# Symphony: Optimized DNN Model Serving using Deferred Batch Scheduling

arXiv: https://arxiv.org/abs/2308.07470 · HTML: https://mlsanity.com/paper/2308.07470/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 27 of its 50 references that are in the map.
- First posted on arXiv 14 Aug 2023.

## Builds on (most cited first, 27 of 27)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Very Deep Convolutional Networks for Large-Scale Image Recognition](https://mlsanity.com/paper/1409.1556/) (2014)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [Densely Connected Convolutional Networks](https://mlsanity.com/paper/1608.06993/) (2016)
- [EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks](https://mlsanity.com/paper/1905.11946/) (2019)
- [Rethinking the Inception Architecture for Computer Vision](https://mlsanity.com/paper/1512.00567/) (2015)
- [MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications](https://mlsanity.com/paper/1704.04861/) (2017)
- [SSD: Single Shot MultiBox Detector](https://mlsanity.com/paper/1512.02325/) (2015)
- [Xception: Deep Learning with Depthwise Separable Convolutions](https://mlsanity.com/paper/1610.02357/) (2016)
- [Searching for MobileNetV3](https://mlsanity.com/paper/1905.02244/) (2019)
- [Identity Mappings in Deep Residual Networks](https://mlsanity.com/paper/1603.05027/) (2016)
- [GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers](https://mlsanity.com/paper/2210.17323/) (2022)
- [Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism](https://mlsanity.com/paper/1909.08053/) (2019)
- [Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning](https://mlsanity.com/paper/1602.07261/) (2016)
- [AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration](https://mlsanity.com/paper/2306.00978/) (2023)
- [EfficientNetV2: Smaller Models and Faster Training](https://mlsanity.com/paper/2104.00298/) (2021)
- [Learning Transferable Architectures for Scalable Image Recognition](https://mlsanity.com/paper/1707.07012/) (2017)
- [LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale](https://mlsanity.com/paper/2208.07339/) (2022)
- [Clipper: A Low-Latency Online Prediction Serving System](https://mlsanity.com/paper/1612.03079/) (2016)
- [Serving DNNs like Clockwork: Performance Predictability from the Bottom Up](https://mlsanity.com/paper/2006.02464/) (2020)
- [TensorFlow-Serving: Flexible, High-Performance ML Serving](https://mlsanity.com/paper/1712.06139/) (2017)
- [PRETZEL: Opening the Black Box of Machine Learning Prediction Serving Systems](https://mlsanity.com/paper/1810.06115/) (2018)
- [Llama: A Heterogeneous & Serverless Framework for Auto-Tuning Video Analytics Pipelines](https://mlsanity.com/paper/2102.01887/) (2021)
- [Perseus: Characterizing Performance and Cost of Multi-Tenant Serving for CNN Models](https://mlsanity.com/paper/1912.02322/) (2019)
- [DistCache: Provable Load Balancing for Large-Scale Storage Systems with Distributed Caching](https://mlsanity.com/paper/1901.08200/) (2019)

## Built on by (most cited first, 2 of 2)

- [Approximate Caching for Efficiently Serving Diffusion Models](https://mlsanity.com/paper/2312.04429/) (2023)
- [EdgeServing: Deadline-Aware Multi-DNN Serving at the Edge](https://mlsanity.com/paper/2605.05527/) (2026)

## Abstract (arXiv)

Having large batch sizes is one of the most critical aspects of increasing the accelerator efficiency and the performance of DNN model inference. However, existing model serving systems cannot achieve adequate batch sizes while meeting latency objectives as these systems eagerly dispatch requests to accelerators to minimize the accelerator idle time. We propose Symphony, a DNN serving system that explores deferred batch scheduling to optimize system efficiency and throughput. Further, unlike other prior systems, Symphony's GPU usage is load-proportional: it consolidates workloads on the appropriate number of GPUs and works smoothly with cluster auto-scaling tools. Symphony consists of two core design points. First, Symphony defines a schedulable window in which a batch of inference requests can be dispatched. This window is computed in order to improve accelerator efficiency while meeting the request's SLO. Second, Symphony implements a scalable, low-latency, fine-grained coordination scheme across accelerators to dispatch and execute requests in the schedulable window. Through extensive scheduler-only benchmarks, we demonstrate that Symphony can schedule millions of requests per second and coordinate thousands of GPUs while also enabling robust autoscaling that adapts to workload changes. Symphony outperforms prior systems by achieving 5x higher goodput when given the same number of GPUs and 60% reduction in GPUs when given the same workload.