Data partitioning strategies for graph workloads on heterogeneous clusters

LeBeane, Michael; Song, Shuang; Panda, Rutuparna; Ryoo, Jee Ho; John, Lizy K.

doi:10.1145/2807591.2807632

Cited by 33 publications

(15 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These systems use graph partitioning to scale out computations on graphs or sparse matrices that do not fit in the memory of a single node. In the graph analytics literature, partitioning strategies are classified into Edge-Cuts [4,[22][23][24]41,42,50,51] and Vertex-Cuts [8,12,18,26,29,37,39,44]. In the matrix literature, they are classified into 1D and 2D partitionings [7,11].…”

Section: Related Workmentioning

confidence: 99%

A study of partitioning policies for graph analytics on large-scale distributed platforms

et al. 2018

View full text Add to dashboard Cite

Distributed-memory clusters are used for in-memory processing of very large graphs with billions of nodes and edges. This requires partitioning the graph among the machines in the cluster. When a graph is partitioned, a node in the graph may be replicated on several machines, and communication is required to keep these replicas synchronized. Good partitioning policies attempt to reduce this synchronization overhead while keeping the computational load balanced across machines. A number of recent studies have looked at ways to control replication of nodes, but these studies are not conclusive because they were performed on small clusters with eight to sixteen machines, did not consider work-efficient data-driven algorithms, or did not optimize communication for the partitioning strategies they studied. This paper presents an experimental study of partitioning strategies for work-efficient graph analytics applications on large KNL and Skylake clusters with up to 256 machines using the Gluon communication runtime which implements partitioning-specific communication optimizations. Evaluation results show that although simple partitioning strategies like Edge-Cuts perform well on a small number of machines, an alternative partitioning strategy called Cartesian Vertex-Cut (CVC) performs better at scale even though paradoxically it has a higher replication factor and performs more communication than Edge-Cut partitioning does. Results from communication micro-benchmarks resolve this paradox by showing that communication overhead depends not only on communication volume but also on the communication pattern among the partitions. These experiments suggest that high-performance graph analytics systems should support multiple partitioning strategies, like Gluon does, as no single graph partitioning strategy is best for all cluster sizes. For such systems, a decision tree for selecting a good partitioning strategy based on characteristics of the computation and the cluster is presented.

show abstract

Section: Related Workmentioning

confidence: 99%

A study of partitioning policies for graph analytics on large-scale distributed platforms

et al. 2018

View full text Add to dashboard Cite

show abstract

“…Such sketches represent data structures that approximate properties of a data stream. LeBeane et al [97] proposed on-line graphpartitioning multiple strategies to optimise data-ingress across heterogeneous clusters. SWORD [98] handles the partitioning and placement for OLTP workloads.…”

Section: Graph-based Data Placementmentioning

confidence: 99%

A survey on data storage and placement methodologies for Cloud-Big Data ecosystem

et al. 2019

View full text Add to dashboard Cite

Over the time, the type of applications has evolved from batch, compute or memory intensive applications to streaming or even interactive applications. As a result, applications are getting more complex and become long-running. Such applications might require frequent-access to multiple distributed data sources. During application deployment and provisioning, the user can face various issues such as (i) where to effectively place both the data and the computation; (ii) how to achieve required objectives while reducing the overall application running cost. Data could be generated from various sources, including a multitude of devices over IoT environments that can generate a huge amount of data, while the applications are running. An application can further produce a large amount of data. In general, data of such size is usually referred to as Big Data. In general, Big Data is characterised by five properties [1, 2]. These are volume, velocity (means rapid update and propagation of data), variety

show abstract

“…With a new composite graph, we further leverage a hybrid-cut algorithm [28], which attempts to perform an either vertex-or edge-cut method for the balance of workload and communication. We first give priority in the use of a vertex-cut partition (with a greedy heuristic algorithm [15]) to minimize the number of vertex mirrors for reducing the total storage as well as communication requirements (of symbol assignment).…”

Section: Activation-induced Graph Partitionmentioning

confidence: 99%

“…The vertex-cut partition [9,15,31] considers evenly placing all edges of graph for mitigating imbalanced communication. LeBeane et al [28] developed severalf graph partitioning strategies to investigate the heterogeneity of graph processing. Further, Song et al [44] presented a profiling-based approach to fully exploit the computational capability of each compute node for guiding better graph partitioning with minimal overhead.…”

Section: Related Workmentioning

confidence: 99%

Efficient and Scalable Graph Parallel Processing With Symbolic Execution

Zheng

Jin

2018

ACM Trans. Archit. Code Optim.

View full text Add to dashboard Cite

Existing graph processing essentially relies on the underlying iterative execution with synchronous (Sync) and/or asynchronous (Async) engine. Nevertheless, they both suffer from a wide class of inherent serialization arising from data interdependencies within a graph.In this article, we present SymGraph, a judicious graph engine with symbolic iteration that enables the parallelism of dependent computation on vertices. SymGraph allows using abstract symbolic value (instead of the concrete value) for the computation if the desired data is unavailable. To maximize the potential of symbolic iteration, we propose a chain of tailored sophisticated techniques, enabling SymGraph to scale out with a new milestone of efficiency for large-scale graph processing. We evaluate SymGraph in comparison to Sync, Async, and a hybrid of Sync and Async engines. Our results on 12 nodes show that SymGraph outperforms all three graph engines by 1.93x (vs. Sync), 1.98x (vs. Async), and 1.57x (vs. Hybrid) on average. In particular, the performance for PageRank on 32 nodes can be dramatically improved by 16.5x (vs. Sync), 23.3x (vs. Async), and 12.1x (vs. Hybrid), respectively. The efficiency of SymGraph is also validated with at least one order of magnitude improvement in contrast to three specialized graph systems (Naiad, GraphX, and PGX.D).

show abstract

Data partitioning strategies for graph workloads on heterogeneous clusters

Cited by 33 publications

References 18 publications

A study of partitioning policies for graph analytics on large-scale distributed platforms

A study of partitioning policies for graph analytics on large-scale distributed platforms

A survey on data storage and placement methodologies for Cloud-Big Data ecosystem

Efficient and Scalable Graph Parallel Processing With Symbolic Execution

Contact Info

Product

Resources

About