ndnstagram ubiquitous consistency ubicon
play

NDNstagram - Ubiquitous Consistency (UbiCon) Joshua Joy, Saro - PowerPoint PPT Presentation

NDNstagram - Ubiquitous Consistency (UbiCon) Joshua Joy, Saro Meguerdichian CS217B - Spring 2012 Outline Introduction Demo Solution Questions Introduction Get the latest version of a piece of data IP -- only


  1. NDNstagram - Ubiquitous Consistency (UbiCon) Joshua Joy, Saro Meguerdichian CS217B - Spring 2012

  2. Outline ● Introduction ● Demo ● Solution ● Questions

  3. Introduction ● Get the latest version of a piece of data ● IP -- only choice: ask for the latest version from the source ● NDN -- data lives everywhere ○ ask for {version, actions} pairs for all versions ○ maybe the latest version comes first, maybe a previous version and set of actions comes first ● NDNstagram -- distributed photo editing

  4. Demo Time ● NDNstagram on multiple Android devices ● Take photo and apply actions on one device ● Fetch on two devices ○ Get the latest version of the data? ○ Get the original version and the list of actions, apply? ● Fetch on original device and one other ○ Original device should be much faster (apply actions)

  5. UbiCon - UBIquitous CONsistency ● Data is ubiquitous and pervasive ● Data surrounds us in the ether in various formats and versions ● 2 dimensional consistency in time and format ● Parallely ask for all data in the ether and get back all versions of transcoded data and diffs ● By transforming the heterogeneous versioned data we arrive at the current version of data ○ v2 + diff2 = v3

  6. How to Name? ● Requesting multiple objects in parallel for various screen sizes and versions ○ Aggregate at the screen size and version ○ Route toward the producer ● Unique version names ○ Use device_id/sequence_number as the version ○ Only need sequence number locally for each image ● Infer quality from the name (e.g. large size --> small size) ○ Include size ● Determine branching structure solely from the names ○ Include previous version name

  7. Naming Convention ● Examples ○ /ndn/ucla. edu/alice/ndnstagram/avengers/hulk/image/1280x720/alicegalaxyii/2/1280x720/ alicegalaxyii/1 ○ /ndn/mit. edu/bob/ndnstagram/avengers/hulk/image/960x480/bobiphone4s/3/1280x720/ali cegalaxyii/2 ○ /ndn/ucla. edu/alice/ndnstagram/avengers/hulk/diff/1280x720/alicegalaxyii/2/1280x720/alic egalaxyii/1 ○ /ndn/mit. edu/bob/ndnstagram/avengers/hulk/diff/960x480/bobiphone4s/3/1280x720/alice galaxyii/2 ● Structure ○ Topology ○ Data Set ○ Content type ○ Image Size (transcoder, original) ○ Version (DeviceId, Sequence Number)

  8. Meta File ● Publish meta file after each edit ○ List of image name and actions name pairs ● Each client syncs latest when fetching ○ Finds out what images and actions to fetch to get to the final version ● Branching is supported ○ Structure can be inferred from meta file ○ Client chooses which branch to fetch ○ Merging is application-specific (if needed)

  9. Multipath ● Robust multipath fetch ● Create separate groups and dispatch ○ first group to return wins ○ cancel all other fetches ○ relies on ccnx pipelining ● Naive example ○ group1 - v3 ○ group2 - v2,d2 ○ group3 - v1,d1,d2

  10. Conclusion -- Research Issues ● Naming ○ How to route toward specific screen sizes / versions of data while maintaining scalable naming hierarchy ● Syncing meta file ● Object request strategy ○ Multipath, multi-channel ● Intermediate object creation strategy ● Experiments: latency, energy, bandwidth

  11. References Mark Handley. Multipath TCP: Goals and Background ● David Eppstein, Michael T. Goodrich, Frank Uyeda, George Varghese. What’s the Difference? Efficient Set ● Reconciliation without Prior Context. Joshua Joy; Youngtae Noh; Uichin Lee; Jihoon An; Mario Gerla. Secure Personal Content Networking over ● Untrusted Devices. Joshua Joy, Ilya Moiseenko. Action Synchronization and Geometric Security. ● ● Satoshi Nakamoto. Bitcoin: A Peer-to-Peer Electronic Cash System. Werner Vogels. Eventually consistent. ● getlatestversion - http://www.ccnx.org/pipermail/ccnx-dev/2010-June/000234.html ● ● Ellis, C.A.; Gibbs, S.J. Concurrency Control In Groupware Systems. Prince Mahajan, Srinath Setty, Sangmin Lee, Allen Clement, Lorenzo Alvisi, Mike Dahlin, and Michael Walfish. ● Depot: Cloud storage with minimal trust. Puneet Kumar & M. Satyanarayanan. Supporting Application-Specific Resolution in an Optimistically Replicated ● File System. "Secrets to Lightning Fast Mobile Design". http://speakerdeck.com/u/mikeyk/p/secrets-to-lightning-fast-mobile-design? ● slide=82 ● ZHENKAI ZHU, CHAOYI BIAN, ALEXANDER AFANASYEV, VAN JACOBSON, LIXIA ZHANG. Chronos: Serverless Multi-User Chat Over NDN.

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend


More recommend