- Weta Digital - Challenges in Data Centre Growth - Paul Gunn
- System deployment and bare metal recovery by Clonezilla - Steven Shiau
- Going mad with MDADM - Steven Ellis
- Samba4: We now do AD replication with windows - Andrew Bartlett
- Linux Containers: virtualization without overhead or strange patches - Sam Vilain
- Being Lazy in a Large Organisation - Documentation by Wiki - Mark Suter
- Behaviour driven monitoring with cucumber-nagios - Lindsay Holmwood
- Lies, damn lies, statistics and benchmarks - Devdas Bhagat
- Edubuntu - supervised and controlled learning in a fun and free environment - Craig Pearce
- Developing and Managing Linux SOE Environments - Matthew Lye
-
Weta Digital - Challenges in Data Centre Growth - Paul Gunn
(or "You need how many processors to finish the movie???")As the scale and complexity of the visual effects shots that Weta Digital produces has increased, so have the demands on the physical and computing infrastructure. I will first give a brief introduction to the visual effects pipeline. I'll then cover the rapid growth in our infrastructure, and the significant technology changes we've had to embrace to keep our data centre running.
Paul Gunn has beenworking for Weta Digital for several years building data centres.
Length: 45m
-
System deployment and bare metal recovery by Clonezilla - Steven Shiau
Most people and companies think they can rebuild system by backing up critical data. However, when catastrophe failure occurs, they lose all, including entire operating system, data, setting and applications. Therefore disaster recovery is always an issue not only for enterprise but also for a person
Clonezilla, licensed under GPL, is a partition or disk cloning tool similar to Symantec Ghost or Acronis True Image. Based on DRBL, Partition Image, ntfsclone, partclone, and udpcast, Clonezilla allows you to do system deployment and bare metal backup and recovery. The file systems supported by Clonezilla are ext2, ext3, ext4, reiserfs, reiser4, xfs, jfs of GNU/Linux, FAT, NTFS of MS Windows, and HFS+ of Mac OS. Therefore you can clone GNU/Linux, MS windows and Intel-based Mac OS. LVM2 under GNU/Linux is supported, too.
Two types of Clonezilla are available, Clonezilla live and Clonezilla SE (server edition). Clonezilla live is suitable for single machine backup and restore. While Clonezilla SE is for massive deployment, it can clone many (40 plus!) computers simultaneously. Clonezilla saves and restores only used blocks in the hard disk. This increases the clone efficiency. E.g. Clonezilla SE was used to clone 41 computers simultaneously, and it took only about 10 minutes to clone a 5.6 GBytes system image to all 41 computers via multicasting
So far the download number of Clonezilla is more than 1,000,000, that implies that it\'s a reliable and popular
In this talk, the speaker will cover how to:
1. Use Clonezilla live to do a bare metal recovery and disk clone
2. Create your own recovery CD or USB flash drive
3. Use Clonezilla SE to do system deployment, e.g. multicast restoringBio
Steven Shiau is a researcher at the NCHC (National Center for High-performance Computing), Taiwan. While there he wrote parallel and distributed program to simulate plasma. This program formed the basic idea for the future development of the free software DRBL (Diskless Remote Boot in Linux), a rapid deployment software for distributed PC cluster and education environment. With Kuo-Lien Huang, Ceasar Sun, Jazz Wang and Thomas Tsai, he developed another free software "Clonezilla", an open source clone system. He has given more than 50 lessons about DRBL and Clonezilla in Taiwan from 2002, and 4 talks in Libre Software Meeting in France since 2006. The project DRBL won first place in 'Public Sector Applications' category at the Free Software Contest in France in 2007. He is the devision leader of Grid technology devision at NCHC now, and spends most of his time in the development of DRBL, Clonezilla and promoting the use of free software and high-performance computing.
Length: 20m
-
Going mad with MDADM - Steven Ellis
Whilst in the enterprise the preference is to utilise hardware raid controllers there are occasions where Software Raid via MDADM is appropriate. So what happens when software raid goes wrong, and what is the best way to troubleshoot.
During the session I'll walk you through a real world situation where I experienced a very wide range of emotions whilst trying to troubleshoot a major issue with Linux software raid. With luck you will learn from some of my mistakes and come away with a good overview of how to analyse and fix potential issues.
