Adaptive Caching and Request Routing for Latency-Sensitive Serverless Applications

Cold-start overhead and uneven placement of warm function instances can amplify request latency in serverless deployments. This paper studies a compact joint policy, AJR-ARF, combining latency-aware routing with a decayed recency-frequency warm-instance cache. At each arrival, the dispatcher estimates network, queueing, execution, and initialization components; after dispatch, a node replaces the cached function with the largest popularity-adjusted idle age. The method does not require offline training or workload forecasts. A deterministic discrete-event simulator was executed on changing-popularity synthetic requests across six nodes. Across five random seeds, the proposed policy produced 298.6 ms mean per-run 95th-percentile response latency, compared with 597.6 ms for nearest-node LRU, 392.3 ms for queue-aware LRU, and 302.8 ms for cache-aware LRU. The experiments also report cold-start frequency, SLO attainment, and sensitivity to cache capacity, traffic intensity, and popularity skew. The large reduction comes primarily from warm-state-aware routing, while adaptive eviction provides a smaller incremental change under this model.

Authors

Institutions

Publication Details

Journal
Academic Journal of Applied Sciences
Published
2026-11-16
DOI
https://doi.org/10.54097/nnt69h42
Primary Topic
Cloud Computing and Resource Management
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Adaptive Caching and Request Routing for Latency-Sensitive Serverless Applications

Iris Jansen, Daan Vermeer, Sophie van Leeuwen
Academic Journal of Applied Sciences
Cloud Computing and Resource Management
article

Adaptive Caching and Request Routing for Latency-Sensitive Serverless Applications

Iris Jansen, Daan Vermeer, Sophie van Leeuwen
article en

Abstract

Cold-start overhead and uneven placement of warm function instances can amplify request latency in serverless deployments. This paper studies a compact joint policy, AJR-ARF, combining latency-aware routing with a decayed recency-frequency warm-instance cache. At each arrival, the dispatcher estimates network, queueing, execution, and initialization components; after dispatch, a node replaces the cached function with the largest popularity-adjusted idle age. The method does not require offline training or workload forecasts. A deterministic discrete-event simulator was executed on changing-popularity synthetic requests across six nodes. Across five random seeds, the proposed policy produced 298.6 ms mean per-run 95th-percentile response latency, compared with 597.6 ms for nearest-node LRU, 392.3 ms for queue-aware LRU, and 302.8 ms for cache-aware LRU. The experiments also report cold-start frequency, SLO attainment, and sensitivity to cache capacity, traffic intensity, and popularity skew. The large reduction comes primarily from warm-state-aware routing, while adaptive eviction provides a smaller incremental change under this model.

Academic Journal of Applied SciencesVol. 3(3)
Utrecht University (NL), Eindhoven University of Technology (NL)
Cloud Computing and Resource Management
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon