At RICON, our goal is to provide attendees a holistic perspective of the distributed systems industry. That expansive lens includes views from industry veterans, insightful newcomers and learned developers; views from large technology leaders as well as hot new companies and consultants. Their expertise is invaluable to providing a well-rounded understanding of both the challenges facing distributed system architecture and solutions for overcoming those challenges.
A holistic view would not be complete without a significant contribution from academia. In our previous post we highlighted talks from customers that shed light on those challenges. Now, here’s a snapshot of some of the innovations from the academic world. Get a preview of what researchers and PhD students believe is the future of distributed systems and how their research will affect our industry in the years to come.
When Weak Consistency is not Enough: From Causality to Transactions
Ensuring transactional consistency across distributed system can be a headache. Often, consistency and low latency exist in a give-and-take relationship — providing both can be difficult. Different types of consistency offer varying guarantees and degrees of semantic strength, and deciding which model to follow is a difficult decision. João Leitão, an integrated researcher at the NOVA Laboratory for Computer Science and Informatics (NOVA LINCS) in Lisbon, will discuss why developers might want to opt for stronger semantics than weak consistency in geo-replicated data stores. Leitão will describe several solutions that offer either causal consistency or transactional semantics, including an example that provides causal+ (causal consistency with eventual convergence guarantees) consistency.
Demystifying Distributed Transactions with the Fairness-Isolation-Throughput Tradeoff
You’re likely attending RICON because you’re aware of the capabilities NoSQL systems offer. While NoSQL system flexibility is a huge benefit for developers, many solutions forego atomic transactions, a programmer-friendly facet of traditional DBMSs. Atomicity in a distributed system requires each machine precisely communicate with the other machines involved in the transaction. As such, many perceive atomic transactions to have prohibitive performance cost, especially in distributed settings. Yet, Jose Faleiro, a fourth year PhD student at Yale University who conducts research on concurrency in parallel and distributed systems, asks, “Does distributed coordination necessarily preclude performant general atomic transactions?” He’ll answer that question in a talk that focuses on key performance and correctness properties.
Minimizing Faulty Executions of Distributed Systems
When distributed systems go buggy, developers must uncover the events that caused the faults before they can squash the bugs. However, troubleshooting these causes is a time-intensive task, especially when developers must comb through thousands of events. Luckily, Colin Scott, a PhD candidate at UC Berkeley who works on developing practice methods to discover and fix bugs in distributed systems, will present a solution that minimizes the amount of time and effort developers must devote to troubleshooting.
The Quest for Invariant Preserving Replication
Distributed systems operate in a world of tradeoffs. For instance, it can be difficult to include both systems correctness and high performance performance in a distributed system, all while maintaining consistent data in NoSQL stores. Valter Balegas, a PhD student at Universidade Nova de Lisboa, works to improve correctness properties in geo-located systems. He’ll divulge how his research is uncovering how developers can include both systems correctness and performance in distributed systems. Balegas’ prototype uses Riak Core to help achieve these goals and his talk will help you understand how to achieve yours.
There will always be tradeoffs when it comes to database development. As organizational needs change, processing capabilities evolve and data continues to mount, developers must constantly search for new ways to provide availability and consistency without sacrificing system performance. We are excited to welcome these researchers and their innovative solutions to problems everyone in the distributed systems world encounters. Their prototypes and methodology are sure to offer insight and inspiration and help provide the wide-angle view of distributed databases RICON aims to deliver.
Learn With Us at RICON 2015
RICON is the perfect mix of practical tips and academic research. Don’t miss this opportunity to join hundreds of the top distributed systems talent.
There’s still time to register for RICON 2015. Register here and use offer code Basho100 for $100 off your ticket price. Full RICON 2015 Agenda and event details are available at www.RICON.io