Bio
* Technical Director OpenMedia Limited, Auckland New Zealand * Director Global Engineering, Bulletin.net Limited
Steven's passion for FOSS comes from both development and operational experience. Within his role as OpenMedia\'s Technical Director he provides technical direction and operational guidance for a number of New Zealand companies on the use of Open Source and Linux. For over three years he ran the Linux operations team at IBM NZ, looking after a number of their enterprise customers in the Asia Pacific market, and recently tool a role with Bulletin.net managing their infrastructure, operations and support.
As developer of the MythTV based consumer appliance myPVR for OpenMedia, Steven has leveraged his over 14 years experience with Linux, and nearly 20 years experience of FOSS. OpenMedia was one of the first companies in the world to offer a truly consumer ready appliance based on MythTV.
Steven gives regular talks on FOSS to the Auckland Linux User Group and Auckland BarCamp, and has given presentations and organised tutorials at IBM NZ and Optimation NZ. He has been an invited speaker at a number of regional and international conferences including OSDC, Linux.conf.au, Linux World and OSCON.
Length: 20m
-
Samba4: We now do AD replication with windows - Andrew Bartlett
Since last years SysAdmin miniconf update on Samba4, we have made great progress, bringing on new team members, and bringing up great functionality. We are still making alpha releases, but importantly we now have a working implementation of the native AD replication technology: DRS.
Using this important advance, we are now aiming to provide a 'read only domain controller', in the hope that we might soon have caching domain controllers on the edges of existing corporate networks, able to safely assist in load management and branch office connectivity. (We are also pushing forward with our full read-write replication with Active Directory).
This talk will outline recent progress, and weather permitting, show a demo of some of these changes in action.
Bio
Andrew has been a member of the Samba Team since 2001, and has worked on the Samba4 project since 2004. Andrew's development focus in on Authentication systems (such as Kerberos) but he takes a keen interest in all the pieces that make Samba4 an Active Directory domain controller. He regularly speaks at the annual SambaXP conference in Germany, and at the annual CIFS conference in San Jose.
Away from the computer Andrew prefers to be cycling or walking the Australian Bush.
Length: 45m
-
Linux Containers: virtualization without overhead or strange patches - Sam Vilain
Virtualization can be achieved in many ways: full-CPU emulation like Qemu has its place, gradually moving through microkernel/hypervisor systems like KVM and Xen, and finally at the simplest end you have jail/container-based systems, from the humble chroot() to Linux-VServer and openvz.
This talk will discuss briefly the differences in these complementary approaches to virtualization, then in more detail explain the "namespaces", "containers" , "controllers" concepts, standard features added between Linux 2.4.9 and 2.6.29 that require no kernel patches to use, and how they combine to form a virtualization solution for Linux userland environments. Their functionality will be compared and contrasted with the closest fully-featured non-mainstream solution, Linux VServer.
Bio
Sam is a Perl programmer and Systems hacker based in Wellington, working for Catalyst IT. He has spoken at many open source conferences around the world, including several Perl conferences, OSDC, OSCON, and the Dunedin Linux.conf.au.
He also implemented some vserver features in the early days of the project, and participated in the early discussions around the design of Linux containers
Length: 20m
-
Being Lazy in a Large Organisation - Documentation by Wiki - Mark Suter
Ever seen an error in documentation but gave up trying to get it fixed or didn't even try because it would be too hard?
Ever despaired about the quality of documentation but felt powerless to improve it?
Ever tried to implement a wiki within a large, conservative organisation?
Getting a wiki going was one of the first things I accomplished when I started with Unisys Australia in July 2008. That wiki has become the authoritative source of truth on many topics (both internal and formal client facing), with over 2,200 articles being maintained by several diverse teams.
This talk covers both technical and business/social issues in getting a wiki starting an an IT Services company contracted to a Government Department.
Bio
Mark Suter is a geek with a deep interest in the how and why of complex systems. I have a strong focus on getting things to work well and contributing to the lifelong learning of my fellow professionals.
Length: 20m
-
Behaviour driven monitoring with cucumber-nagios - Lindsay Holmwood
Setting up monitoring for web applications can be complicated - tests tend to lack expressiveness, and quite often don't even test the right problem in the first place.
cucumber-nagios lets you write behavioural tests for your web apps in plain English and outputs the test results in the Nagios plugin format, letting Nagios notify you when your production apps aren't behaving.
