On external-memory MST, SSSP and multi-way planar graph separation

We consider dynamic subgraph connectivity problems for planar undirected graphs. In this model there is a fixed underlying planar graph, where each edge and vertex is either off (failed) or (recovered). We wish to answer connectivity queries with respect to the subgraph. The model has two natural variants, one in which there are d edge/vertex failures that precede all connectivity queries, and one in which failures/recoveries and queries are intermixed. We present a d-failure connectivity oracle for planar graphs that processes any d edge/vertex failures in sort(d,n) time so that connectivity queries can be answered in pred(d,n) time. (Here sort and pred are the time for integer sorting and integer predecessor search over a subset of [n] of size d.) Our algorithm has two discrete parts. The first is an algorithm tailored to triconnected planar graphs. It makes use of Barnette's theorem, which states that every triconnected planar graph contains a degree-3 spanning tree. The second part is a generic reduction from general (planar) graphs to triconnected (planar) graphs. Our algorithm is, moreover, provably optimal. An implication of Pǎtrascu and Thorup's lower bound on predecessor search is that no d-failure connectivity oracle (even on trees) can beat pred(d,n) query time. We extend our algorithms to the subgraph connectivity model where edge/vertex failures (but no recoveries) are intermixed with connectivity queries. In triconnected planar graphs each failure and query is handled in O(logn) time (amortized), whereas in general planar graphs both bounds become O(log2n).

Breadth-first search (BFS) and single source shortest paths (SSSP) are two fundamental graph problems with countless real-world applications. It is of major interest to develop efficient parallel algorithms and implementations to solve these problems on modern multicore architectures. The challenge is that these computations are often irregular, and there is no known algorithm that guarantees polylogarithmic depth for all graphs. For BFS, this paper describe several performance engineering methods to develop a practical parallel implementation for all classes of real-world graphs. For SSSP, we introduce a unique contraction based preprocessing method which significantly speeds up queries on high diameter graphs, like road networks. Our method is generic and can be used with any algorithm for SSSP queries.

We present linear-I/O algorithms for fundamental graph problems on embedded outerplanar graphs. We show that breadth-first search, depth-first search, single-source shortest paths, triangulation, and computing an � separator of size O(1/� )t akeO(scan(N)) I/Os on embedded outerplanar graphs. We also show that it takes O(sort(N)) I/Os to test whether a given graph is outerplanar and to compute an outerplanar embedding of an outerplanar graph, thereby providing O(sort(N))-I/O algorithms for the above problems if no embedding of the graph is given. As all these problems have linear-time algorithms in internal memory, a simple simulation technique can be used to improve the I/O-complexity of our algorithms from O(sort(N)) to O(perm(N)). We prove matching lower bounds for embedding, breadth-first search, depth-first search, and singlesource shortest paths if no embedding is given. Our algorithms for the above problems use a simple linear-I/O time-forward processing algorithm for rooted trees whose vertices are stored in preorder.

Even though a large number of I/O-efficient graph algorithms have been developed, a number of fundamental problems still remain open. For example, no space- and I/O-efficient algorithms are known for depth-first search or breadth-first search in sparse graphs. In this paper we present two new results on I/O-efficient depth-first search in an important class of sparse graphs, namely undirected embedded planar graphs. We develop a new efficient depth-first search algorithm and show how planar depth-first search in general can be reduced to planar breadth-first search. As part of the first result we develop the first I/O-efficient algorithm for finding a simple cycle separator of a biconnected planar graph. Together with other recent reducibility results, the second result provides further evidence that external memory breadth-first search is among the hardest problems on planar graphs.

