Graph Systems and Algorithms
Programmable Unified Memory Architecture (PUMA)
Large scale graph analytics is essential to analyze relationships in big data sets. Thereto, the DARPA HIVE program targets a leap in power efficient graph analytics. In response to this program, Intel proposes the Programmable Unified Memory Architecture (PUMA). Based on graph workload analysis insights, PUMA consists of many multi-threaded cores, fine-grained memory and network accesses, a globally shared address space and powerful offload engines. In this talk, we will describe the PUMA architecture, both in terms of hardware and the software ecosystem. We will provide initial simulation based performance estimations, showing that for graph analysis applications, a PUMA node will outperform a conventional compute node by one to two orders of magnitude. Additionally, PUMA will continue to scale across multiple nodes, which is a challenge in conventional multinode setups.
Additional information
| Type | devroom |
|---|
More sessions
| 2/1/20 |
Python has proven to be a popular choice for data scientists in the domain of graph analytics. The multitude of freely available frameworks and python packages allow to develop applications quickly through ease of expressibility and reuse of code. With petabytes of data generated everyday and an ever evolving landscape of hardware solutions, we observe a graph processing framework should expose the following characteristics: ease of use, scalability, interoperability across data formats, and ...
|
| 2/1/20 |
Graffiti is the graph engine of Skydive - an open source networking analysis tool. Graffiti was created from scratch to provide the features required by Skydive : distributed, replicated, store the whole history of the graph, allow subcribing to events on the graph using WebSocket and visualization.
|
| 2/1/20 |
Graph algorithms play an increasingly important role in real-world applications. The Neo4j Graph Algorithms library contains a set of ~50 graph algorithms covering a lot of different problem domains. In our talk, we’ll present the architecture of the library and demonstrate the different execution phases using a real world example.
|
| 2/1/20 |
Gunrock is a CUDA library for graph-processing designed specifically for the GPU. It uses a high-level, bulk-synchronous, data-centric abstraction focused on operations on vertex or edge frontiers. Gunrock achieves a balance between performance and expressiveness by coupling high-performance GPU computing primitives and optimization strategies, particularly in the area of fine-grained load balancing, with a high-level programming model that allows programmers to quickly develop new graph ...
|
| 2/1/20 |
Graph databases and applications have attracted much attention in the past few years due to the efficiency with which they can represent big data, connecting different layers of data structures and allowing analysis while preserving contextual relationships. This has resulted in a fast-growing community that has been developing various database and algorithmic innovations in this area, many of which will be gathering together in this conference. We joined this field as computer architecture ...
|
| 2/1/20 |
In this talk we will introduce enhancements to the Cypher graph query language, enabling queries spanning multiple graphs, intended for use in sharding and federation scenarios. We will also present our experience with sharding the LDBC Social Network Benchmark dataset.
|
| 2/1/20 |
Temporal graphs capture the development of relationships within data throughout time. This model fits naturally within a streaming architecture, where new events can be inserted directly into the graph upon arrival from a data source, being compared to related entities or historical state. However, the vast majority of graph processing systems only consider traditional graph analysis on static data, with some outliers supporting batched updating and temporal analysis across graph snapshots. This ...
|
