Systems Research Group seminars

The Systems Research Group is re-starting their seminars series from 6th May 2022. Seminars will take place every two weeks at 1pm, on Fridays. From May to July the seminars will be online (SRG Teams), while from September onward we aim to move them to a hybrid format. More information on the schedule can be found on the seminars page of the Systems Research Group site.

Donald Robertson awarded Brendan Murphy Prize at MSN/Cosener’s 2019!

Each year in July, the (broadly-defined) computer networking community converges at Cosener’s House for the MSN workshop. The workshop is an informal gathering where attendees – students in particular – are encouraged to present on-going work and/or crazy ideas. From among the  presentations, the Brendan Murphy Award is given to the best student presentation, generally for work that has yet to be scrutinized or peer-reviewed.

Congratulations to Donald Robertson who, this year, has brought that honour to St Andrews as co-recipient of the award (alongside Naomi Arnold from QMUL).

http://coseners.net/history/brendan-murphy-prize/

(In the interest of transparency, Marwan Fayed was on the judging panel but recused himself during discussion of Donald’s presentation.)

SRG Seminar: “Large-Scale Hierarchical k-means for Heterogeneous Many-Core Supercomputers” by Teng Yu

We present a novel design and implementation of k-means clustering algorithm targeting supercomputers with heterogeneous many-core processors. This work introduces a multi-level parallel partition approach that not only partitions by dataflow and centroid, but also by dimension. Our multi-level ($nkd$) approach unlocks the potential of the hierarchical parallelism in the SW26010 heterogeneous many-core processor and the system architecture of the supercomputer.
Our design is able to process large-scale clustering problems with up to 196,608 dimensions and over 160,000 targeting centroids, while maintaining high performance and high scalability, significantly improving the capability of k-means over previous approaches. The evaluation shows our implementation achieves performance of less than 18 seconds per iteration for a large-scale clustering case with 196,608 data dimensions and 2,000 centroids by applying 4,096 nodes (1,064,496 cores) in parallel, making k-means a more feasible solution for complex scenarios.
This work is to be presented in the International Conference for High Performance Computing, Networking, Storage, and Analysis (SC18).

Event details

  • When: 1st November 2018 13:00 - 14:00
  • Where: Cole 1.33b
  • Series: Systems Seminars Series
  • Format: Seminar, Talk

SRG Seminar: “Using Metric Space Indexing for Complete and Efficient Record Linkage” by Özgür Akgün

Record linkage is the process of identifying records that refer to the same real-world entities, in situations where entity identifiers are unavailable. Records are linked on the basis of similarity between common attributes, with every pair being classified as a link or non-link depending on their degree of similarity. Record linkage is usually performed in a three-step process: first groups of similar candidate records are identified using indexing, pairs within the same group are then compared in more detail, and finally classified. Even state-of-the-art indexing techniques, such as Locality Sensitive Hashing, have potential drawbacks. They may fail to group together some true matching records with high similarity. Conversely, they may group records with low similarity, leading to high computational overhead. We propose using metric space indexing to perform complete record linkage, which results in a parameter-free record linkage process combining indexing, comparison and classification into a single step delivering complete and efficient record linkage. Our experimental evaluation on real-world datasets from several domains shows that linkage using metric space indexing can yield better quality than current indexing techniques, with similar execution cost, without the need for domain knowledge or trial and error to configure the process.

Event details

  • When: 18th October 2018 13:00 - 14:00
  • Where: Cole 1.33b
  • Series: Systems Seminars Series
  • Format: Seminar

SRG Seminar: “Efficient Cross-architecture Hardware Virtualisation” by Tom Spink

Virtualisation is a powerful tool used for the isolation, partitioning, and sharing of physical computing resources. Employed heavily in data centres, becoming increasingly popular in industrial settings, and used by home-users for running alternative operating systems, hardware virtualisation has seen a lot of attention from hardware and software developers over the last ten?fifteen years.

From the hardware side, this takes the form of so-called hardware assisted virtualisation, and appears in technologies such as Intel-VT, AMD-V and ARM Virtualization Extensions. However, most forms of hardware virtualisation are typically same-architecture virtualisation, where virtual versions of the host physical machine are created, providing very fast isolated instances of the physical machine, in which entire operating systems can be booted. But, there is a distinct lack of hardware support for cross-architecture virtualisation, where the guest machine architecture is different to the host.

I will talk about my research in this area, and describe the cross-architecture virtualisation hypervisor Captive that can boot unmodified guest operating systems, compiled for one architecture in the virtual machine of another.

I will talk about the challenges of full system simulation (such as memory, instruction, and device emulation), our approaches to this, and how we can efficiently map guest behaviour to host behaviour.