Even though a large number of I/O-efficient graph algorithms have been developed, a number of fundamental problems still remain open. For example, no space- and I/O-efficient algorithms are known for depth-first search or breadth-first search in sparse graphs. In this paper we present two new results on I/O-efficient depth-first search in an important class of sparse graphs, namely undirected embedded planar graphs. We develop a new efficient depth-first search algorithm and show how planar depth-first search in general can be reduced to planar breadth-first search. As part of the first result we develop the first I/O-efficient algorithm for finding a simple cycle separator of a biconnected planar graph. Together with other recent reducibility results, the second result provides further evidence that external memory breadth-firstsearch is among the hardest problems on planar graphs.KeywordsPlanar GraphExternal MemorySimple CycleOuterplanar GraphBoundary CycleThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Motivated by the increasing need for fast distributed processing of large-scale graphs such as the Web graph and various social networks, we study a number of fundamental graph problems in the message-passing model, where we have k machines that jointly perform computation on an arbitrary n-node (typically, n > k) input graph. The graph is assumed to be randomly partitioned among the k ≥ 2 machines (a common implementation in many real world systems). The communication is point-to-point, and the goal is to minimize the time complexity i.e., the number of communication rounds, of solving various fundamental graph problems.We present lower bounds that quantify the fundamental time limitations of distributively solving graph problems. We first show a lower bound of Ω(n/k) rounds for computing a spanning tree (ST) of the input graph. This result also implies the same bound for other fundamental problems such as computing a minimum spanning tree (MST), breadth-first tree (BFS), and shortest paths tree (SPT). We also show an Ω(n/k2) lower bound for connectivity ST verification and other related problems. Our lower bounds develop and use new bounds in random-partition communication complexity.To complement our lower bounds, we also give algorithms for various fundamental graph problems, e.g., PageRank, MST, connectivity, ST verification, shortest paths, cuts, spanners, covering problems, densest subgraph, subgraph isomorphism, finding triangles, etc. We show that problems such as PageRank, MST, connectivity, and graph covering can be solved in O(n/k) time (the notation O hides polylog(n) factors and an additive polylog(n) term); this shows that one can achieve almost linear (in k) speedup, whereas for shortest paths, we present algorithms that run in O(n/Ok) time (for (1 + e)-factor approximation) and in O(n/k) time (for O(log n)-factor approximation) respectively.Our results are a step towards understanding the complexity of distributively solving large-scale graph problems.

In the Steiner Tree problem one is given an undirected graph, a subset T of its vertices, and an integer k and the question is whether there is a connected subgraph of the given graph containing all the vertices of T and at most k other vertices. The vertices in the subset T are called terminals and the other vertices are called Steiner vertices. Recently, Pilipczuk et al. (55th IEEE Annual Symposium on Foundations of Computer Science, FOCS, 2014) gave a polynomial kernel for Steiner Tree in planar graphs and graphs of bounded genus, when parameterized by $$|T|+k$$ , the total number of vertices in the constructed subgraph. In this paper we present several polynomial time applicable reduction rules for Steiner Tree in graphs of bounded genus. In an instance reduced with respect to the presented reduction rules, the number of terminals |T| is at most cubic in the number of other vertices k in the subgraph. Hence, using and improving the result of Pilipczuk et al., we give a polynomial kernel for Steiner Tree in graphs of bounded genus for the parameterization by the number k of Steiner vertices in the solution. We give better bounds for Steiner Tree in planar graphs.

Thanks to their massive computational power and their SIMT computational model, Graphics Processing Units (GPUs) have been successfully used to accelerate a wide variety of regular applications (linear algebra, stencil computations, image processing and bioinformatics algorithms, among others). However, many established and emerging problems are based on irregular data structures, such as graphs. Examples can be drawn from different application domains: networking, social networking, machine learning, electrical circuit modeling, discrete event simulation, compilers, and computational sciences. It has been shown that irregular applications based on large graphs do exhibit runtime parallelism; moreover, the amount of available parallelism tends to increase with the size of the datasets. In this work, we explore an implementation space for deploying a variety of graph algorithms on GPUs. We show that the dynamic nature of the parallelism that can be extracted from graph algorithms makes it impossible to find an optimal solution. We propose a runtime system able to dynamically transition between different implementations with minimal overhead, and investigate heuristic decisions applicable across algorithms and datasets. Our evaluation is performed on two graph algorithms: breadth-first search and single-source shortest paths. We believe that our proposed mechanisms can be extended and applied to other graph algorithms that exhibit similar computational patterns.

