Talk "Low latency vs scalability in causal consistency: trade-offs and protocols" by Diego Didona

Event details
Date | 04.12.2018 |
Hour | 14:15 › 15:15 |
Location | |
Category | Conferences - Seminars |
Abstrac :
Causal consistency is emerging as an interesting consistency level for building for geo-distributed systems, as it strikes a sweet point in the trade-off between performance and programming complexity. In this talk we present theoretical and practical results on causal consistency. On the theoretical side, we show a fundamental trade-off of causal consistency: it is impossible for a causal consistency implementation to achieve the lowest latency and the highest scalability.
In light of this result, we identify causal consistency designs that achieve feasible combinations of low latency and high scalability. We then show the implementation of such designs in Contrarian, a system that implements causal consistency, and in Wren, a system that supports generic transactions with causal consistency.
Bio :
Diego Didona received the MSc degree in computer engineering in 2010 from Sapienza Università di Roma, and the PhD degree in computer engineering from Instituto Superior Técnico, Universidade de Lisboa in 2015. He is a postdoctoral researcher at EPFL where he works on data center job scheduling, key-value stores, consistency in geo-replicated data platforms, and performance modeling applied to self-tuning systems.
Causal consistency is emerging as an interesting consistency level for building for geo-distributed systems, as it strikes a sweet point in the trade-off between performance and programming complexity. In this talk we present theoretical and practical results on causal consistency. On the theoretical side, we show a fundamental trade-off of causal consistency: it is impossible for a causal consistency implementation to achieve the lowest latency and the highest scalability.
In light of this result, we identify causal consistency designs that achieve feasible combinations of low latency and high scalability. We then show the implementation of such designs in Contrarian, a system that implements causal consistency, and in Wren, a system that supports generic transactions with causal consistency.
Bio :
Diego Didona received the MSc degree in computer engineering in 2010 from Sapienza Università di Roma, and the PhD degree in computer engineering from Instituto Superior Técnico, Universidade de Lisboa in 2015. He is a postdoctoral researcher at EPFL where he works on data center job scheduling, key-value stores, consistency in geo-replicated data platforms, and performance modeling applied to self-tuning systems.
Practical information
- Expert
- Free
- This event is internal
Organizer
- EcoCloud
Contact
- Valérie Locca