Finally, I will discuss our plans for open-sourcing the hypervisor, the work we are currently doing and what future work we have planned.

Event details

  • When: 11th October 2018 13:00 - 14:00
  • Where: Cole 1.33b
  • Series: Systems Seminars Series
  • Format: Seminar, Talk

Dasip 2018 Keynote: Professor Simon Dobson

Head of School Simon Dobson will deliver a keynote at Dasip, the Conference on Design and Architectures for Signal and Image Processing in October in Porto. Dasip provides an international forum for innovation and developments in the field of embedded signal processing systems. Simon’s keynote will focus on making the transition from sensors to sensor systems software.

Abstract: Signal processing underpins everything we do with sensors. The physical limits of sensors, and the effects of their exposure to their environment, in turn constrain their accuracy, and therefore affect the trust we can place in sensor-driven systems. But this is a long pipeline, and it’s by no means clear how to trace from low-level errors and inaccuracies to their high-level consequences. In this talk I will try to tease-out some of the desiderata we might look for in such a pipeline, with a view to understanding how we can go about building sensor systems that deserve our trust.

An Academic’s Observations from a Sabbatical at Google

Professor Adam Barker is featured in this month’s Communications of the ACM Magazine (CACM) discussing his recent Visiting Faculty appointment at Google. The Viewpoints article summarises his experiences working in software engineering on the Borgmaster team, and some of the core lessons which can be brought back to academia.

Borg is Google’s cluster management framework, which runs hundreds of thousands of jobs, across a number of clusters each with up to tens of thousands of machines.

SRG Seminar: “Application of Bayesian Nonparametric in household human activity recognition” by Lei Fang

Abstract

In this talk, I will talk about the possibility of using Bayesian nonparametric clustering, or Dirichlet Process Mixture model to solve human activity recognition problem. In particular, I will discuss how the technique can be useful when the activity labels are not annotated and/or the activity evolves over the time. This initial study is built on an existing work on using directional statistical models (von Mises-Fisher) distribution, called Hierarchical Mixture of Conditional Independent von Mises Fisher distribution (HMCIvMFs), for unknown events detection and learning. Markov chain Monte Carlo sampling based learning algorithm will be presented together with some initial experiment results.

Event details

  • When: 12th April 2018 13:00 - 14:00
  • Where: Cole 1.33b
  • Series: Systems Seminars Series
  • Format: Seminar

SRG Seminar: “Introduction to Apache Mesos and the DataCenter Operating System” by Matt Jarvis

Abstract
Data processing paradigms are undergoing a paradigm shift as we move more and more towards real time processing. Emerging software models such as the SMACK stack are at the forefront of this change, focused on a pipeline processing model, but are also introducing new levels of operational complexity in running multiple complex distributed systems such as Spark, Kafka and Cassandra. In this talk, I’ll introduce both Apache Mesos and DC/OS as a solution to this growing problem, and describe the benefits are of running these new kinds of systems for emerging cloud native workloads.
 
Bio
Matt Jarvis is Senior Director of Community and Evangelism at Mesosphere, engaging with the communities around DC/OS and Mesos. Matt has spent more than 15 years building products and services around open source software, on everything from embedded devices to large scale distributed systems. Most recently he has been focused on the open cloud infrastructure space, and in emerging patterns for cloud native applications. 

Event details

  • When: 24th April 2018 13:00 - 14:00
  • Where: Cole 1.33b
  • Series: Systems Seminars Series
  • Format: Seminar

SRG Seminar: “On Engineering Unikernels” by Ward Jaradat

We have explored data coordination techniques that permit distributed systems to be constructed by interconnecting services. In such systems the network latency is often a problem. For example, large data volumes might have to be transmitted across the network if computation cannot be co-located close to data sources. One solution to this problem is the ability to deploy services in appropriate geographical locations and compose them together to create distributed ecosystems. Hence we seek to be able to deploy such services rapidly and dynamically enact and orchestrate them. However, this goal is hindered by the size of the deployments. Currently, virtual machine appliances that host such services on top of monolithic kernels are very large, thus are potentially slow to deploy as they may need to be transmitted across a network.

Our principles led us to take the route of re-engineering the standard software stack to create self-contained applications that are less-bloated and consequently much smaller based on Unikernels. Unikernels are compact library operating systems that enable a single application to be statically linked against a simple kernel that manages the underlying resources presented by a hypervisor. In this talk I will present Stardust – a specialised Unikernel that aims to support the deployment of application services based on the Java programming language.

Event details

  • When: 15th March 2018 13:00 - 14:00
  • Where: Cole 1.33b
  • Series: Systems Seminars Series
  • Format: Seminar