.png){kind=icon}
.png){kind=icon}
Return
CGP-Graphless: Towards Efficient Serverless Graph Processing via CPU-GPU Pipelined Collaboration
.png){kind=icon}
PRE
AI
DOI：10.1007/978-3-031-99854-6_23.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The serverless computing model offers users flexible, pay-as-you-go services. However, existing frameworks face challenges such as resource over-subscription and workload over-scaling when deploying graph processing jobs in serverless environments. To address these limitations, we introduce CGP-Graphless, a CPU-GPU heterogeneous computing framework designed for efficient vertical scaling. This approach divides graph processing into two phases: querying on a core proxy graph and correction on the full graph. GPU containers execute the querying phase, while CPU containers perform the correction. Furthermore, we propose an adaptive pipelined scheduling strategy for these phases, which leverages pressure-aware intra-pipeline scaling to convert excessive horizontal scaling into vertical scaling, enhancing serverless graph computation efficiency. Experiments show that CGP-Graphless improves end-to-end performance by up to 2.00x over FaaSGraph under concurrent stress evaluations, while reducing CPU core allocation by half through GPU container utilization. In short-interval query scenarios, CGP-Graphless further reduces average request latency by 3.30x compared to FaaS-Graph.
Keywords:
Serverless Computing
Graph Processing
Heterogeneous computing
Vertical Scaling
Journal
E
IF:
0
Papers:
28
Citations:
0