Lindsay will be covering how to get up and running with the tool, writing tests for your web apps, and why it's important to test the behaviour (and not just the availability) of your production web apps.
Bio
Lindsay Holmwood is a sysadmin/developer from Sydney, Australia. He's the creator of cucumber-nagios, Flapjack, Visage, and Gastro.
Length: 20m
-
Lies, damn lies, statistics and benchmarks - Devdas Bhagat
System administrators are often tasked with figuring out where things go wrong, and why. They may or may not have the resources to fix things offhand, but they still need to be able to justify their requirements for hardware/software upgrades (and fixes).
Using simple, standard tools, system administrators can setup monitoring systems which track important business metrics (not just hosts or services), and make it far easier to show the business value in obtaining better resources.
Knowing what metrics the business values also makes it easier for operations staff not to over-optimise minor constraints, and instead focus on those bits which gives most value.
This talk will attempt to provide a broad overview of what operations teams can (and should) do to monitor systems, what sort of reporting measures need to be applied and how they can be fitted into the business side of things. The same reporting systems will also help them make decisions on sourcing new hardware, or trying to fix issues in some other way.
Bio
Devdas Bhagat is just another recovering sysadmin, and is currently playing database administrator for a large domain registration system. In his copious spare time, he works with other business units at his workplace to help tune their systems, add business benchmarks and other general monitoring stuff.
He has spoken at the sysadmin miniconf at LCA09 on system automation, and on DNS benchmarking. He has spoken at various network operator group conferences on spam control, DNS, operational issues, large scale email systems, etc. He has also been a speaker on databases at FOSS.IN and it\'s previous avatar, Linux-Bangalore.
When not busy working with computers, he spends time in the hills trekking and photographing.
Length: 20m
-
Edubuntu - supervised and controlled learning in a fun and free environment - Craig Pearce
Edubuntu is becoming a popular platform for childhood learning activities. It is designed with a core requirement to operate out-of-the-box with minimal administrative overheads. As shown in this presentation, the project has not yet fully captured and implemented requirements around parental controls. Many parents consider this an important component in order to put in place supervision and protections for children using Edubuntu.
Bio
Craig Pearce is an enterprise IT security designer for a major-4 Australian bank and a sessional lecturer at RMIT University in computer security, privacy and distributed computing. He is also a CompTIA Linux+ exam writer. Craig last presented at LCA back in 2005 and has presented at AusCERT and several international conferences in topics of computer security and privacy.
Craig is also an official member of the Ubuntu Australian Team, Edubuntu advocacy team and GChildCare projects. Craig has a wife and two very young children of whom he attempts to use as guineau pigs for various Linux-based hobbies around the house.
Length: 20m
-
Developing and Managing Linux SOE Environments - Matthew Lye
An overview of Griffith Universities SOE Linux teaching environments and hopefully a guide to how to develop your own.
We release a new version every 6 months (to coincide with the start of semesters) that is deployed into teaching lab environments. With minor changes this is also usable by staff as well. This version is typically the latest version of Ubuntu released previous to our development cycle. This was mostly due to problems with older kernels supporting newer hardware (no longer a problem in Ubuntu)
Simplified Maintenance:
We utilize rc.local to automate updates and changes required to groups of machines through a process of updating itself off a server on boot. We use CSSH to make minor changes to individual labs.Networked Logins and Home Directories:
We have recently transitioned our labs from NDS to AD for logins via PAM modules. Information on how to do this both ways will be available. We use networked home directories so that users can store information that is available on any lab machine. We also use a passwd file containing unix attributes for all networked homes to both verify that users are meant to have access to the linux environments, and to keep unix attributes separate to the networked login environment variables.Gnome Customization and Themes
Configuring login screens, preventing users from changing settings using gconf-tool2 and gconfeditor to preserve corporate branding. Editing menus to hide categories such as Games. Setting MOTD/Acceptable Usage Policy messages to show up before login.Bio
Matthew Lye works for Griffith University managing linux environments. His experience is largely based around Ubuntu and is a Ubuntu Member and involved in many Ubuntu community teams and projects.
Matthew currently has a very small baby daughter and is looking forward to teaching her linux...
Length: 20m