Cache Performance Benchmark
The Cache Performance Benchmark evaluates how Timbr’s 4-tier semantic cache accelerates query performance across different storage tiers.
Caching is critical for BI workloads because repeated queries and aggregations on large datasets can overwhelm compute resources, driving up costs and slowing down dashboards.
CAPABILITIES – 4 TIER CACHE
Timbr introduces a semantic-aware caching layer that understands ontology hierarchies and pre-aggregates results intelligently. This allows metrics to be answered faster while reducing compute load, improving both performance and cost-efficiency.
benchmark settings
For this benchmark, we used a dataset of over 100 GB in Delta tables and Microsoft Power BI as the target application. We compared response times when running complex metrics across Synapse views and Timbr’s materialized views, stored at three tiers: ADLS, SSD, and In-Memory.
Measuring Performance
in Views vs.
Materialized Views
Representative metrics such as Risk Score, Invoice Amount, and Customer Cases were executed against both Synapse views and Timbr’s cache tiers. The results demonstrate how materialized views, especially SSD and In-Memory, consistently reduced execution times from minutes to sub-seconds.
This benchmark highlights how semantic caching transforms BI performance at scale: instead of overloading compute engines with repeated aggregations, results are served instantly from optimized caches while preserving accuracy and governance.
Execution Time Comparison Across Tiers
| View | Materialized View | |||
|---|---|---|---|---|
| Metric / Source | Synapse | ADLS | SSD | In-Memory |
| Metric#1: Risk Score | >10 min | 11.2 sec | 2.6 sec | 0.2 sec |
| Metric#2: Invoice Amount/count | 17.6 sec | 4.2 sec | 0.2 sec | |
| Metric#3: PR Amount/count | 15.5 sec | 3.2 sec | 0.1 sec | |
| Metric#4: PO Amount/count | 17.8 sec | 4.1 sec | 0.29 sec | |
| Metric#5: PAY Amount/count | 18.0 sec | 4.4 sec | 0.12 sec | |
| Metric#6: Total Cases | 9.1 sec | 2.1 sec | 0.7 sec | |
| Metric#7: Open Cases | 2.3 sec | 0.57 sec | 0.36 sec | |
| Metric#8: Closed Cases | 1.96 sec | 0.46 sec | 0.33 sec | |
Across representative metrics, Timbr’s semantic cache reduced execution times from >10 minutes in Synapse to sub-seconds in In-Memory cache, demonstrating the impact of semantic caching for BI at scale.
