Sustainable serverless computing with cold-start optimization and automatic workflow resource scheduling

S Pan, H Zhao, Z Cai, D Li, R Ma… - IEEE Transactions on …, 2023 - ieeexplore.ieee.org
In recent years, serverless computing has garnered significant attention owing to its high
scalability, pay-as-you-go billing model, and efficient resource management provided by …

SPSC: Stream processing framework atop serverless computing for industrial big data

Z Cai, Z Chen, X Chen, R Ma, H Guan… - IEEE Transactions on …, 2024 - ieeexplore.ieee.org
With the advance of smart manufacturing and information technologies, the volume of data
to process is increasing accordingly. Current solutions for big data processing resort to …

Chitu: accelerating serverless workflows with asynchronous state replication pipelines

Z Lei, X Shi, C Lv, X Yu, X Zhao - … of the 2023 ACM symposium on cloud …, 2023 - dl.acm.org
Serverless workflows are characterized as multi-stage computing, while downstream
functions require accessing intermediate states or the output of upstream functions for …

FasDL: An Efficient Serverless-Based Training Architecture with Communication Optimization and Resource Configuration

X Chen, Z Cai, R Buyya - IEEE Transactions on Computers, 2024 - ieeexplore.ieee.org
Deploying distributed training workloads of deep learning models atop serverless
architecture alleviates the burden of managing servers from deep learning practitioners …

Measuring the impact of gradient accumulation on cloud-based distributed training

Z Huang, B Jiang, T Guo, Y Liu - 2023 IEEE/ACM 23rd …, 2023 - ieeexplore.ieee.org
Gradient accumulation (GA) is a commonly adopted technique for addressing the GPU
memory shortage problem in model training. It reduces memory consumption at the cost of …

On the performance and memory footprint of distributed training: An empirical study on transformers

Z Lu, F Wang, Z Xu, F Yang, T Li - ar** and deploying resource-aware AI models presents a compelling optimization
challenge in edge computing and serverless domains. Current research mainly focuses on …