The problem of processing shortest pa th queries in graphs arises in application areas such as intelligent t ranspor ta t ion system (ITS), geographic information system (GIS), and robotics. In this paper, we present an efficient algorithmic solution for processing shortest pa th queries in undirected planar graphs with non-negative edge weights. Previous algorithms for this problem on planar graphs all have a tradeoff between the query t ime and the da ta structure space used by the solutions. The previously best known trade-off on an n-vertex planar graph G in the worst case is O(v/-r) t ime for a pa th length query and O(v/-r + L) t ime for report ing an actual shortest path, with a da ta structure of O(n2/x/~) space, where r is an integer parameter with 1 < r _< n and L is the number of edges on the output shortest path. We present a new scheme, called frame search, that exploits the graph planari ty in a novel fashion. Wi th this scheme, we build an improved da ta structure of O(n + pv/-r + p2/r) space (in O(n + p2/v/~ + pr 3/4) time), where p is the minimum cardinality of a face-covering of the planar embedding of G with 1 < p < O(n), and r is an integer parameter with 1 < r < p. This da ta structure enables us to process each length query in O(v/-rlog r + a(n)) t ime and report an actual shortest path in an addit ional O(L) time, where a(n) is the inverse of Ackermann's function. In the worst case, our approach reduces the previously best query t ime by a factor of up to O(n 1/4) if the same amount of space is used. Our technique can also be applied to improve the previous query algorithms for some geometric shortest pa th problems on the plane.

Many emerging real-world applications require fast processing of large-scale data represented in the form of graphs. In this paper, we design a Field-Programmable Gate Array (FPGA) framework to accelerate graph algorithms based on the edge-centric paradigm. Our design is flexible for accelerating general graph algorithms with various vertex attributes and update propagation functions, such as Sparse Matrix Vector Multiplication (SpMV), PageRank (PR), Single Source Shortest Path (SSSP), and Weakly Connected Component (WCC). The target platform consists of large external memory to store the graph data and FPGA to accelerate the processing. By taking an edge-centric graph algorithm and hardware resource constraints as inputs, our framework can determine the optimal design parameters and produce an optimized Register-Transfer Level (RTL) FPGA accelerator design. To improve data locality and increase parallelism, we partition the input graph into non-overlapping partitions. This enables our framework to efficiently buffer vertex data in the on-chip memory of FPGA and exploit both inter-partition and intra-partition parallelism. Further, we propose an optimized data layout to improve external memory performance and reduce data communication between FPGA and external memory. Based on our design methodology, we accelerate two fundamental graph algorithms for performance evaluation: Sparse Matrix Vector Multiplication (SpMV) and PageRank (PR). Experimental results show that our accelerators sustain a high throughput of up to 2250 Million Traversed Edges Per Second (MTEPS) and 2487 MTEPS for SpMV and PR, respectively. Compared with several highly-optimized multi-core designs, our FPGA framework achieves up to 20.5× speedup for SpMV, and 17.7× speedup for PR, respectively; compared with two state-of-the-art FPGA frameworks, our designs demonstrate up to 5.3× and 1.8× throughput improvement for SpMV and PR, respectively.

We present the buffer heap , a cache-oblivious priority queue that supports Delete-Min , Delete , and a hybrid Insert / Decrease-Key operation in O (1/ B log 2 N / M ) amortized block transfers from main memory, where M and B are the (unknown) cache size and block size, respectively, and N is the number of elements in the queue. We introduce the notion of a slim data structure that captures the situation when only a limited portion of the cache, which we call a slim cache , is available to the data structure to retain data between data structural operations. We show that a buffer heap automatically adapts to such an environment and supports all operations in O (1/λ + 1/ B log 2 N /λ) amortized block transfers each when the size of the slim cache is λ. Our results provide substantial improvements over known trivial cache performance bounds for cache-oblivious priority queues with Decrease-Keys . Using the buffer heap, we present cache-oblivious implementations of Dijkstra’s algorithm for undirected and directed single-source shortest path (SSSP) problems for graphs with non-negative real edge-weights. On a graph with n vertices and m edges, our algorithm for the undirected case performs O ( n + m / B log 2 n / M ) block transfers and for the directed case performs O (( n + m / B ) ċ log 2 n / B ) block transfers. These results give the first non-trivial cache-oblivious bounds for the SSSP problem on general graphs. For the all-pairs shortest path (APSP) problem on weighted undirected graphs, we incorporate slim buffer heaps into multi-buffer-buffer-heaps and use these to improve the cache-aware cache complexity. We also present a simple cache-oblivious APSP algorithm for unweighted undirected graphs that performs O ( mn / B log M / B n / B ) block transfers. This matches the cache-aware bound and is a substantial improvement over the previous cache-oblivious bound for the problem.

