Stuart McRobert

Computing Manager

Department of Statistics

University of Oxford

Get Started

Picture of Stuart McRobert

Overview

Stuart spent many happy years’ at Imperial College in the Computing Support Group (CSG) for the Department of Computing before accepting a position at the University of Oxford in 2009
Experienced computing, team and project manager
With more than 40 years' in-depth IT expertise
Proficient technical staff and project management
Accomplished delivery of leading edge computing solutions at world-class universities
Much of what follows relates to Stuart's work now at the University of Oxford

Learn More

Network Switch

Summary

Supporting

University teaching, research and administrative computing requirements

further details

HPC

CPU server procurement, configuration, deployment and support of Linux compute servers

further details

GPU

Servers usually configured with multiple NVIDIA GPUs with CUDA and associated software

further details

Slurm

Installation and configuration of the free open-source Slurm Workload Manager (job scheduler)

further details

System Administration

Chiefly using Linux operating system (currently Fedora, Debian and Ubuntu distributions)

further details

Networking

High speed Ethernet switches with links at up to 100 Gbps, routers, firewalls, and extensive Wi-Fi

further details

Data Centre

Chair of University Data Centre User Group. Review operation, deployment and refurbishment projects

further details

Speaker

Invited speaker / tutor at computing and networking conferences worldwide

further details

Data Storage

Provision and deployment of both local storage and network file servers plus online backup solutions

further details

Lots more

Including web, audio visual solutions, virtualization, account management, email, database, and cloud computing...

further details

Previous Roles

In and managing the Computing Support Group (CSG), Department of Computing, Imperial College London for over 20 years'

further details

Contact Details

sm ... stats.ox.ac.uk
+44 1865 272 862
Department of Statistics, University of Oxford, 24-29 St Giles', Oxford, UK, OX1 3LB

Supporting
HPC Servers: HPC server procurement, configuration, deployment and support

Extensively involved in server specification, selection, procurement, data centre deployment, configuration and routine system administration work. Servers are chiefly dual processor with plenty of memory and fast local RAID storage available to assist in achieving good overall computational performance.
GPU Servers: GPU servers are usually configured with multiple NVIDIA GPUs with CUDA and associated software

One of our latest GPU server is a Dell EMC DSS 8440 complete with ten NVIDIA Quadro RTX 6000 GPUs and a combination of NVMe and SSD for very fast local storage

A significant amount of machine learning work is currently performed on Supermicro servers each with eight Geforce GTX 1080 or 1080 Ti or 2080 or 2080 Ti GPUs. Newer platforms are now also being explored

One of the first dedicated GPU servers was a Dell PowerEdge C4130 with four Tesla K80 dual Kepler GK210 GPUs which was then upgraded to a Dell C4140 with four Tesla V100-PCIE-32GB GPUs. Prior to that a number of early passive GPU cards were simply installed into existing servers, perhaps with a power supply upgrade
Slurm: Installation and configuration of the Slurm Workload Manager.

Slurm is an open source, fault-tolerant, and highly scalable cluster management and job scheduling system for large and small Linux clusters. Now used by default for both CPU and GPU servers to help ensure good server resource utilisation and convenient job scheduling, whilst avoiding overloading systems.
System Administration: Extensive system administration and development experience

Chiefly with Linux systems (currently Fedora, Debian and Ubuntu distributions) with regular patch and release updates across desktops and servers to help ensure a consistent up-to-date research environment

Software patch management along with deployment of release updates

Virtualization chiefly using KVM and QEMU, also previously Virtualbox and Solaris containers with zones

Implementation of storage backup solutions both to tape and now mostly online

Use and development of automated management tools to consistently manage systems and monitor availability

Provision of Kerberos and LDAP for account management using the FreeIPA project, although already familiar with Kerberos since the 1980s as part of MIT Project Athena

Email system configuration with Exim, Dovecot and SquirrelMail

Website installation and maintenance, including sites using WordPress

Installation and operation of replicated databases using both MariaDB (MySQL) and PostgreSQL

Previously extensively deployed BSD, SunOS / Solaris and AT&T Bell Labs Unix releases, plus some Apple OS X and Microsoft Windows
Networking: High speed Ethernet switches with links up to 100 Gbps, routers, firewalls, and extensive Wi-Fi deployment

