| Paper | Presentation |
There is a renewed interest in transactional database systems. In addition to large-scale cloud-native OLTP systems, there are novel proposals for HTAP and transactional lakehouse systems. At the same time, research proposals on highly available transactions are emerging, which avoid coordination and are particularly useful in high-concurrency and geo-replicated settings that have yet to see mainstream industrial adoption. However, in contrast to data storage and processing, in which there is an increasing interest in composability and there are widely available and used components that can be combined, transactional guarantees are monolithically built in each system.
In this paper we conduct a thought experiment on how to design database systems with composable transactional guarantees by examining and combining existing database systems with Cure, a research proposal for highly available transactional causal consistency. As a result, we derive lessons on desirable interfaces and useful building blocks for composable transactional implementations in database systems.
agent- The core agent implementation, as described in the paper.benchmark- Benchmarking client for evaluating the agent.bin- Useful scripts for interacting with all the components.deploy- Deployment scripts for running the agent and benchmarks.resources- Resources used in the paper, including figures.schema- Database schema used in PostgreSQL deployment.
Install Rust, Docker (with Docker Compose), and Ansible. For details on how to perform a deploy, see deploy/README.md.
Details on how to run benchmarks can be found in benchmark/README.md.
...
This work is licensed under a Creative Commons Attribution 4.0 International License.