Organizations: University of California, Los Angeles, Los Angeles, USA · University of California, Berkeley, Berkeley, USA · New York University, New York, USA · Independent Researcher, Mukilteo, USA · Fordham University, New York, USA · Independent Researcher, Washington, USA
Serverless computing eliminates infrastructure management overhead but introduces significant challenges regarding cold start latency and resource utilization. Traditional static resource allocation often leads to inefficiencies under variable workloads, resulting in performance degradation or excessive costs. This paper presents an adaptive engineering framework that optimizes serverless performance through event-driven architecture and probabilistic modeling. We propose a dual-strategy mechanism that dynamically adjusts idle durations and employs an intelligent request waiting strategy based on slot survival predictions. By leveraging sliding window aggregation and asynchronous processing, our system proactively manages resource lifecycles. Experimental results show that our approach reduces cold starts by up to 51.2% and improves cost-efficiency by nearly 2x compared to baseline methods in multi-cloud environments.
Figures & tables
Figure 1: System architecture overview showing the five-layer design from request sources through the core optimization engine to resource allocation decisions.
Figure 2: Dynamic resource lifecycle management showing state transitions and adaptive parameter formulations for idle duration, wait probability, and online learning.
Figure 3: Data preprocessing pipeline visualization. (a) Multi-resolution temporal analysis showing four aggregation metrics computed from adaptive time bins over a 24-hour period, revealing distinct activity patterns during peak hours and burst events. (b) Request categorization results via online clustering and incremental PCA, identifying five resource consumption profiles with 95% confidence ellipses.
Figure 4: Training convergence of different methods across three workload patterns. Columns represent four evaluation metrics: Cold Start Reduction Rate (CSRR), Resource Utilization Efficiency (RUE), Adaptive Response Latency (ARL), and Cost-Performance Index (CPI). Rows correspond to Dataset-1 (high-frequency uniform), Dataset-2 (periodic burst), and Dataset-3 (irregular sparse). ASRM (Ours) demonstrates faster convergence and superior final performance across all configurations.
Method
CSRR
RUE
ARL
CPI
Dataset-1: High-frequency Uniform Workload
OpenWhisk-Default
0.000
0.423
0.872
1.000
SOCK
0.284
0.512
0.658
1.342
Firecracker-snap
0.317
0.538
0.621
1.428
Knative-KPA
0.256
0.487
0.734
1.218
ASRM (Ours)
0.512
0.687
0.421
1.876
Table 1: Performance Comparison Across Different Workload Patterns
Department of Engineering, University of Perugia, Perugia, Italy · Consorzio Nazionale Interuniversitario per le Telecomunicazioni (CNIT), 43124 Parma, Italy