Comprehensive LAN and campus network deployment experience and configuration expertise with both wired and wireless networking

Fortinet firewall high availability configuration supporting

Border Gateway Protocol (BGP) and IP routing

Access policies

NAT, DHCP, DNS, and VPN services

The search in late 2020 for Multigigabit edge switches has seen the initial selection of the Dell Networking N2248PX-ON PowerSwitch offering RJ45 ports with speeds up to 2.5 Gbps along with PoE providing 30W 802.3at (PoE+ or Type 2) and up to 60W 802.3bt (PoE++ or Type-3) on some ports to power even more demanding devices, and 25 GbE uplinks to help ensure plenty of bandwidth to and from the edge

In summer 2018 introduced the next generation of network switching chiefly to support new servers and now edge switching too

Dell EMC Networking (Force10) S5048F-ON switches each providing 48 x 25 GbE and 6 x 100 GbE ports

Twin 100 GbE switch-to-switch local connectivity

Mellanox Technologies (now part of NVIDIA in 2019) dual port 25 GbE network adaptors for server links

Dell EMC Virtual Link Trunking (VLT) between local switch pairs

Link Aggregation Control Protocol (LACP) used for switch-to-switch and switch-to-server links offering both resiliency and increased bandwidth

Wi-Fi network configuration currently using Ruckus Wireless (part of CommScope in 2019) technology

SmartZone controllers supporting numerous dual-band technology access points e.g. ZoneFlex R710 Wave 2 802.11ac (4x4:4 stream MU-MIMO) and ZoneFlex R500 802.11ac (2x2:2 stream MIMO)

Experience with earlier Ruckus ZoneDirector controllers and access points, and prior to that extensive deployment of Lucent and Proxim Wi-Fi solutions from 1999 onwards

Extreme Networks switch configuration and management

Multi-Chassis Link Aggregation Group (MLAG / MC-LAG)

VLANs

MAC based NetLogin with RADIUS support

RIPE Atlas probes

"RIPE Atlas employs a global network of probes that measure Internet connectivity and reachability, providing an unprecedented understanding of the state of the Internet in real time."
Data Centre: Data Centre – Chair of University Data Centre User Group

The University has data centres on campus as well as capacity in the off-site state-of-the-art Jisc shared data centre hosted by Virtus in Slough

The department has several racks of computational equipment located in the data centre with dedicated dark fibre connectivity back to our building in St Giles' where the network room links everything together
Data Storage: Provision and deployment of file servers and online backup solutions

Most compute servers have traditional local RAID storage with some of the latest GPU servers have NVMe and/or SSD storage, especially for "hot" data being used by GPUs with machine learning datasets

Software and hardware RAID using where appropriate using one of level 0 (stripe), 1 (mirror), 5, 6, 50 or 60

NFS and Samba

File systems including ext4 and ZFS, with previously ext2, ext3, ReiserFS, UFS, Veritas, and XFS were all extensively used

Volume managers e.g. Linux Volume Manager LVM
Implementation of storage backup solutions both to tape and now mostly online with snapshots with the next generation based on ZFS storage using a Dell ME844 JBOD
Lots more...: Area Summary

Audio Visual Systems For lecture theatres, IT Suite and meeting rooms

Cloud Computing Experience managing and using Cloud resources

Finance Budget management and order generation

Cooling projects Both for computer server rooms and meeting rooms

Building projects Chiefly determination of requirements and overseeing project delivery. Projects have included building structured wiring and additional power installation, improving lecture threate and teaching room fresh air flow with cooling, and moving a library

Security systems Both for IT e.g. Multi-Factor Authentication (MFA), and building CCTV
Previously: Various roles in the Computing Support Group (CSG), Department of Computing, Imperial College London

Project Manager
October 2006 – September 2009
Project manager delivering innovative IT solutions enabling world-class research

Manager, Computing Support Group
1999 – 2006
Manager of the Computing Support Group with a dozen full-time staff primarily covering networking and extensive systems development work, user support help desk and technical library, along with numerous software and hardware projects. Also encouraged several enthusiastic part-time student helpers to focus on, and develop very interesting IT projects.
Management of a data centre with around 270 kW of cooling capacity hosting some 80 racks of compute servers, storage and networking (1 and 10 Gb Ethernet plus InfiniBand) supporting academic computer science research, teaching and administrative computing. High performance computing clusters (including GPUs) and large Sun shared memory systems. Extensive copper and fibre connected networking supporting some 500 dual-boot desktops across the department, with locations across campus.
Project management, purchasing and budget control, strategic direction and forward planning vision.

