Loading…

Sign up or log in to bookmark your favorites and sync them to your phone or calendar.

Compute [clear filter]
Monday, May 12
 

11:15am

Multi-Tenant Bare Metal Provisioning with Ironic

Rackspace deploys a lot of servers. Over the years we’ve developed a huge variety of tools that we leverage to automate bare-metal deployment on behalf of customers. Recently we’ve undertaken a project to add Ironic to our arsenal of bare-metal deployment tools.

Ironic is an OpenStack project which offers an API for bare metal provisioning. Ironic's first "customer" has been the TripleO project, which targets Cloud operators searching for a means of easily bootstrapping hardware in a trusted environment.

This session will discuss the challenges we have faced in adapting Ironic for use in a multitenant environment. Topics covered will include network isolation, hardware security, operational reliability and how we have harnessed Ironic’s flexible driver model to automate provisioning for our customers. This session will also include a discussion of the direction and scope of Ironic and where the boundary should lie between Ironic and Nova.




Speakers
avatar for Russell Haering

Russell Haering

Manager, Software Development, Rackspace
Russell is an engineering manager building infrastructure tools at Rackspace. His interests include building high performance distributed systems, product strategy and scaling engineering teams.


Monday May 12, 2014 11:15am - 11:55am
Room B206

12:05pm

Under the Hood with Nova, Libvirt, and KVM
The Libvirt driver for Nova is one of the most comprehensive implementations across all supported hypervisors. It is quite common for new functionality to be added Libvirt before it hits the others. API calls to manage instances generally translate to a complex set of interactions between Nova, Libvirt and QEMU/KVM.
During this talk, we will go under the hood to understand exactly what our KVM hypervisors are doing when various requests are made. This includes a deep dive into the sequence of operations when an instance is spawned, resized, migrated (live and cold), rebuilt or snapshot, to name a few. Which areas are most brittle and why? What happens when these operations fail of complete impartially? How can we recover? You will be able to answer these and many other related questions once we are through this technical jam session.
We will also cover some of the functionality added for Ice House and the new Libvirt and KVM capabilities driving Nova-related development.

Speakers
RK

Rafi Khardalian

CTO, Metacloud, Metacloud
Rafi Khardalian is Chief Technical Officer and a member of the founding team at Metacloud. Prior to joining Metacloud, Rafi was the Director of Systems Architecture at Ticketmaster Entertainment. He was responsible for driving innovation while designing highly scalable, globally distributed... Read More →


Monday May 12, 2014 12:05pm - 12:45pm
Room B206

2:00pm

Continuous Integration Testing for Your Database Migrations
This isn't a talk about OpenStack. This is a talk about how OpenStack does database migration continuous integration, and how other projects might be able to learn from what we've done.

OpenStack is an interesting project in that it runs in a variety of different scale scenarios. There are very small deployments of only a few machines, up to large public clouds of thousands of machines. The OpenStack project takes Continuous Integration (CI) very seriously, and runs a suite of unit and integration tests on all proposed patches.

One thing we weren't very good at testing was database upgrades though. We have unit tests for many of them, but we don't know how these upgrades affect real users in the field. As a response, a small team I manage have worked on adding database migration testing to our CI environment. This new test infrastructure is open source (https://github.com/stackforge/turbo-hipster), and extensible. It currently runs tests on smaller databases, as well as a largish real user database. It has already caught two significant performance problems in database migrations.

This talk will cover how the test infrastructure is built, what we've learned from the six months of running it that we'll have had by the time of the conference, as well as telling a story about some of the bugs we've found and how we fixed them.

