We show that certain Graph Laplacian linear sets of equations exhibit optimal accuracy, guaranteeing that the relative error is no larger than the norm of the relative residual and that optimality occurs for carefully chosen right-hand sides. Such sets of equations arise in PageRank and Markov chain theory. We establish new relationships among the PageRank teleportation parameter, the Markov chain discount, and approximations to linear sets of equations. The set of optimally accurate systems can be separated into two groups for an undirected graph – those that achieve optimality asymptotically with the graph size and those that do not – determined by the angle between the right-hand side of the linear system and the vector of all ones. We provide supporting numerical experiments.

Triangle counting is a fundamental building block in graph algorithms. In this article, we propose a block-based triangle counting algorithm to reduce data movement during both sequential and parallel execution. Our block-based formulation makes the algorithm naturally suitable for heterogeneous architectures. The problem of partitioning the adjacency matrix of a graph is well-studied. Our task decomposition goes one step further: it partitions the set of triangles in the graph. By streaming these small tasks to compute resources, we can solve problems that do not fit on a device. We demonstrate the effectiveness of our approach by providing an implementation on a compute node with multiple sockets, cores and GPUs. The current state-of-the-art in triangle enumeration processes the Friendster graph in 2.1 seconds, not including data copy time between CPU and GPU. Using that metric, our approach is 20 percent faster. When copy times are included, our algorithm takes 3.2 seconds. This is 5.6 times faster than the fastest published CPU-only time.

Designing flexible graph kernels that can run well on various platforms is a crucial research problem due to the frequent usage of graphs for modeling data and recent architectural advances and variety. In this work, we propose a novel graph processing framework, PGAbB (Parallel Graph Algorithms by Blocks), for modern shared-memory heterogeneous platforms. Our framework implements a block-based programming model. This allows a user to express a graph algorithm using kernels that operate on subgraphs. PGAbB support graph computations that fit in host DRAM but not in GPU device memory, and provides simple but effective scheduling techniques to schedule computations to all available resources in a heterogeneous architecture. We have demonstrated that one can easily implement a diverse set of graph algorithms in our framework by developing five algorithms. Our experimental results show that PGAbB implementations achieve better or competitive performance compared to hand-optimized implementations. Based on our experiments on five graph algorithms and forty-four graphs, in the median, PGAbB achieves 1.6, 1.6, 5.7, 3.4, 4.5, and 2.4 times better performance than GAPBS, Galois, Ligra, LAGraph Galois-GPU, and Gunrock graph processing systems, respectively.

Some past and future supercomputer nodes incorporate High- Bandwidth Memory (HBM). Compared to standard DRAM, HBM has similar latency, higher bandwidth and lower capacity. In this paper, we evaluate algorithms for managing High- Bandwidth Memory automatically. Previous work suggests that, in the worst case, performance is extremely sensitive to the policy for managing the channel to DRAM. Prior theory shows that a priority-based scheme (where there is a static strict priority-order among p threads for channel access) is O(1)-competitive, but FIFO is not, and in the worst case is Ω(p) competitive. Following this theoretical guidance would be a disruptive change for vendors, who currently use FIFO variants in their DRAMcontroller hardware. Our goal is to determine theoretically and empirically whether we can justify recommending investment in priority-based DRAM controller hardware. In order to experiment with DRAM channel protocols, we chose a theoretical model, validated it against real hardware, and implemented a basic simulator. We corroborated the previous theoretical results for the model, conducted a parameter sweep while running our simulator on address traces from memory bandwidth-bound codes (GNU sort and TACO sparse matrix-vector product), and designed better channel-access algorithms.

Some past and future supercomputer nodes incorporate High- Bandwidth Memory (HBM). Compared to standard DRAM, HBM has similar latency, higher bandwidth and lower capacity.

