Message Area
 
 
Seminar
Ubiquitous Computing Lab at University at Buffalo\n210 Furnas Hall,\nBuffalo, NY
meeting time: Wed 1pm-3pm, Bell 224
PapersToDiscuss in the remaining of the semester\n
[[Syllabus]] PapersToDiscuss [[Reviews]] [[Announcements]]\n[[Links]]
[[Announcements]]\nPapersToDiscuss\n[[Syllabus]]
CSE 708
!! 2 paragraphs for executive summary\n* what is the paper trying to do?\n* what is potential contribution of paper?\n* summary of strengths and weaknesses\n\n!! several paragraphs of details (listed in order of importance)\n* technical flaws?\n* structure of paper?\n* are key ideas brought out?\n* motivation and justification of approach -- why are these ideas important?\n* presentation? (ex: undefined terms, unclear sections…)\n* comparison with relevant work? \n\n!! questions and issues raised in the class \n* Include answers to the most interesting questions raised in the class\n* What other issues are raised?\n* How could you improve the paper?\n* Potential follow-up work, future work?\n\n(The format is borrowed from Ousterhout's advice for paper reviewing.) \n\n!!! Finish your review within 2 days of the class, and post your review in the blog. Include your name as the note-taker for the paper.
The seminar hours are __Friday 10:00-12:00 at Davis 113a__.\nRead the [[Syllabus]].
Follow our course page at Piazza.
!!! AutoDropboxify all web/url shared documents\n!!! Use FUSE to build a simple distributed filesystem\n!!! Extend GlusterFS to add a meaningful feature\n!!! Deploy a wide area network Paxos solution\n!!! Implement a self-healing and convergent replicated data structure \n!!! Implement a 2-3 locality Giraph program compare with mapreduce implementation\n\n
I often download pdf presentations from the web, but then when I reopen them in my laptop later, I wonder if they got updated/changed in the meanwhile. It would be nice to have a service where my local copy also gets automatically updated (a la dropbox) when the authoritative source pdf gets changed.\n\nSo here is my project idea: auto-dropboxify all web/url downloaded documents, so the downloaded local copy of a file is kept up-to-date with the authoritative copy on the web. Please build this, I think many people will find this useful. Some other examples of documents downloaded from the web and would benefit from auto-updating include: city documents (e.g., street-parking rules, garbage collection rules, etc.), event calendars, tax documents, CVs. (Do you have any other examples?)\n\nThis shouldn't be too hard to build. A completely client-side solution is possible. A client-side software that keeps track of web-downloaded documents, and periodically checks with http to detect whether any of these documents got changed would do it. If a change is detected, the software should download a copy, and should prompt the user when she opens the local document next about whether the updated or the old copy to be used.\n\nOf course a cloud-hosted push-based synchronization solution would be more scalable and efficient. This would be also more gentle for the original content provider as instead of thousands of client-side software periodically checking for updates, only the cloud-hosted synchronization service will check for the update to the document. Upon detecting an update the cloud service will push the updated document to all clients that posses this document. I am sure there are a lot of details to work out to make this service efficient. And there may even be a way to monetize this service eventually, who knows? This cloud hosted synchronization service idea may be seen as extending the CDNs to reach out and embrace the pc/laptops as the last hop for content-replication.\n\nA further generalization of this "auto-dropboxify documents downloaded from the web" idea is to allow any host to modify/update the document (yet the other copies should still be kept in sync/up-to-date). This then turns into an application-level virtual-filesystem that is web-scale. And implementing that would be very tricky.