Head of Systems, Computing Support Group
1986 – 1999
Providing world-class IT solutions for research and teaching requirements across a wide range of systems from an IBM mainframe to hundreds of desktops. Platforms supported included Apple, Fujitsu, GEC (UK), Gould (Encore), HP, IBM, ICL, Onyx, PDP, Sequent, SGI, Sun, VAX, and numerous PCs.
Widespread deployment of Sun Microsystems desktops and servers, ranging from some of the earliest Sun 2 systems through to large clusters of SPARC CPU servers, supporting both compute (HPC) and file serving.
Extensive Unix systems experience from kernel level upwards (Version 7, System III and V, BSD 2.x and 4.x, SunOS and Solaris, HP-UX, and Linux).
Development with a colleague of one of the largest public FTP archives available on the early Internet (containing freely distributable software and information) which became SunSITE Northern Europe.
Speaker: Invited speaker at computing and networking conferences worldwide including USENIX LISA, SANS, UKUUG FLOSSUK, and AUUG

A combination of tutorials, lectures, presentations, round tables, seminars and works-in-progress spanning around 30 years'

Moving Department − IT secrets to achieving a smooth migration
March 2017 • UKUUG FLOSS
UKUUG FLOSS Spring Conference, Manchester, England
Presentation covering the successful relocation of all servers and core networking to a new data centre with minimal downtime, followed by the migration of the department to its new home.

Moving Home − Replacing Sun Solaris with Debian Linux file servers
March 2017 • UKUUG FLOSS
UKUUG FLOSS Spring Conference, Manchester, England
Presentation about replacing an old Sun SPARC Solaris home directory file server with three Dell Intel servers running Debian Linux and using LVM2, ext4fs, and NFS.

Apple Pie − A New Recipe
November 2003 • UKUUG
Mac OS X Technology briefing, Apple UK, Stockley Park, England.
Presentation about the Apple G5 and the installation of a G5 laboratory for teaching high-performance computing.

A TOE in the Water
February 2003 • UKUUG
UKUUG LISA Conference, London, England.
A presentation chiefly around TCP Offload Engines (TOE) and the challenges encountered matching network and server performance. Also covered aspects of iSCSI and Security Offload Engines (SOE).

Wireless Ethernet
October 2002 • Imperial College
Wireless Seminar, ICTAP, Imperial College, London, England.
A presentation covering wireless Ethernet standards (802.11abg) and experience implementing existing 2.4 GHz and early 5 GHz products. Wireless security issues and 802.1x authentication.

Wire Speed Networks
March 2002 • Extreme Networks
Extreme Networks meeting, The RSA, London, England.
Presentation covering the successful deployment of reliable high performance Gigabit Ethernet networking in a demanding university computing environment. Also covered near future 10 Gigabit Ethernet developments.

Faster and Faster Networks
February 2002 • UKUUG
UKUUG LISA Winter Conference, London, England.
Conference presentation covering "improvements in hardware and networking technology, particularly in the field of Gigabit and 10 Gigabit Ethernet and wireless networking".

Users' Round Table Presentation
April 2001 • Gigabit Ethernet Conference
GEC 2001, San Jose, CA, USA.
Presentation covering the successful introduction and deployment of Gigabit Ethernet in a University environment.

Wireless Ethernet @ 11 Mbps
February 2001 • UKUUG
UKUUG Winter Conference, Newcastle, England.

Users' Round Table Position Statement
March 2000 • Gigabit Ethernet Conference
GEC 2000, San Jose, CA, USA.
Position statement covering the successful early adoption of Gigabit Ethernet in a University environment.

Migration to Gigabit Ethernet, How? and Why?
December 1999 • UKUUG
UKUUG Winter Conference, Internet Technologies, Cambridge University, Cambridge, England.

