Talks are listed in descending order of acceptance, newest talks at the top. Note that Slides/Papers are on the Schedule, where provided.
- Top 10 distributed storage systems
- The Benefits of SQL and PostgreSQL
- Performance characteristics and monitoring of NoSQL databases from the cloud
- A New OSI For A New Decade
- Design and implementation of BigV, a modern virtual machine hosting service
- Live Guest Relocation for Clustered z/VM Systems
- Automated Planning for Configuration Changes
- UK Government Open Source Action Plan
- The First Thing Tak Did - Elegant Remote Control For Sysadmins
- SSH Can Do That? Productivity Tips for Working with Remote Servers
- Ganeti - Enterprise virtualisation
- From Zero to Cloud in 90 Days with Chef
- OpenAFS: Reinventing yesterday's file system for tomorrow's applications
- Riding the Flume; Feeding your Logs into Hadoop
- Highly Available Virtualisation on a Highly Virtual Budget
- Large Scale Devolved Configuration: Realising the Benefits
- Configuration management for the masses with Rudder
- A browser based database client for the admin office
- Operating Systems - Virtualisation and system deployment using 'cloud' technology (Dr. Gerald Pfeifer)
- Central monitoring of anything, anywhere, anytime
- Large data management in virtual environment with small footprint virtual machines
- How to implement KVM with other guest systems and a web manager
- Open Enterprise Server
- Icinga at Deutsche Welle: Large Scale Distributed Monitoring for Germany's International Broadcaster
- 10 reasons for choosing Rear as a DR strategy
- Practical workflow for technical documentation
- Memory-mapped Database for OpenLDAP
- Authentication, fun times for all. A tour of Linux Examples
- How to (hopefully) avoid being r00ted
- Seven Tools for your devops stack
Top 10 distributed storage systems
This talk will give a broad summary of 10 (or more) popular distributed storage systems primarily from the perspective of cross-datacenter replication and high-availability. It will cover Ceph, GlusterFS, Riak, Cassandra, HDFS, Swift and more.
The Benefits of SQL and PostgreSQL
The SQL language offers considerable benefits and its worth reviewing what those are to help understand why 99% of data in large enterprises continues to be held in relational databases such as MySQL, Oracle and SQLServer. From there, we move on to PostgreSQL, the widely used and most advanced open source database to show the widest range of implementation options available in any database management system.
Performance characteristics and monitoring of NoSQL databases from the cloud
We’ll explore the key factors affecting NoSQL database performance, how they differ from traditional SQL databases & the challenges of monitoring their health at scale - focusing on the importance of relevant metrics in achieving high availability & performance. Using an end-to-end monitoring approach we’ll demonstrate how these metrics help improve performance & end-user experience.
A New OSI For A New Decade
The OSI (Open Source Initiative) is reorganizing its governance from a board-only organization into a member-based structure in 2012. Simon will give us a brief explanation of what that is, what it means, and how to join.
Design and implementation of BigV, a modern virtual machine hosting service
BigV is a new hosting platform designed for selling virtual machines (or VPSs or "cloud" servers), currently in beta test. It's based around Linux's KVM supervisor, intended to allow customers paid access to all of KVM's features in the context of a fast ISP network and high-performance hardware. The software has been in development since August 2010, and running live machines since August 2011. Its designer will talk through decisions made during its specification, and how successive versions have been developed, deployed and rolled out to customers with little down time. Bytemark have built several similar smaller VM systems over the years and Matthew will show how Free software, a little scripting, and a clued-up system administrator can perform tricks that a commercial solutions can't.
Live Guest Relocation for Clustered z/VM Systems
On the 12th October IBM announced the much anticipated z/VM 6.2 release which its clustering capability and ability to move live guest virtual machines from one VM hypervisor to another. Big deal? Well yes, System-Z presented interesting challenges for the developers of LGR, not least of which emanate from the design criteria for continuous availability and non-disruptive h/w update. This, coupled with the need to create a practical solution for managing fine-grained architecture variation in a Z cluster led to a novel implementation. This talk expands on the challenges met by the zVM developers and shows the degree of flexibility introduced into their design. So why talk about this at FLOSS, well for one, LGR was developed specifically to support Linux, and two zVM is semi-open source anyway.
Automated Planning for Configuration Changes
In this talk, we will present a prototype implementation of a configuration system which uses Automated Planning technique to automatically compute the workflow between two declarative states. The resulting workflow is executed using the popular combination of ControlTier and Puppet. This allows the prototype to be used in unattented "autonomic" situations where manual workflow specification is not feasible. It also ensures that critical operational constraints are maintained throughout the execution of the workflow. We will show how the configuration system solves the Cloud-Burst problem i.e. migrating a web application services from private to public cloud.