Speakers
avatar for Joshua Hesketh

Joshua Hesketh

Software Developer, Rackspace
Joshua Hesketh is a software developer for Rackspace Australia working on upstream OpenStack. He works from his home in Hobart, Tasmania. Joshua is currently President of Linux Australia, previously the co-chair for PyCon Australia and a key organiser for linux.conf.au. He has an... Read More →


Monday May 12, 2014 2:00pm - 2:40pm
Room B206

2:50pm

Nova's March Towards Live Upgrade Capability
In large deployments, some piece of Nova is deployed almost everywhere.
Until recently, the story for how you upgrade to using a newer version of this fast-paced project has been turn off the cloud" -- an unacceptable answer for something that aims to be deployed at a scale large enough to make such a plan unfeasible. Over the last several releases, "upgrade-ability" has been a major theme of development for the Nova team, shooting for both reduced downtime of core components and elimination of the need to "turn off the cloud" while thousands of nodes are upgraded atomically.
This talk will cover the incremental steps we have made in recent cycles, as well as reveal the major new capability that the Icehouse release brings to the table.
"

Speakers
avatar for Dan Smith

Dan Smith

Speaker, Red Hat


Monday May 12, 2014 2:50pm - 3:30pm
Room B206

3:40pm

Hosting Hybrid (Bare-metal + Virtualized) Applications on OpenStack
OpenStack is quickly gaining momentum as a general-purpose IaaS platform to host a variety of applications. However, for some applications, running on shared or general-purpose virtualized hardware may be suboptimal or even unacceptable for a number of reasons, including performance or security. Workloads that can benefit from dedicated, physical systems include extremely I/O intensive and/or CPU intensive applications (analytics, 3D modeling, media transcoding, HPC, etc) or applications designed to leverage specialized hardware such as GPUs. Moreover, many applications have 'hybrid' nature, comprising different components with different infrastructure requirements, that can be best met with different kinds of hardware -- virtualized or bare-metal.
In this talk we will survey the main challenges in hosting 'bare-metal' and 'hybrid' applications in the cloud in general, and using OpenStack in particular, will outline possible design approaches, and will demonstrate how OpenStack can be enhanced to optimally host such applications.


Speakers
avatar for Ofer Biran

Ofer Biran

STSM, Manager - Virtualization and Systems Management, IBM Research - Haifa
http://researcher.watson.ibm.com/researcher/view.php?person=il-BIRAN
avatar for Alex Glikson

Alex Glikson

Manager, Cloud Operating System Technologies, IBM Research
Mr Alex Glikson is leading a research group at IBM Haifa Research Lab working on research and development of innovative cloud infrastructure management technologies - focusing on OpenStack (http://www.research.ibm.com/haifa/dept/stt/cloud_sys.shtml). In this role Alex closely col... Read More →
avatar for Ezra Silvera

Ezra Silvera

Cloud & Virtualization architect, IBM Research - Haifa Labs


Monday May 12, 2014 3:40pm - 4:20pm
Room B206

4:40pm

Enterprise-Grade Scheduling: Enterprise-Grade OpenStack from a Scheduler Perspective
Enterprise-Grade Openstack is all about the delivery of cloud services to meet the service-level requirements of enterprise mission critical and performance critical application in large scale.
In our presentation we will focus on the different Openstack schedulers' implications of the enterprise-grade requirements. We will talk about hierarchical-scheduling", "cross-scheduling" and "re-scheduling," in the context of providing highly-available and high-performance compute, networking and storage services in large scale. We will also cover recent developments in Openstack schedulers in the context of the general Enterprise-Grade Openstack roadmap.

This is a joint work of Gary Kotton from VMware and Gilad Zlotkin from Radware.
"

Speakers
avatar for Gary Kotton

Gary Kotton

Staff Engineer at VMware, VMware
Gary is a core Neutron developer working at VMware who also spends a lot of his time these days writing + reviewing code for Nova. Prior to working at VMware Gary worked at Red Hat, Radware and at Algorithmic Research. Gary holds a Bs.C in Mathematics and Computer Science from the... Read More →
avatar for Gilad Zlotkin

Gilad Zlotkin

VP Virtualization & Management Products, Radware
Dr. Gilad Zlotkin is responsible for leading Radware?s Cloud strategy, product development, roadmap, and management solutions. Gilad's experience in cloud service delivery, virtualization operations management, network management and distributed systems has contributed to the development... Read More →


