Middleware-enforced Timed Causal Consistency for Apache Cassandra: An energy–performance–consistency evaluation against static consistency levels using YCSB

Key findings from this study

• The study demonstrates that middleware-enforced Strict Timed Causal Consistency enables adaptive runtime consistency tuning in Cassandra without modifications to the core database engine.
• The authors report that their framework provides integrated tools for measuring consistency-energy-performance trade-offs across geo-replicated clusters using standard benchmarking and monitoring infrastructure.
• The researchers validate that external consistency layers can seamlessly integrate with unmodified Cassandra deployments while supporting configurable freshness bounds and causal guarantees.

Overview

The study presents a middleware-enforced consistency model called Strict Timed Causal Consistency (STCC) that operates as an external layer for Apache Cassandra. STCC combines client-side and server-side causal guarantees with a configurable freshness bound delta, enabling adaptive consistency tuning at runtime without modifying the core database engine. The middleware integrates with unmodified Cassandra deployments across geo-replicated nodes, addressing limitations of static consistency levels (ALL, QUORUM, ONE) that cannot adapt to workload intensity variations or network condition changes.

Methods and approach

The researchers deployed a 24-node Cassandra cluster distributed across three data centers and executed YCSB-based workloads against both STCC and static consistency levels. The evaluation framework includes automation tools for workload execution, CPU frequency control via cpupower, and real-time energy monitoring using SNMP-enabled power distribution units and dstat. The benchmarking suite measures consistency-latency-energy trade-offs under realistic deployment conditions while maintaining full reproducibility through open-source release with complete documentation.

Results

The evaluation quantified energy, performance, and consistency trade-offs between STCC and built-in static levels across the 24-node test cluster. STCC demonstrated the ability to adapt consistency guarantees at runtime based on configurable freshness bounds without requiring core database modifications. The middleware achieved seamless integration with existing Cassandra instances, enabling tunable consistency levels while monitoring real-time power consumption and latency metrics across geo-replicated deployments.

Implications

The reproducible framework enables cloud engineers and distributed systems researchers to systematically analyze consistency-latency-energy trade-offs in NoSQL deployments. Middleware-enforced consistency models offer an alternative to static configuration approaches, allowing runtime optimization responses to workload variations and network conditions. The public release with full tooling and documentation supports extensible experimentation for evaluating adaptive consistency strategies in production-scale clusters.

Scope and limitations

This summary is based on the study abstract and available metadata. It does not include a full analysis of the complete paper, supplementary materials, or underlying datasets unless explicitly stated. Findings should be interpreted in the context of the original publication.