Effective Techniques for Message Reduction and Load Balancing in Distributed Graph Computation

Abstract

Massive graphs, such as online social networks and communication networks, have become common today. To efficiently analyze such large graphs, many distributed graph computing systems have been developed. These systems employ the "think like a vertex" programming paradigm, where a program proceeds in iterations and at each iteration, vertices exchange messages with each other. However, using Pregel's simple message passing mechanism, some vertices may send/receive significantly more messages than others due to either the high degree of these vertices or the logic of the algorithm used. This forms the communication bottleneck and leads to imbalanced workload among machines in the cluster. In this paper, we propose two effective message reduction techniques: (1)vertex mirroring with message combining, and (2)an additional requestrespond API. These techniques not only reduce the total number of messages exchanged through the network, but also bound the number of messages sent/received by any single vertex. We theoretically analyze the effectiveness of our techniques, and implement them on top of our open-source Pregel implementation called Pregel+. Our experiments on various large real graphs demonstrate that our message reduction techniques significantly improve the performance of distributed graph computation.

A flexible method for optimising sharing of healthcare resources and demand in the context of the COVID-19 pandemic

“…In anticipation of these scenarios, here we design and implement a simple and flexible load sharing procedure which we hope can help to alleviate the level of stress of healthcare systems and implement and test with information for the UK National Health Service (NHS) and the Spanish health system. Graph-embedded load balancing [9,10] has been mainly explored in computer science, usually taking a 'vertex perspective' for graphical computation with the aim of achieving a centralised solution to load allocation, subject to locality and availability constraints [11]. Interestingly, this line usually relates to minimise large-scale computational efforts, rather than actually sharing physical resources.…”

Composing Optimization Techniques for Vertex-Centric Graph Processing via Communication Channels

“…We test Pregel+'s basic implementation, Pregel+'s ghost mode (a.k.a. the mirroring technique [29]), the standard channel version ( Fig. 1) and the scatter-combine channel version.…”

“…While Pregel provides a friendly interface for processing massive graphs, current research shows that it is important to introduce optimizations for dealing with various performance issues such as imbalanced workload (a.k.a. skewed degree distribution) [2], [8], [29], redundancies in communication [3], [19], [29] and low convergence speed [23], [24], [28]. However, there remains one challenge: although the usefulness of these optimizations are well demonstrated in solving simple algorithms such as PageRank and single-source shortest path (SSSP) 1 , it is, however, hard to combine them together to implement complex algorithms, where we may have to deal with multiple performance issues at the same time.…”

“…Essentially, it is an iterative algorithm with two key operations -pointer jumping and tree merging. For the pointer jumping operation, the communication suffers from imbalanced workload [29], and in the meantime in the tree merging operation, the neighborhood communication (every vertex broadcasts a message to all of its own neighbors) could be potentially very heavy. Although there are techniques in separate systems [3], [19], [29] dealing with each case, there is no system capable of optimizing away both issues at the same time.…”

Big Graph Analytics Platforms

“…For example, MOCgraph , GraphD [Yan et al, 2016d], and the superstep-splitting technique of Giraph all propose aggregating messages earlier instead of buffering them for later processing, in order to save memory space; while PowerGraph , GraphChi and X-Stream [Roy et al, 2013] assume that data values are aggregated at each vertex from its incoming edges, in their model design. We, however, would like to indicate that not all algorithms with edge-based communication allow its vertices to aggregate received values, such as the attribute broadcast algorithm of Yan et al [2015].…”