Monday May 12, 2014 4:40pm - 5:20pm
Room B206

5:30pm

Bridging The Gap: OpenStack For VMware Administrators
As OpenStack continues to grow, Enterprises are beginning to explore and to implement OpenStack as their Cloud platform of choice. Often, these companies have existing investments and expertise with VMware technologies. In order to prepare for this new world, these people who are familiar with VMware concepts and terminology will need to understand the parallel concepts and terminology in OpenStack. In this keynote, a VMware veteran who have moved from the VMware space into the OpenStack space will talk about how to bridge the gap between vSphere and OpenStack concepts and. This session will be valuable for anyone who needs a better grasp of how to talk about both VMware and OpenStack in an enterprise context.
We will also explore the concept of Cloud-native applications and talk about the differences between operating a VMware environment and an OpenStack Cloud platform. Finally, we will discuss how VMware, as a company, has been bridging these 2 world by integrating their technologies with OpenStack in both the Havana and Icehouse releases.

Speakers
CB

Cody Bunch

Principal Architect, Private Cloud at Rackspace, Rackspace
Cody Bunch is a Principal Architect with Rackspace. Cody has authored or co-authored several OpenStack and VMware books. Cody also regularly speaks at industry events and local user groups. You can follow Cody on Twitter @cody_bunch.
avatar for Kenneth Hui

Kenneth Hui

Technology Evangelist, Rackspace
Kenneth Hui is a Technology Evangelist with Rackspace. His passion is to help IT deliver value to their customers through collaboration, automation, and cloud computing. Ken blogs about cloud computing, OpenStack, and VMware at http://cloudarchitectmusings.com and at http://www.rackspace.com./blog/author/kenneth-hui... Read More →
avatar for Scott Lowe

Scott Lowe

Engineering Architect, VMware, Inc.
Scott Lowe is a blogger, speaker, best-selling author, podcaster, and IT industry veteran. In his role as an engineering architect on the NSX team at VMware, Inc., he focuses on open source, cloud computing, networking, and network virtualization. Scott has authored a number of books... Read More →


Monday May 12, 2014 5:30pm - 6:10pm
Room B206
 
Tuesday, May 13
 

11:15am

Enhancing High Availability in Context of OpenStack
There are 4 layers of HA to consider in a cloud deployment: the application itself, the virtual machine the application runs on, and the control plane of the cloud that allows provisioning and managing resources. The OpenStack development community is largely focused on HA of OpenStack itself (i.e. the control plane). VM-level HA has some limited support in OpenStack, but it is not complete. And, while application level HA is largely considered out of scope of OpenStack, Savanna, Heat, and Trove are beginning to touch on this.
This session will focus on the current status of HA at each level, technical challenges that remain, and the research and development activities to close them.

Speakers
avatar for Qiming Teng

Qiming Teng

Research Scientist, IBM
I have been working on virtualization technologies since 2005, though this is the first time I get myself involved in an open community. It has been an exciting experience so far where I can work with so many talents around the globe, all with different backgrounds, interests, expertise... Read More →


Tuesday May 13, 2014 11:15am - 11:15am
Room B206

12:05pm

OpenStack Python and the Holy Grail: A New Proposal for Image Portability
We've all heard it: portability, federation, interconnected. The Holy Grail of the open cloud movement is arguably compute instance portability, but it may be possible to port virtual machine images now. While open source tools exist to covert raw VM disk images, image portability may be best achieved by OpenStack providers by adopting some common bootstrapping and data transfer patterns. In this session, John Garbutt and Brian Rosmaita from Rackspace will explore what that may entail. And there will be much rejoicing.