Experiences with Gigabit Ethernet Networks
May 1999 • SANS
SANS99, The Eighth Annual Conference on System Administration, Networking and Security, Baltimore, USA.

Faster and Faster − Gigabit Ethernet Networks, File Servers, and Users
December 1998 • Usenix
April 1999 • Usenix
Tutorial first presented at LISA'98 (December 1998), 12th Systems Administration Conference, Usenix, Boston, MA, USA.
Tutorial also presented at the USENIX Networking '99 conference, Santa Clara, CA, USA.

Veritas and Sun SITE Northern Europe
October 1998 • Veritas
Veritas Vericon.98 conference, Las Vegas, Nevada, USA.
Presentation about successfully utilizing the Veritas file system and volume manager with Sun SITE Northern Europe.

Rector's Award
October 1998 • Imperial College
Rector's Award for Outstanding Work, Imperial College, London, England.
Presentation and Rector's Award for "exceptional practical skills" with the early deployment of a multi gigabit Ethernet superhighway in the Department of Computing.

Veritas and Sun SITE Northern Europe
May 1998 • Veritas UK
Veritas UK User Group meeting, Egham, England.
Presentation about successfully utilizing the Veritas file system and volume manager with Sun SITE Northern Europe.

Larger and Larger File Systems − From 200 MB to 200 GB
April 1998 • UKUUG
UKUUG LISA 98, London, England.
Technical paper and presentation about ever larger file systems.

High Performance Sun SITE Northern Europe
June 1997 • Sun User Group, USA
Sun User Group Conference, Boston, MA, USA.
Technical conference paper and presentation about the Sun Enterprise 6000. Topics included hardware, system configuration, disk layout, Veritas file and volume management software.
The Archive and Sun SITE Northern Europe were projects undertaken with colleagues at Imperial College, especially Lee McLoughlin, and Sun Microsystems with assistance from Veritas and other interested parties.

Experiences of Running a Large Archive Site
October 1996 • Usenix
Invited Talk, Usenix, 10th Systems Administration Conference LISA 1996, Chicago, Illinois, USA.
Invited talk taking a fascinating look behind the scenes of, at the time, one of the Internet's richest and most popular free access archive sites, sunsite.doc.ic.ac.uk. This busy site was then powered by an 8-way Gbyte SS1000 with some 70+GB of trans (logged) RAID5 disk space, Ethernet and FDDI networking. Efforts to improve network and server performance were also discussed.

Sun SITE Northern Europe − The Archive Continues...
September 1996 • AUUG
AUUG Technical Conference, Melbourne, Australia.
Technical paper and presentation about the development and expansion of the Internet archive site.

Sun SITE Northern Europe − The Archive Continues...
September 1996 • Australian National University
Centre for Networked Information and Publishing, Australian National University, Canberra, Australia.
Seminar looking behind the scenes at many of the disk and file system issues involved in supporting and attempting to expand the size of a large and active archive site − Sun SITE Northern Europe.

From Twisting Country Lanes to MultiLane Ethernet Super Highways
September 1995 • Usenix
Ninth Systems Administration Conference (LISA '95), Monterey, CA, USA.
Technical paper and presentation explaining the utilization of Sun servers each with multiple Ethernet connections to deliver a much higher performance teaching laboratory network.

Sun SITE Northern Europe
September 1994 • Sun Microsystems
Presentation at the Sun Education and Research Conference, San Francisco, CA, USA.

The Archive Strikes Back
September 1993 • UKUUG
UKUUG LISA 93, Coping with Change, London, England.
Presentation about The Archive from the start in 1989, followed by the preannouncement of the donation of hardware for a new archive, Sun SITE Northern Europe.

The Archive − src.doc.ic.ac.uk
June 1993 • Usenix
USENIX Technical Conference, Cincinnati, OH, USA.
A work-in-progress presentation covering the development of the software archive.

The Replacement of a Central Server by Distributed SPARC systems − Divide and Conquer
June 1991 • Sun User Group, USA
Technical Conference, Atlanta, Georgia, USA.
Technical conference paper, presentation and newsletter article covering the replacement of a single central server with multiple Sun servers.

MIT Project Athena
1990 • UKUUG
UKUUG Security Workshop, London, England.
Presentation about MIT's Project Athena, including Kerberos.