Motivated by the increasing need for fast distributed processing of large-scale graphs such as the Web graph and various social networks, we study a message-passing distributed computing model for graph processing and present lower bounds and algorithms for several graph problems. This work is inspired by recent large-scale graph processing systems (e.g., Pregel and Giraph) which are designed based on the message-passing model of distributed computing. Our model consists of a point-to-point communication network of $k$ machines interconnected by bandwidth-restricted links. Communicating data between the machines is the costly operation (as opposed to local computation). The network is used to process an arbitrary $n$-node input graph (typically $n \gg k > 1$) that is randomly partitioned among the $k$ machines (a common implementation in many real world systems). Our goal is to study fundamental complexity bounds for solving graph problems in this model. We present techniques for obtaining lower bounds on the distributed time complexity. Our lower bounds develop and use new bounds in random-partition communication complexity. We first show a lower bound of $\Omega(n/k)$ rounds for computing a spanning tree (ST) of the input graph. This result also implies the same bound for other fundamental problems such as computing a minimum spanning tree (MST). We also show an $\Omega(n/k^2)$ lower bound for connectivity, ST verification and other related problems. We give algorithms for various fundamental graph problems in our model. We show that problems such as PageRank, MST, connectivity, and graph covering can be solved in $\tilde{O}(n/k)$ time, whereas for shortest paths, we present algorithms that run in $\tilde{O}(n/\sqrt{k})$ time (for $(1+\epsilon)$-factor approx.) and in $\tilde{O}(n/k)$ time (for $O(\log n)$-factor approx.) respectively.

The importance of high-performance graph processing to solve big data problems targeting high-impact applications is greater than ever before. Recent graph processing frameworks target different hardware platforms (e.g., shared memory systems, accelerators such as GPUs, and distributed systems) and differ with respect to the programming model they adopt (e.g., based on linear algebra formulations of graph algorithms or enabling direct access to the graph structure). To better understand the impact of these choices, this paper, presents a comparative study of five state-of-the-art graph processing frameworks: two CPU-only frameworks - GraphMat and Galois, two GPU-based frameworks - Nvgraph and Gunrock; and Totem, a hybrid (CPU+GPU) framework. We use three popular graph algorithms (PageRank, Single Source Shortest Path, and Breadth-First Search), and massive scale graphs with up to billions of edges. Our evaluation focuses on three performance metrics: (i) execution time, (ii) scalability and (iii) energy consumption.

FPGAs are promising platforms to efficiently execute distributed graph algorithms. Unfortunately, they are notoriously hard to program, especially when the problem size and system complexity increases. In this paper, we propose GraVF, a high-level design framework for distributed graph processing on FPGAs. It leverages the vertex-centric paradigm, which is naturally distributed and requires the user to define only very small kernels and their associated message semantics for the target application. The user design may subsequently be elaborated and compiled to the target system automatically by the framework. To demonstrate the flexibility and capabilities of the proposed framework, 4 graph algorithms with distinct requirements have been implemented, namely breadth-first search, PageRank, single source shortest path, and connected component. Results show that the proposed framework is capable of producing FPGA designs with performance comparable to similar custom designs while requiring only minimal input from the user.

Faster Shortest-Path Algorithms for Planar Graphs