Fograph: Enabling Real-Time Deep Graph Inference with Fog Computing

Zeng, Liekang; Huang, Peng; Luo, Ke; Zhang, Xiaoxi; Zhou, Zhi; Chen, Xu

doi:10.1145/3485447.3511982

Cited by 10 publications

(4 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Distributed GNN systems. To support efficient GNN processing, a few frameworks have been developed in both research [6], [36]- [41] and industry communities [11], [12], [27], [42]- [44] to optimize performance at different levels. Towards general kernel-level optimizations, several works [12], [44], [45] endeavor to abstract a user-friendly programming model, and then apply a set of techniques under the unified interfaces by exploiting the execution primitives of GNN models.…”

Section: B Related Workmentioning

confidence: 99%

“…data collection) cannot be omitted. To unleash the architectural benefits of edge computing, Fograph [41] first investigates GNN processing with vicinal fog servers, and proposes a distributed inference system for real-time serving. Yet it merely pursues performance indicators on latency and throughput, where a comprehensive cost model for DGPE still lacks, demanding a new formulation for analysis.…”

Section: B Related Workmentioning

confidence: 99%

“…In our modeling, these cross-edge connections can be identified by e uv w ij x vi x uj , ∀u, v ∈ V, ∀i, j ∈ D, where e uv ∈ E and w ij ∈ W check whether there is subsistent links of u, v or i, j , respectively. Existing distributed GNN systems [11], [27], [41] typically leverage the Bulk Synchronous Parallel (BSP) model [78] to orchestrate parallelism over distributed servers, where a synchronization step is triggered when interserver data exchange is needed. We adopt this setting and refer to such synchronization as cross-edge traffic, with a unit communication cost τ ij for edge pairs i, j as depicted in Fig.…”

Section: B Cost Factorsmentioning

confidence: 99%

See 2 more Smart Citations

GNN at the Edge: Cost-Efficient Graph Neural Network Processing over Distributed Edge Servers

Zeng¹,

Chongyu²,

Huang³

et al. 2022

Preprint

View full text Add to dashboard Cite

Edge intelligence has arisen as a promising computing paradigm for supporting miscellaneous smart applications that rely on machine learning techniques. While the community has extensively investigated multi-tier edge deployment for traditional deep learning models (e.g. CNNs, RNNs), the emerging Graph Neural Networks (GNNs) are still under exploration, presenting a stark disparity to its broad edge adoptions such as traffic flow forecasting and location-based social recommendation. To bridge this gap, this paper formally studies the cost optimization for distributed GNN processing over a multitier heterogeneous edge network. We build a comprehensive modeling framework that can capture a variety of different cost factors, based on which we formulate a cost-efficient graph layout optimization problem that is proved to be NP-hard. Instead of trivially applying traditional data placement wisdom, we theoretically reveal the structural property of quadratic submodularity implicated in GNN's unique computing pattern, which motivates our design of an efficient iterative solution exploiting graph cuts. Rigorous analysis shows that it provides parameterized constant approximation ratio, guaranteed convergence, and exact feasibility. To tackle potential graph topological evolution in GNN processing, we further devise an incremental update strategy and an adaptive scheduling algorithm for lightweight dynamic layout optimization. Evaluations with real-world datasets and various GNN benchmarks demonstrate that our approach achieves superior performance over de facto baselines with more than 95.8% cost reduction in a fast convergence speed.

show abstract

Section: B Related Workmentioning

confidence: 99%

Section: B Related Workmentioning

confidence: 99%

Section: B Cost Factorsmentioning

confidence: 99%

See 1 more Smart Citation

GNN at the Edge: Cost-Efficient Graph Neural Network Processing over Distributed Edge Servers

Zeng¹,

Chongyu²,

Huang³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Firstly, available resources may be limited. It is common to perform heavy operations in the fog such as ML inference [7] and real-time video analysis [8]. Hence, such…”

Section: Introductionmentioning

confidence: 99%

TransScale: Combined-Approach Elasticity for Stream Processing in Fog Environments

Pagliari,

Pierre

2023

2023 11th IEEE International Conference on Mobile Cloud Computing, Services, and Engineering (MobileCloud)

View full text Add to dashboard Cite

Real-time data processing is a standard requirement in Fog Computing. Dynamically adapting data stream processing frameworks is an essential functionality to handle time-varying workloads efficiently and to optimize resource consumption. However, horizontal scaling alone, by adapting the parallelism and number of provisioned nodes, faces limits when available compute resources are scarce. We propose TransScale, a combined-approach auto-scaler that combines horizontal scaling to approximation computing, controlling it through transprecision computing. We design TransScale to make the approximation method transparent to the system and support context-specific requirements through QoS-driven re-configuration decisions. Based on the policy's objective, we show that it can reduce re-configuration occurrences, optimize resource utilization and sustain high workloads in resource-constrained environments.

show abstract

MOGR: Multi-task Offloading via Graph Representation in Heterogeneous Computing Network

Mal,

Gong,

Zeng

et al. 2024

ICC 2024 - IEEE International Conference on Communications

View full text Add to dashboard Cite

Fograph: Enabling Real-Time Deep Graph Inference with Fog Computing

Cited by 10 publications

References 52 publications

GNN at the Edge: Cost-Efficient Graph Neural Network Processing over Distributed Edge Servers

GNN at the Edge: Cost-Efficient Graph Neural Network Processing over Distributed Edge Servers

TransScale: Combined-Approach Elasticity for Stream Processing in Fog Environments

MOGR: Multi-task Offloading via Graph Representation in Heterogeneous Computing Network

Contact Info

Product

Resources

About