100 likes | 230 Vues
This weekly report by Brett Geren discusses several methods for acquiring academic papers, including Google Scholar and CiteSeer, and introduces bibliography tools like BibTeX and CiteULike. It explores data management in mobile environments, emphasizing the importance of ACID properties. The report highlights consistency solutions such as two-tier replication and the importance of Mobile Ad-Hoc Networks (MANETs) in data distribution. Various methods for dynamic social grouping and probabilistic routing are also covered, along with a reminder of the significance of maintaining organization in data management.
E N D
Weekly Report Brett Geren
Academic Notes • Several sites provide methods for the acquisition of academic papers: • Google Scholar • CiteSeer • ACM Digital Library (ACM Portal) • Online MST Library (http://library.mst.edu/) • LaTex provides a bibliography engine called BibTex. It’s pretty sweet (so says Dylan) • CiteULike can be used to auto generate bibliography stuff.
TinyOS • TinyOS.net is the “go-to-guy” • Uses NesC, C for networked, embedded systems • Non-blocking and doesn’t preempt like larger OS’s • Normally blocking operations are done asynchronously through callback events • Tasks are provided to allow preemptable, long-running background tasks
Data Compression • Data compression allows us to minimize the amount of data sent, a costly endeavor • Methods: • Pipelined In-Network – chop off shared data • LZW/LSZW – replace character sequences with indices into a dictionary • GAMPS – group like signals, possibly scaling some data • Delta Huffman Compression – express data as changes from a previous values; compress using Huffman coding
Data Management in Mobile Environments • Highly reminiscent of distributed file systems or databases • Good database systems follow the ACID properties: • Atomicity • Consistency • Isolation • Durability • How do we manage data on many mobile devices without violating any ACID properties? • Inconsistent state across devices • Isolating changes is difficult
Data Management in Mobile Environments (Cont.) • Consistency Solutions: • Two Tier Replication • Tentative local update • Lazy update primary • Pitoura’s Method • Cluster “strongly connected” devices, nominate a leader • Intra-cluster requires mutual consistency • Group leaders store system-wide values (core copies) • Inter-cluster allows inconsistency • Group followers can have conditionally committed data (quasi copies)
MANET • Mobile Ad-Hoc Networks (MANET’s) play crucial roles in many scenarios (such as disaster relief) • How do we effectively distribute needed data: • SAF – every MH replicates highly sought after data • DAFN – neighbors take care not to have duplicate data • DCG – “stable” groups are formed, groups do not have duplicate data • Consistency: • Establish quorums – every two quorums must share at least one common node • Updated data is routed through these common nodes
Dynamic Social Grouping • We are interested in dynamically grouping nodes for routing and other purposes • Bubble-rap • Form cliques with “full contact” nodes • Merge sufficiently close groups • Socialcast Model • Calculate utility value based on your “closeness” to a group. • Probabilistic Routing • Successful deliveries increase probability • Timeouts decrease probability
End Notes…. • Torvalds has an ‘s’ in it • VIM is pretty cool • Questions?