PWLNYC is extremely excited to be hosting Deniz Altınbüken (http://www.cs.cornell.edu/~deniz/), a Ph.D. candidate in Distributed Systems at Cornell University. We were first introduced to her work upon seeing her amazing Ricon 2015 talk (https://www.youtube.com/watch?v=vW8pN07cH8M) and reading through Paxos: Made Moderately Complex (http://paxos.systems/). She's travelling from Cornell to join us in talking about the topic of Chain Replication (CR) (http://www.cs.cornell.edu/home/rvr/papers/osdi04.pdf) and their new and improved protocol for it, which has a formally defined end-to-end specification.
For those unfamiliar with CR, Christopher Meiklejohn (https://christophermeiklejohn.com/) gave a fantastic survey talk (http://paperswelove.org/2015/video/nathan-taylor-chris-meiklejohn-os-scalability-chain-replication/) at Papers We Love San Francisco (http://paperswelove.org/chapter/sanfrancisco/) a little while back (http://www.meetup.com/papers-we-love-too/events/224178870/). Here's an overview (http://paperswelove.org/2015/topic/christopher-meiklejohns-a-brief-history-of-chain-replication/) of some of those papers on CR. We can't wait to see Deniz and her collaborators' new work in the field!
Influenced a bit by some of our other chapters (http://www.meetup.com/papers-we-love-too/), we're adding a special 15-minute PWL-mini presentation by Wes Chow, CTO of Charbeat (https://chartbeat.com/), on Tiered Replication: A Cost-effective Alternative to Full Cluster Geo-replication (https://www.usenix.org/system/files/conference/atc15/atc15-paper-cidon.pdf)!
7:30 ~ 7:45-> PWL Mini w/ Wes Chow on Tiered Replication: A Cost-effective Alternative to Full Cluster Geo-replication (https://www.usenix.org/system/files/conference/atc15/atc15-paper-cidon.pdf)
7:50 ~ ->Deniz Altınbüken on Chain Replication (old and new)
• Chain Replication
Chain Replication (CR) is a variant of Primary-Backup Replication that supports high throughput and fast recovery from failures. CR has been widely used in both commercial systems and academic research prototypes. In so doing, various shortcomings of the original CR protocol have come to light. In this talk, I will summarize these findings and present a new version of CR that addresses the shortcomings. Our improved CR protocol supports different consistency guarantees, avoids the tail bottleneck for reads, and introduces autonomous reconfiguration of the system without requiring an external master. Additionally, we have developed a formal end-to-end specification of the protocol, including the actions of clients, detailing reconfiguration and linearizable execution of client requests. Through this specification, we are able to reason about the new protocol more precisely and implement the protocol effortlessly. Lastly, I will contrast our approach to the related work.
• Tiered Replication
Tiered Replication, by Cidon et. al, explores the problem of effective data replication strategies first introduced in the Copysets paper (https://www.usenix.org/system/files/conference/atc15/atc15-paper-cidon.pdf), awarded 2013 Usenix ATC Best Student Paper (https://www.usenix.org/conference/atc15/technical-session/presentation/cidon). While Copysets introduced a randomized algorithm for solving NP Hard redundancy and load balancing constraints around placement of data in distributed filesystem, Tiered Replication proposes a greedy algorithm for solving the same problem and also adding the ability to bake in real world constraints such as rack awareness. Wes will summarize the problem Copysets proposed, show Tiered Replication’s solution (https://github.com/chartbeat-labs/trepl), and examine a real world deployment of the algorithm at Chartbeat.
Deniz Altınbüken (http://www.cs.cornell.edu/~deniz/) is a Ph.D. candidate in Distributed Systems at Cornell University, working with Robbert van Renesse (https://www.cs.cornell.edu/home/rvr/). Her interests are in distributed systems and the theory of distributed computing with a focus on building infrastructure services for large-scale distributed systems. She loves breaking distributed systems down to their basic components, understanding the bits and pieces of what makes them work and explaining these in precise ways with the objective of making complicated systems easier to understand and implement.
Wes Chow has a B.S. in Electrical Engineering & Computer Science from UC Berkeley. He spent eight years building technical infrastructure for high frequency trading shops. One day, he stared into his dark soul and realized he needed to move into the startup light. Thus S7 Labs sprang into being, and he led teams that built Storybox, a Seedcamp NY finalist, and Songza Radio, subsumed by Google Music. He's now at Chartbeat serving out his term as CTO.
TwoSigma (https://www.twosigma.com/) - Platinum Sponsor of the New York chapter
Doors open at 7 pm; the presentation will begin at 7:30 pm; and, yes, there will be refreshments of all kinds and pizza.
A little different than previous PWLs, you'll have to check-in with security with your Name/ID. Definitely sign-up if you’re going to attend–unfortunately people whose names aren’t entered into the security system in advance won’t be allowed in.
After Deniz presents her work, we will open up the floor to discussion and questions.
We hope that you'll read the paper(s) before the meetup, but don't stress if you can't. If you have any questions, thoughts, or related information, please visit #pwlnyc (https://paperswelove.slack.com/messages/pwlnyc/) on slack (http://papersweloveslack.herokuapp.com/), our GitHub repository (https://github.com/papers-we-love/papers-we-love), where you can also find the papers (https://github.com/papers-we-love/papers-we-love/blob/baed7e16c0456b6395796fe0af70f4668391e9b9/distributed_systems/README.md), or add to the discussion on this event's thread.
Additionally, if you have any papers you want to add to the repository above (papers that you love!), please send us a pull request (https://github.com/papers-we-love/papers-we-love/pulls). Also, if you have any ideas/questions about this meetup or the Papers-We-Love org, just open up an issue.