Given an input stream S of size N , a ɸ-heavy hitter is an item that occurs at least ɸN times in S . The problem of finding heavy-hitters is extensively studied in the database literature. We study a real-time heavy-hitters variant in which an element must be reported shortly after we see its T = ɸ N-th occurrence (and hence it becomes a heavy hitter). We call this the Timely Event Detection ( TED ) Problem. The TED problem models the needs of many real-world monitoring systems, which demand accurate (i.e., no false negatives) and timely reporting of all events from large, high-speed streams with a low reporting threshold (high sensitivity). Like the classic heavy-hitters problem, solving the TED problem without false-positives requires large space (Ω (N) words). Thus in-RAM heavy-hitters algorithms typically sacrifice accuracy (i.e., allow false positives), sensitivity, or timeliness (i.e., use multiple passes). We show how to adapt heavy-hitters algorithms to external memory to solve the TED problem on large high-speed streams while guaranteeing accuracy, sensitivity, and timeliness. Our data structures are limited only by I/O-bandwidth (not latency) and support a tunable tradeoff between reporting delay and I/O overhead. With a small bounded reporting delay, our algorithms incur only a logarithmic I/O overhead. We implement and validate our data structures empirically using the Firehose streaming benchmark. Multi-threaded versions of our structures can scale to process 11M observations per second before becoming CPU bound. In comparison, a naive adaptation of the standard heavy-hitters algorithm to external memory would be limited by the storage device’s random I/O throughput, i.e., ≈100K observations per second.

Both the data science and scientific computing communities are embracing GPU acceleration for their most demanding workloads. For scientific computing applications, the massive volume of code and diversity of hardware platforms at supercomputing centers has motivated a strong effort toward performance portability. This property of a program, denoting its ability to perform well on multiple architectures and varied datasets, is heavily dependent on the choice of parallel programming model and which features of the programming model are used. In this paper, we evaluate performance portability in the context of a data science workload in contrast to a scientific computing workload, evaluating the same sparse matrix kernel on both. Among our implementations of the kernel in different performance-portable programming models, we find that many struggle to consistently achieve performance improvements using the GPU compared to simple one-line OpenMP parallelization on high-end multicore CPUs. We show one that does, and its performance approaches and sometimes even matches that of vendor-provided GPU math libraries.

In this project we developed and validated algorithms for privacy-preserving linear regression using a new variant of Secure Multiparty Computation (MPC) we call ?Hybrid MPC? (hMPC). Our variant is intended to support low-power, unreliable networks of sensors with low-communication, fault-tolerant algorithms. In hMPC we do not share training data, even via secret sharing. Thus, agents are responsible for protecting their own local data. Only the machine learning (ML) model is protected with information-theoretic security guarantees against honest-but-curious agents. There are three primary advantages to this approach: (1) after setup, hMPC supports a communication-ef?cient matrix multiplication primitive, (2) organizations prevented by policy or technology from sharing any of their data can participate as agents in hMPC, and (3) large numbers of low-power agents can participate in hMPC. We have also created an open-source software library named ?Cicada? to support hMPC applications with fault-tolerance. The fault-tolerance is important in our applications because the agents are vulnerable to failure or capture. We have demonstrated this capability at Sandia?s Autonomy New Mexico laboratory through a simple machine-learning exercise with Raspberry Pi devices capturing and classifying images while ?ying on four drones. 1 Sandia National Laboratories 2 University of New Mexico

