Mitigating Cold-Start Latency in Serverless Edge Deployments via Adaptive Container Checkpointing and Predictive Pre-Warming
Keywords:
serverless computing, cold-start latency, container checkpointing, edge computing, predictive pre-warming, LSTM-based scheduling, function-as-a-service, Markov chain workload modeling, SLA complianceAbstract
Serverless computing at the network edge introduces critical cold-start latency penalties that undermine real-time service-level agreement (SLA) compliance in latency-sensitive workloads. This paper addresses the challenge by proposing AdaptWarm, a hybrid framework combining adaptive container checkpointing with a Long Short-Term Memory (LSTM)-driven predictive pre-warming scheduler. AdaptWarm continuously profiles invocation frequency distributions, constructs Markov-based transition models for function call sequences, and restores snapshotted execution contexts ahead of predicted demand spikes. Evaluated across heterogeneous edge nodes using the OpenFaaS and Knative runtimes with production-grade workload traces, AdaptWarm reduces mean cold-start latency by 73.4% and tail latency (P99) by 68.1% compared to baseline reactive provisioning, while maintaining memory overhead below 11%. These results establish AdaptWarm as a viable, deployment-ready solution for sub-millisecond function initialization in multi-tenant edge infrastructures.
References
Kumar, N., & Kataria, V. Enhanced Sentiment Classification using a Multi-layered Stacked Ensemble Architecture.
Kumar, Nitin, and Vipin Kataria. "Enhanced Sentiment Classification using a Multi-layered Stacked Ensemble Architecture."
Kumar, N., & Kataria, V. (2023). Enhanced Sentiment Classification using a Multi-layered Stacked Ensemble Architecture. International Journal of Intelligent Systems and Applications in Engineering, 11(4s), 304–311.
Satyanarayana, D. and Rao, S. V. (2011), “Fault-tolerant spanners for ad hoc networks” International Journal of Network Management, Vol: 22, Issue: 4, Pages: 311-329, July-2012, Wiley Publishers. https://doi.org/10.1002/nem.807
D. Satyanarayana, Sathyashree. S “Mobility Management in Voronoi based Cell Structure for Wireless networks”, IEEE proceedings for the International Conferences on Currents trends in Information Technology (CTIT), Pages: 204-208, Dubai, December-2013. DOI: 10.1109/CTIT.2013.6749504; INSPEC Accession Number: 14131032
D. Satyanarayana and J. M. H. Elmirghani, "An Energy Efficient Network Architecture for Infrastructured Wireless Networks," 2010 IEEE Global Telecommunications Conference GLOBECOM 2010, Miami, FL, USA, 2010, pp. 1-6, doi: 10.1109/GLOCOM.2010.5683234.
D. Satyanarayana and J. M. H. Elmirghani, "A Voronoi Based Energy Efficient Architecture for Wireless Networks," 200S Third International Conference on Next Generation Mobile Applications, Services and Technologies, Cardiff, UK, 2009, pp. 377-382, doi: 10.1109/NGMAST.2009.79.
D. Satyanarayana and S. V. Rao, "Local Delaunay Triangulation for Mobile Nodes," 2008 First International Conference on Emerging Trends in Engineering and Technology, Nagpur, India, 2008, pp. 282-287, doi: 10.1109/ICETET.2008.253.
D. Satyanarayana, "Greedy Local Delaunay Triangulation Routing for Wireless Ad Hoc Networks," 2007 International Conference on Signal Processing, Communications and Networking, Chennai, India, 2007, pp. 49-53, doi: 10.1109/ICSCN.2007.350694
Satyanarayana, D., Sathyashree, A. A. K., & Alkalbani, A. (2015, November). Voronoi LEACH for Energy Efficient Communication in Wireless Sensor Networks. In Proceedings for the International Conference on Electronics Computer networks and Information Technology, Dubai.
Rahmanov, N., Guliyev, H., İbtahimov, F., & Mammadov, Z. (2024). Determination of optimal dimensions of hybrid AC/DC distributed generation system with renewable sources for autonomous power supply of remote locations. In E3S Web of Conferences (Vol. 584, p. 01020). EDP Sciences.
Hashimov, A. M., Babayeva, A. R., & Guliyev, H. B. (2019). Shunt reactors control algorithm using fuzzy sets theory. International Journal on Technical and Physical Problems of Engineering, 11(1), 10-15.
