Auto-reconfiguration for Latency Minimization in CPU-based DNN Serving

Ankit Bhardwaj, Amar Phanishayee, Deepak Narayanan, Ryan Stutsman
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:4115-4129, 2025.

Abstract

In this paper, we investigate how to push the performance limits of serving Deep Neural Network (DNN) models on CPU-based servers. Specifically, we observe that while intra-operator parallelism across multiple threads is an effective way to reduce inference latency, it provides diminishing returns. Our primary insight is that instead of running a single instance of a model with all available threads on a server, running multiple instances each with smaller batch sizes and fewer threads for intra-op parallelism can provide lower inference latency. However, the right configuration is hard to determine manually since it is workload- (DNN model and batch size used by the serving system) and deployment-dependent (number of CPU cores on server). We present Packrat, a new serving system for online inference that given a model and batch size ($B$) algorithmically picks the optimal number of instances ($i$), the number of threads each should be allocated ($t$), and the batch sizes each should operate on ($b$) that minimizes latency. Packrat is built as an extension to TorchServe and supports online reconfigurations to avoid serving downtime. Averaged across a range of batch sizes, Packrat improves inference latency by 1.43$\times$ to 1.83$\times$ on a range of commonly used DNNs.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-bhardwaj25a, title = {Auto-reconfiguration for Latency Minimization in {CPU}-based {DNN} Serving}, author = {Bhardwaj, Ankit and Phanishayee, Amar and Narayanan, Deepak and Stutsman, Ryan}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {4115--4129}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/bhardwaj25a/bhardwaj25a.pdf}, url = {https://proceedings.mlr.press/v267/bhardwaj25a.html}, abstract = {In this paper, we investigate how to push the performance limits of serving Deep Neural Network (DNN) models on CPU-based servers. Specifically, we observe that while intra-operator parallelism across multiple threads is an effective way to reduce inference latency, it provides diminishing returns. Our primary insight is that instead of running a single instance of a model with all available threads on a server, running multiple instances each with smaller batch sizes and fewer threads for intra-op parallelism can provide lower inference latency. However, the right configuration is hard to determine manually since it is workload- (DNN model and batch size used by the serving system) and deployment-dependent (number of CPU cores on server). We present Packrat, a new serving system for online inference that given a model and batch size ($B$) algorithmically picks the optimal number of instances ($i$), the number of threads each should be allocated ($t$), and the batch sizes each should operate on ($b$) that minimizes latency. Packrat is built as an extension to TorchServe and supports online reconfigurations to avoid serving downtime. Averaged across a range of batch sizes, Packrat improves inference latency by 1.43$\times$ to 1.83$\times$ on a range of commonly used DNNs.} }
Endnote
%0 Conference Paper %T Auto-reconfiguration for Latency Minimization in CPU-based DNN Serving %A Ankit Bhardwaj %A Amar Phanishayee %A Deepak Narayanan %A Ryan Stutsman %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-bhardwaj25a %I PMLR %P 4115--4129 %U https://proceedings.mlr.press/v267/bhardwaj25a.html %V 267 %X In this paper, we investigate how to push the performance limits of serving Deep Neural Network (DNN) models on CPU-based servers. Specifically, we observe that while intra-operator parallelism across multiple threads is an effective way to reduce inference latency, it provides diminishing returns. Our primary insight is that instead of running a single instance of a model with all available threads on a server, running multiple instances each with smaller batch sizes and fewer threads for intra-op parallelism can provide lower inference latency. However, the right configuration is hard to determine manually since it is workload- (DNN model and batch size used by the serving system) and deployment-dependent (number of CPU cores on server). We present Packrat, a new serving system for online inference that given a model and batch size ($B$) algorithmically picks the optimal number of instances ($i$), the number of threads each should be allocated ($t$), and the batch sizes each should operate on ($b$) that minimizes latency. Packrat is built as an extension to TorchServe and supports online reconfigurations to avoid serving downtime. Averaged across a range of batch sizes, Packrat improves inference latency by 1.43$\times$ to 1.83$\times$ on a range of commonly used DNNs.
APA
Bhardwaj, A., Phanishayee, A., Narayanan, D. & Stutsman, R.. (2025). Auto-reconfiguration for Latency Minimization in CPU-based DNN Serving. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:4115-4129 Available from https://proceedings.mlr.press/v267/bhardwaj25a.html.

Related Material