UK Government Open Source Action Plan
Tariq Rashid will explain the UK Government Open Source Action Plan, in the context the wider aims of the ICT Strategy 2011. He will cover the reasons we are interested in open source, explain why open source has not been adopted more widely, and discuss how Government aims to address these barriers. In particular, Tariq will explain the Open Source Procurement Toolkit, and key messages around IT security. The open source programme is open to ideas and feedback from all sections of the ecosystem and community. You are encouraged to participate by sharing your experiences of cultural change, addressing barriers and myths, and suggesting improvements to the Action Plan
The First Thing Tak Did - Elegant Remote Control For Sysadmins
Tak is a program for running everything from system commands and shell scripts up to complex multi step workflow-based requests on multiple systems simultaneously.
The First Thing Tak Did, He Wrote Himself.
Able to bootstrap cleanly on any LSB compliant system (and, frankly, every common server OS I've yet tried it on), Tak is able to push its core components over an ssh connection so while little installation is required on the master machine, none at all is necessary on those being controlled.
The Second Thing Tak Did, He Wrote The Laws.
Inter-node communications are handled by an extremely simple protocol encoded on the wire as one-line JSON objects, which makes socat a viable if not particularly thrilling client - but combine it with a little convention and the messaging is rich enough to put together extremely capable systems in very little code.
The Third Thing Tak Did, He Wrote The World.
Having established how the building blocks fit together ... well, you'll just have to come along and find out what comes next.
SSH Can Do That? Productivity Tips for Working with Remote Servers
Running commands on remove servers constitutes a large part of a system administrator's life. SSH makes this possible, but used in its default state can make working on a server rather cumbersome: doing something over SSH feels more awkward than performing the same task locally.
Fortunately SSH has lots of features which can make our lives much easier. Unfortunately some of these don't appear to be widely known. So this talk is here to spread the good news and enhance our productivity.
It starts with relatively basic topics, such as how to avoid repeatedly entering your password on each connection, then goes on to cover automating ‘chained’ connections through intermediate servers, running graphical editors remotely, running local programs on remote files, working easily with remote web servers and databases. It includes tips for keeping track of which window is which, and a variety of techniques for making SSH connections faster.
Many of us know much of this stuff in theory, just haven't quite yet bothered to getting round to setting it up and figuring out the precise configuration runes required. So this talk comes with copy-and-paste recipes to make that easier for you, plus a big heap of encouragement to actually do it.
Let's make life easier for ourselves!
Ganeti - Enterprise virtualisation
Ganeti is a tool for managing virtualisation on in a cluster-based model. The talk will detail the model that Google uses to deploy Ganeti, the use cases that are fulfilled by it and the internal architecture of Ganeti.
From Zero to Cloud in 90 Days with Chef
The "Cloud Paradigm" is Agile. You can't effectively get there and/or stay there if you're not using Agile methods. This means Agile development, as well as Agile operations (a.k.a., DevOps). Using "Infrastructure as Code" management tools like Opscode Chef is a cornerstone of the methods that can help you build into the cloud and stay there.
In this business, much is made of the term "hockey stick growth". Well, when you are tasked with developing a next-generation systems infrastructure for a small startup to help them prepare for the potential of this kind of growth in the future, and you're trying to do that in the cloud using tools and methodologies you've never used before, then you are faced with a "hockey stick learning curve".
This talk is a quick survey of the learning process that the speaker has been going through for his most recent customer, including lessons learned and "What Not To Do". In DevOps terminology, both patterns and anti-patterns will be shown and discussed. Brad Knowles
OpenAFS: Reinventing yesterday's file system for tomorrow's applications
AFS has been in use as an enterprise-wide distributed filesystem for the over 20 years. OpenAFS, the open source distribution of the original AFS code recently celebrated its 10 year anniversary. That's a lot of history, and a lot of baggage, however AFS remains in use, and active development, today.
We'll take a look at the original AFS feature set, and why it remains relevant to today's computing environments. We'll look at some of the new features, such as disconnected operation, which are now available within OpenAFS, and talk about the ways in which it will develop in the future. We'll also discuss some of the trials and tribulations of developing a system with such a long history, and of satisfying a user community with incredibly diverse demands and expectations. Simon Wilkinson
Riding the Flume; Feeding your Logs into Hadoop
Analysing logs is a key part of gaining information about your live system, and in this talk I will discuss how We7 have used Flume and Hadoop to move their log processing beyond the realms of shell scripts, sed and awk. Stuart Teasdale
Highly Available Virtualisation on a Highly Virtual Budget
Most system administrators have some understanding of virtualisation, and many have begun to investigate projects using tools such as libvirt and KVM, while considering high availability and redundancy as aims best left to enterprise-level providers.
This talk starts at the point most people begin - a simple virtual machine - and works through the process of moving towards a highly available, redundant, virtual machine pool, built with open-source solutions on commodity hardware. On the way we’ll cover issues relating to virtual machine management and migration using libvirt; redundant, replicated, network block devices using DRBD with iSCSI; high availability clustering with Pacemaker; and a selection of other related tools and methods.