Motivated by the properties of unending real-world cybersecurity streams, we present a new graph streaming model: XStream. We maintain a streaming graph and its connected components at single-edge granularity. In cybersecurity graph applications, input streams typically consist of edge insertions; individual deletions are not explicit. Analysts maintain as much history as possible and will trigger customized bulk deletions when necessary Despite a variety of dynamic graph processing systems and some canonical literature on theoretical sliding-window graph streaming, XStream is the first model explicitly designed to accommodate this usage model. Users can provide Boolean predicates to define bulk deletions. Edge arrivals are expected to occur continuously and must always be handled. XStream is implemented via a ring of finite-memory processors. We give algorithms to maintain connected components on the input stream, answer queries about connectivity, and to perform bulk deletion. The system requires bandwidth for internal messages that is some constant factor greater than the stream arrival rate. We prove a relationship among four quantities: the proportion of query downtime allowed, the proportion of edges that survive an aging event, the proportion of duplicated edges, and the bandwidth expansion factor. In addition to presenting the theory behind XStream, we present computational results for a single-threaded prototype implementation. Stream ingestion rates are bounded by computer architecture. We determine this bound for XStream inter-process message-passing rates in Intel TBB applications on Intel Sky Lake processors: between one and five million graph edges per second. Our single-threaded prototype runs our full protocols through multiple aging events at between one half and one a million edges per second, and we give ideas for speeding this up by orders of magnitude.

wholly owned subsidiary of Honeywell International Inc., for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-NA0003525.

We introduce HITMIX, a new technique for network seed-set expansion, i.e., the problem of identifying a set of graph vertices related to a given seed-set of vertices. We use the moments of the graph's hitting-time distribution to quantify the relationship of each non-seed vertex to the seed-set. This involves a deterministic calculation for the hitting-time moments that is scalable in the number of graph edges and so avoids directly sampling a Markov chain over the graph. The moments are used to fit a mixture model to estimate the probability that each non-seed vertex should be grouped with the seed set. This membership probability enables us to sort the non-seeds and threshold in a statistically-justified way. To the best of our knowledge, HITMIX is the first full statistical model for seed-set expansion that can give vertex-level membership probabilities. While HITMIX is a global method, its linear computation complexity in practice enables computations on large graphs. We have a high-performance implementation, and we present computational results on stochastic blockmodels and a small-world network from the SNAP repository. The state of the art in this problem is a collection of recently developed local methods, and we show that distinct advantages in solution quality are available if our global method can be used. In practice, we expect to be able to run HITMIX if the graph can be stored in memory.

This paper develops an algorithmic foundation for automated management of the multilevel-memory systems common to new supercomputers. In particular, the High-Bandwidth Memory (HBM) of these systems has a similar latency to that of DRAM and a smaller capacity, but it has much larger bandwidth. Systems equipped with HBM do not fit in classic memory-hierarchy models due to HBM's atypical characteristics. Unlike caches, which are generally managed automatically by the hardware, programmers of some current HBM-equipped supercomputers can choose to explicitly manage HBM themselves. This process is problem specific and resource intensive. Vendors offer this option because there is no consensus on how to automatically manage HBM to guarantee good performance, or whether this is even possible. In this paper, we give theoretical support for automatic HBM management by developing simple algorithms that can automatically control HBM and deliver good performance on multicore systems. HBM management is starkly different from traditional caching both in terms of optimization objectives and algorithm development. Since DRAM and HBM have similar latencies, minimizing HBM misses (provably) turns out not to be the right memory-management objective. Instead, we directly focus on minimizing makespan. In addition, while cache-management algorithms must focus on what pages to keep in cache; HBM management requires answering two questions: (1) which pages to keep in HBM and (2) how to use the limited bandwidth from HBM to DRAM. It turns out that the natural approach of using LRU for the first question and FCFS (First-Come-First-Serve) for the second question is provably bad. Instead, we provide a priority based approach that is simple, efficiently implementable and $O(1)$-competitive for makespan when all multicore threads are independent.

A previously obscure area of cryptography known as Secure Multiparty Computation (MPC) is enjoying increased attention in the field of privacy-preserving machine learning (ML), because ML models implemented using MPC can be uniquely resistant to capture or reverse engineering by an adversary. In particular, an adversary who captures a share of a distributed MPC model provably cannot recover the model itself, nor data evaluated by the model, even by observing the model in operation. We report on our small project to survey current MPC software and judge its practicality for fielding mission-relevant distributed machine learning models.