synchronous collaboration support. Multiple input sources, such as touch, gestural, pen- based, tangible, and full-body interaction. Multiple output sources, such as wall-mounted, table- top, mobile, volumetric, and tangible displays. Visualization mechanisms and patterns (e.g., [60]). Multiple visual representations and views [11].
- #10: In our research, we created Munin, a framework for ubiquitous analytics spaces. Munin is a middleware framework Munin is decentralized. It connects the devices over a p2p infrastructure, no client/servers involved. It has a three layers. Java
- #11: Looking closer at the architecture, We have multiple devices connected over a shared space, generating multiple services, which are accessed by a visualization layer.
- #12: The shared state is essentially created through multicast in the P2P infrastructure. It contains an associative memory to store shared objects (for example) and an event space to share the generated events. Lets take a quick look at the code for accessing shared space. Moving on to the Service layer--
- #13: It can contain services such as for display, input handling, and even for managing computation. These services commute with the shared events and sometimes the shared objects to update the view on each device. Other than generic services such as display, input, computation, we can also create custom services, for example, zoom and pan, with the Munin framework.
- #14: The visualization layer sits on top of the service and shared state layers to support development of visualization. It provides a shared relational data structure to store structured data, and a distributed scene graph to draw the graphical abstractions for visualization.
- #16: We made three example applications using the Munin framework.
- #17: One of our most sophisticated - Visual search of real estate information visualized on map. – shared and public view – personal and private view (segue)
- #18: Tabletop for coupled work and tablet to individual exploration, provides a branch explore merge kinda scenario for visual exploration – the second example is a distributed version of an existing technique (segue)
- #19: The first example is an multi-screen or distributed version of PolyZoom. This lets you build hierarchies of focus regions such that each subsequent level in this tree shows a higher magnification.