We'll also look at some of the caveats and gotchas that are faced on the journey, including fencing and split-brain, network topology as well as planned and unplanned downtime.
This talk aims to be a broad overview of the technologies involved and how best to approach them. It aims to be partly biographical in nature: following the development of our own systems while highlighting the successes and failures encountered along the way. Matthew Richardson
Large Scale Devolved Configuration: Realising the Benefits
Devolved administration within a centralised configuration system is often invoked to help standardise a service across diverse parts of an organisation. In real life devolution is far from cost free. The University of Edinburgh has ten years experience of this model on multiple desktop operating systems: MS Windows, Linux and Mac OS X.
This talk concentrates on the centrally run devolved LCFG system used to manage several thousand Linux and Mac computers managed by tens of organisational units across the University.
I will discuss whether the expected business and community benefits were realised, and if not why not. Unforeseen benefits, new opportunities and challenges will also be presented. Kenneth MacDonald
Configuration management for the masses with Rudder
Sharing and reusing configurations, rolling out upgrades, ensuring a security policy is correctly applied, automating repetitive tasks, preparing for disaster recovery... these are all missions for configuration management tools.
Rudder is a new, open source approach to this domain, built on existing and reliable components. By allowing experts and power-users to create reusable templates and configurations based on best practices, it enables other actors in the IT department to benefit from the advantages of configuration management: using a web-based interface, junior sysadmins can quickly setup new servers while learning and respecting best practices and company policy, while service managers and security officers can get instant reports on their policies compliance level.
This talk will introduce Rudder and show some illustrative use cases before describing the architecture of it's main components and how they interact (a web interface written in Scala, the CFEngine 3 infrastructure used to manage hosts, OpenLDAP as an inventory and configuration data store...), including how to write your own policy templates and extend existing ones. Jonathan Clarke
A browser based database client for the admin office
Abstract: Large corporate applications provide end user interfaces to their back-end databases that are problematic for de-centralised multi-platform organisations. The browser based alternatives are not much better, and worse are not tailored for regular use by administrative staff facing the need to lookup or enter large amounts of data quickly. We have designed a generic lightweight browser based database client that is centrally configured, has a level of functionality similar to Microsoft Access and an end-user performance and usability comparable to dedicated platform specific applications. The result is a fast and simple interface that our administrative staff use to manage complex data sets both at work and remotely from home and that can be easily deployed and maintained. Tim Colles
Operating Systems - Virtualisation and system deployment using 'cloud' technology (Dr. Gerald Pfeifer)
Abstract: The OS is dead. In the future all workloads are going to run in the cloud anyway. Both of these statements are commonly heard these days and they are both true (to a point) and wrong (and very much so). The OS is as relevant as ever, just not as self contained. In this presentation we will show how the lifecycle of an OS instance flows from being created in SUSE Studio, deployed onto cloud infrastructure like Amazon EC2 or OpenStack, and managed using SUSE Manager. We will also see how we are leveraging and advancing core OS functionality like clustered storage and OpenStack to create and operate entire clouds. Dr. Gerald Pfeifer
Central monitoring of anything, anywhere, anytime
Abstract: Robert will explore the suite of OSS tools and techniques used to monitor & control network, server, power and other critical infrastructure, whether it's in a wiring closet on the other side of the campus, or bolted to a fish cage in the Baltic Sea. Specific topics include distributed Nagios and monitoring over an expensive or firewalled cellular link. Robert Waldie
Large data management in virtual environment with small footprint virtual machines
Abstract: This presentation is about large data management with lean agile virtual machines. The virtualization technique used is KVM with a libvirt backend. NFS shares using NFSv4 with TCP connections to allow better performance and secure remote connections over ssh connections over the public internet. Samba is used to serve these files in transparent manner and an OpenLDAP backend for single sign-on experience from the user perspective. The MySQL databases run multiple masters for read/write operation, while slaves do read-only operations. The same principle is used for the PostGreSQL database. Again Apache is the connector to the user, allowing the user to consult and/or login to the website. The whole installation uses SELinux and IPTables to secure the system while network optimization is used to keep the speed of the system very high. The whole setup, based mainly on scripts, a decentralized initiator based on git, to allow nodes to fail while the entire systems remains alive. Toshaan Bharvani
How to implement KVM with other guest systems and a web manager
Abstract: KVM is the new default virtualization technology used in Redhat based Linux distributions. A short overview on how to technology works. An explanation on how to implement virtual systems, using the default command line tools included in Redhat based Linux distributions. How to build server or desktop images, both Linux and Windows. How to implement the networking connections and how to secure the connections. A short overview of the benefits of these implementation and some drawbacks. The portability of these systems and how to migrate the systems amongst physical machines. This takes us more into how this implementation can grow to the next step and create a private cloud infrastructure for workstations. Using simple web manager to allow end user interaction and automated deployments. Toshaan Bharvani
Open Enterprise Server
Abstract: Email, scheduling, collaboration, file & document management, customer management are very important tools in running a business, however compatibility with other companies in a global business world is important. A system which is open, scalable and affordable can be built with the same features included in proprietary systems. An out-of-the-box solution doesn't exists, however it can be very easily implemented in an open source environment, based on CentOS, Zarafa, Alfresco. Each product can be used in it's open-source version or with paid options and support. The client integration uses the idea of 'if it ain't broke, don't fix it', many users do not like change, however silent changes, which do not touch the client side completely can be achieved easily, especially when saving costs. The solution accommodates both smaller and bigger implementations as the system includes scalability options. Toshaan Bharvani
Icinga at Deutsche Welle: Large Scale Distributed Monitoring for Germany's International Broadcaster
Abstract: 700 hosts, 5000 services with 2 instances in a multiple-redundant, double-cluster distributed across two German cities East and West.
