View Proposal

---

Proposer

Rob Stewart

Title

Special purpose hardware for RDF stream processing

Goal

Develop a hardware design to support high performance RDF stream processing and inference

Description

Stream processing is about doing one-pass execution of continuous queries over a potentially infinite stream of values. Streams come with different characteristics, e.g. data rate from low (one or two values an hour) to high (thousands of values a second), and is published either in steady stream or in highly irregular bursts. Many use cases require joining values across streams and stored sources, as well as computing aggregation functions. To support these operations over potentially infinite streams, windowing operators are used to provide a scope for the operation. Additionally with RDF streams, there is also the possibility of performing inferencing over the stream of values, i.e. generating new data based on the content of the stream. All these needs pose interesting challenges for stream processing: how much data needs to be windowed to fire inference rules? And crucially, what processing hardware should be used to support throughput of thousands of RDF data items a second? Field Programmable Gate Arrays (FPGAs) are programmable hardware chips, which when configured have a circuit that is precisely design to meet an algorithmic need. For streaming domains they offer extremely high throughput, and use very little energy. This project will investigate the use of FPGAs for developing special purpose a RDF stream processor. A key hardware design decision will be to decide how "programmable" the hardware is at runtime. In other words should the hardware design allow runtime FPGA programming to (1) switch window operators, (2) change the window size, (3) upload new inference rules. A desirable artefact from this project will be an open source hardware design for a RDF stream processing hardware accelerator.

Resources

Background

Url

Difficulty Level

High

Ethical Approval

None

Number Of Students

1

Supervisor

Rob Stewart

Keywords

Degrees