Speakers
JG

John Garbutt

Software Developer IV, Rackspace
John Garbutt is a Senior Software Developer at Rackspace, part of the team building their Open Cloud, and has core status in OpenStack Nova. His work concentrates on how XCP and XenServer integrate with OpenStack. Previously, he was working on similar things at Citrix. John started... Read More →
avatar for Brian Rosmaita

Brian Rosmaita

Senior Software Developer, Rackspace
Brian Rosmaita is a Senior Software Developer at Rackspace.  He's been an active technical contributor to OpenStack since the Folsom release and was a software developer on the Rackspace first generation cloud.  During a stint as a product manager, he was instrumental in launching... Read More →


Tuesday May 13, 2014 12:05pm - 12:45pm
Room B206

2:00pm

Divide and Conquer: Resource Segregation in the OpenStack Cloud
As a cloud deployment grows both in size and in variety of workloads it is useful to be able to segregate compute resources based on factors such as physical location, networking connections, or the availability of specialized hardware. OpenStack provides a number of options for segregating compute resources in the cloud including:
  • Cells,
  • Host Aggregates,
  • Regions, and
  • Availability Zones.
The differences between each of these approaches are nuanced and often misunderstood. Additionally the capabilities available when using some of these techniques, particularly Host Aggregates and Cells, have been expanded in recent OpenStack releases.
This session will provide an indepth look into these each of these options for segregating resources in the OpenStack cloud including their pros and cons, example use cases, and performance considerations.

Speakers
avatar for Steve Gordon

Steve Gordon

Principal Product Manager, Red Hat
Geographically displaced Australian. Focused on building infrastructure solutions for compute use cases using a spectrum of virtualization, containerization, and bare-metal provisioning technologies. Stephen is currently a Principal Product Manager at Red Hat based in Toronto, Canada... Read More →


Tuesday May 13, 2014 2:00pm - 2:40pm
Room B206

2:50pm

OpenStack at DreamHost: Fantasies, Realities, Dreams, and Nightmares
DreamHost was founded in 1997 as a web hosting provider, and has been helping developers and entrepreneurs to reach their dreams for over 16 years with its shared, VPS, and dedicated hosting services. Since 2011, DreamHost has been developing, deploying, and selling cloud services, including its OpenStack Swift and Amazon S3 compatible object storage service, DreamObjects, and DreamCompute, an OpenStack-based public cloud deployment.
DreamCompute leverages OpenStack, and is built with unique features such as custom virtual networking, Ceph-backed block storage, and an open architecture. Creating DreamCompute with a very small team has been a challenging effort, and this talk will focus on some of the main issues that DreamHost has faced during development, including:
  • The creation and architecture of Akanda, DreamHost’s open source Layer 3+ virtual networking solution.
  • Scaling of OpenStack services, especially Nova, Neutron, and Cinder.
  • Deployment and management of a multi-petabyte Ceph storage cluster.
  • Continuous Integration and Deployment, leveraging Chef, Jenkins, Jenkins Job Builder, and fpm.
Come join us as we investigate the fantasies, realities, dreams, and nightmares we’ve encountered as we’ve developed DreamCompute.

Speakers
JH

Jeremy Hanmer

Security/Network/System Administration, DreamHost
avatar for Jonathan LaCour

Jonathan LaCour

VP, Cloud, DreamHost
Jonathan has been programming from a young age, finessing his way through many programming languages. After cutting his teeth with several smaller jobs and personal projects, Jonathan joined VertiSoft Corp. in 1997, catapulting him into the culture of "software as a service" well... Read More →


Tuesday May 13, 2014 2:50pm - 3:30pm
Room B206

3:40pm

Increasing Read Scalability in Openstack Using Database Replication
A key scalability challenge for almost all OpenStack components is a reliance on a single authoritative store of service related information. While not always the case, generally this comes in the form of a MySQL or PostgreSQL database. As production deployments become larger and Openstack continues to move into the enterprise space we will need to be able to scale this central store. A quick familiar method of decentralizing traditional relational databases has been through use of replications slaves. In this presentation I will explain how support for replication slaves works in Openstack and what currently uses it. I will also go over some production experience as Bluehost has been using replication slaves for over a year.

Speakers
avatar for Mike Wilson

Mike Wilson

Systems Architect at Mirantis, Mirantis
Mike Wilson comes from a background of large scale multi-datacenter computing and has been doing big OpenStack since the Folsom release. His experience includes upstream development as well as being the principal architect and lead in setting up one of the largest OpenStack deployments... Read More →


Tuesday May 13, 2014 3:40pm - 4:20pm
Room B206