Deutsche Welle, the internationally acclaimed broadcaster needed to consolidate their disparate monitoring activities spanning Bonn and Berlin into one enterprise system. Monitoring at each location needed to be self-sufficient while centrally accessible, and of course high availability was raison d'être.
This talk will tell the story of how monitoring at Deutsche Welle progressed from the design phase to the first Nagios implementation and finally a migration to a double-clustered Icinga infrastructure distributed between in Bonn and Berlin.
It will share lessons learned in clustering, configuration management, business process monitoring, performance graphing and network visualisation - with a good dose of insight into studio production, broadcasting and transmission monitoring. Above all, it will show how monitoring can be operationally independent between two cities, yet centrally managed at the same time. Bernd Erk
10 reasons for choosing Rear as a DR strategy
Abstract: Relax and Recover (Rear) is an Open Source tool that implements a simple, yet effective, workflow for bare metal disaster recovery. The design approach taken by its authors delivers a setup-and-forget solution that is easy to deploy, offers a high success rate and makes recovery simple as possible.
So there, I already gave away 3 reasons ! Maybe I'll higher the stakes and bring it to 13 reasons, who knows ? So if you are looking for a solution to complement you existing backup strategy, or if you're looking for a one-stop solution, look no more !
Practical workflow for technical documentation
Abstract: Writing technical documentation can be fun, even if you have to produce DOC, ODF or PDF files that need to conform to a visual corporate identity. In this presentation we will look at the various options and we will highlight a specific solution that brings git/subversion, vim/emacs, LibreOffice/OpenOffice and Makefiles together in a beautiful (subject to eye of the beholder) marriage of Open Source technology.
If you often have to write proposals, design documents or technical procedures, you will enjoy this presentation and proposed solution. The toolchain increases efficiency, makes collaboration possible. Styling and formatting become an afterthought.
Memory-mapped Database for OpenLDAP
Abstract: While OpenLDAP already provides a reliable high performance transactional backend database (using BerkeleyDB), it requires careful tuning to get good results and the tuning aspects can be quite complex. Data comes through three separate layers of caches before it may be used, and each cache layer has a significant footprint. Balancing the three layers against each other can be a difficult juggling act.
This talk presents the design and implementation of a new "back-mdb" memory-mapped database backend for OpenLDAP. This is built on top of a new mdb library written from scratch for the purpose. The library implements B-trees with multi-version concurrency support, and all "reads" are performed by mapping the entire database into virtual memory. Howard Chu
Authentication, fun times for all. A tour of Linux Examples
Abstract: Authentication take many forms on modern unix systems, in this talk I'll cover the setups of a few authentication systems (like SSL and krb5) and then show how they are implimented in databases, openldap, webservers and many other services which run on a variety of unix platforms. This talk will cover practical examples with an enphasis on getting the job done and making day to day system administration easier. Faye Gibbins
How to (hopefully) avoid being r00ted
Abstract: This talk will review the techniques commonly used by attackers to compromise Linux systems. It will then cover the ways in which a system administrator can use this knowledge to make life very difficult for attackers. As no fortress can be utterly impenetrable whilst remaining accessible to authorised people I will also discuss how attempts to compromise a system can be detected quickly. Stephen Quinney
Seven Tools for your devops stack
Abstract: Technology moves fast, most people are so busy they don't have time to keep up with what's new, or sometimes don't really understand the need for these tools, until they take 5 minutes and listen to somebody using them.
This talk will go over a bunch of unmissable open source system tools tools, some of them didn't even exist 2 years ago,
We won't spill which ones we'll cover .. but rest assured .. you'll learn a few, Kris Buytaert
Note that Slides/Papers are on the Schedule, where provided.