!! Paper discussion\nWe will discuss 20+ papers in 14 weeks, sparing 1 hour for each paper.(Some weeks we will use the second hour for project discussion.) Each student will serve as a presenter for one paper, as a notetaker for the same paper, and as a participant for all the remaining papers. We will follow this format:\n\n* By Thursday midnight, each participant should contribute 1 or 2 questions about the paper to the course webpage at Piazza.com. You should state a question not stated before. The question should have some substance and depth, otherwise I will delete the question and you will not get any credit from the question.\n\n* The presenter will use 30 minutes to discuss the heart (the most important and useful part) of the paper. The presenter is allowed to use up to 10 slides.\n\n* In the remaining 30 minutes, the presenter will be answering the questions collected at Piazza, and we will have further discussion about the paper.\n\n* After the class (by Sunday), the notetaker is responsible for writing a review of the paper based on her notes and the discussion in the class. The review should follow these FormattingGuidelines.\n\nAs a participant, you should expect to spend at least 3 hours reading each paper. When it is your turn to become the presenter, the preparation time should take to the north of 12 hours. Go through a couple papers in the list to gauge if you can handle the material technically.\n\n!!Projects\nEach student will participate in a project, regardless of whether she is taking the class for 1 credit or 3 credits. Project groups will consist of 3 students. I will provide the ProjectIdeas. The groups will have some say in the project selection; I will, however, have the final say to enforce an even assignment of project ideas to the groups. I will assign each project to two groups, which will compete with each other. The groups will provide presentations throughout the semester. At the end of the semester, the groups will give a project demo and a final project report. The report will include itemized contributions of each participant. I will also get assessment from each group member about the contributions of other group members.\n\n!! Grading\nEach student will be evaluated the same way, regardless of whether she is taking the class for 1 credit or 3 credits. I will assign the S/U grade based on paper discussion performance and project performance. (As department policy, we do not assign letter grades for 700-level seminar courses.)\n\n15% is paper presentation, 10% writing the paper review, and another 25% for paper discussion participation---you will lose 4% for each paper without a good question. The remaining 50% is for the project.\n\nThe seminar will be intensive and will require a lot of hard work. The S/U grades are of course non-negotiable.If you have a problem with this, you should not take this seminar. The grading will follow strict rules, and there will be no room for pleas. I am fed up with the entitlement attitude from some of the students. http://t.co/YzRUw7XV
# [[Conflict-free replicated data types|http://pagesperso-systeme.lip6.fr/Marc.Shapiro/papers/CRDTs_SSS-2011.pdf]]\n# [[Key-CRDT Stores|http://run.unl.pt/bitstream/10362/7802/1/Sousa_2012.pdf]]\n# [[Abstract unordered and ordered trees CRDT|http://arxiv.org/pdf/1201.1784v1]]\n# [[CAP 12 years later|http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=6133253]]\n# [[A Stabilizing Search Tree with Availability Properties|http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=917445]]\n# [[Stabilizing Replicated Search Trees|http://link.springer.com/chapter/10.1007%2F3-540-45414-4_22?LI=true]]\n# [[Paxos made moderately complex|http://www.cs.cornell.edu/courses/cs7412/2011sp/paxos.pdf]]\n# [[MDCC Multi-Data Center Consistency|http://arxiv.org/pdf/1203.6049v1.pdf]]\n# [[Mencius: building efficient replicated state machines for Wide-Area-Networks (WANs)|http://www.sysnet.ucsd.edu/sysnet/miscpapers/mencius-osdi.pdf]]\n# [[Egalitarian Paxos|http://www.pdl.cmu.edu/PDL-FTP/associated/CMU-PDL-12-108.pdf]]\n# [[Transactional storage for geo-replicated systems|http://research.microsoft.com/en-us/people/aguilera/walter-sosp2011.pdf]]\n# [[Surviving congestion in geo-distributed storage systems|http://research.microsoft.com/en-us/people/aguilera/vivace-atc2012.pdf]]\n# [[Gnothi: Separating Data and Metadata for Efficient and Available Storage Replication|http://www.cs.utexas.edu/~yangwang/Gnothi_Wang.pdf]]\n# [[Dual-Quorum: A Highly Available and Consistent Replication System for Edge Services|http://www.cs.utexas.edu/~lorenzo/papers/middleware05.pdf]]\n# [[Probabilistically Bounded Staleness for Practical Partial Quorums|http://vldb.org/pvldb/vol5/p776_peterbailis_vldb2012.pdf]]\n# [[The Potential Dangers of Causal Consistency and an Explicit Solution|http://db.cs.berkeley.edu/papers/socc12-explicit.pdf]]\n# [[File system on CRDT|http://arxiv.org/pdf/1207.5990.pdf]]\n# GlusterFS: http://pl.atyp.us/papers/summit-glusterfs-2012.odp\nhttp://hekafs.org/index.php/2012/03/glusterfs-algorithms-distribution/\nhttp://hekafs.org/index.php/2012/03/glusterfs-algorithms-replication-present/\n# [[Distributed GraphLab: A Framework for Machine Learning and Data Mining in the Cloud|http://vldb.org/pvldb/vol5/p716_yuchenglow_vldb2012.pdf]]\n# [[PowerGraph: Distributed Graph-Parallel Computation on Natural Graphs|http://www.select.cs.cmu.edu/publications/paperdir/osdi2012-gonzalez-low-gu-bickson-guestrin.pdf]]\n# [[Idempotence is not a medical condition|http://dl.acm.org/citation.cfm?id=2187821]]\n# [[Fault Tolerance via Idempotence|http://research.microsoft.com/pubs/173885/popl38-ramalingam.pdf]]\n# [[Engagements: Building Eventually ACiD Business Transactions|http://www.cidrdb.org/cidr2013/Papers/CIDR13_Paper142.pdf]]\n# [[Consistency Tradeoffs in Modern Distributed Database System Design: CAP is Only Part of the Story|http://cs-www.cs.yale.edu/homes/dna/papers/abadi-pacelc.pdf]]\n