OpenStack is the fastest growing open source movement in history, but its marketing momentum has largely outrun its technology growth. Why are organizations so eager to embrace OpenStack? Some components – like Swift – are ready for prime time. But others – like Horizon and Quantum – are still evolving. What needs the most attention: networking, storage, compute, or something else? Where are the reference
architectures and real world deployments? How are different product and service companies implementing OpenStack in production today? We'll go beyond the hype and dig deep on OpenStack, exploring all that is great and all that needs serious work. Attendees will leave with a firsthand account of the State of the Stack, ready to help their organizations embrace OpenStack armed with practical knowledge.
Cloud brings the promise of quickly spinning up virtual machines and application, but network services often lag behind lacking the automation and rapid provisioning capabilities of compute. We will be discussing the benefits of enabling advanced network services in your cloud such as Server Loadbalancing, site and user VPNs, and Firewalls via Quantum's REST APIs that would enable on-demand provisioning of these services at the time of application deployment. We will also explore the benefits of using virtual appliances to deliver these services on top of standard x86 servers to reduce specialized hardware requirements for cloud build-out to decouple network service feature delivery from hardware installs, procurement, and forklift upgrades.
To conclude, we'll be showing an early integration demo previewing some of VMware's networking and security services in action through Quantum which you won't want to miss!
On a beefy machine (24 cores, 96 GB RAM, SSD), booting a single instance (from "nova boot" to ACTIVE) takes seconds. However, when you try booting 20 instances in parallel, the last instance might not be ACTIVE for minutes! While you're waiting, you notice that the host's CPUs and disk are mostly idle and there's plenty of free RAM. While your instances are BUILDING, you wonder what's going on -- why's this taking so long?
It turns out that lengthy portions of the boot process are serialized by contention for software resources, like iptables, database connections, libvirt, and the python interpreter! In this talk, we show how tools like strace and Tracelytics can be used to identify bottlenecks in Openstack. We present techniques for eliminating these bottlenecks, such as coalescing updates to iptables and avoiding greethreads pitfalls, and demonstrate how boot can scale!
OpenStack currently enables the provisioning and operation of applications on cloud VMs, but what if you want to provision onto real hardware instead of VMs? Wouldn't it be nice to use OpenStack for this, too?
Going one step further, what if you could easily use OpenStack itself to orchestrate the deployment and scaling of an OpenStack cloud onto new machine nodes?
At HP Cloud Services, we think that the possibility of a single suite of tools and a unified API for managing both physical and virtual resources is very exciting. Working with others in the community, we have added the "baremetal" compute driver to the Grizzly release of OpenStack to provide a common framework for deploying images to real hardware. We will present the current design of the driver, some limitations it currently has, how to simulate an environment for dev/test purposes, and cover some of the quirks in deploying it. After all, this isn't your average hypervisor.
Can we build an OpenStack HA solution that allows the same type of automation as Amazon AWS? This session compares an OpenStack HA reference architecture to Amazon to see how OpenStack stacks up.
As OpenStack goes real world production environment, the real world problems like reliability, high availability need to be addressed. In this session, we will explore various time-tested techniques for improving availability of your applications on OpenStack private cloud.
We'll be joined by RIghtScale customer, Samsung SDS, to discuss their revolutionary OpenStack project. Samsung SDS has devised a multi-cloud architecture that leverages OpenStack and AWS to enable cloud-bursting while eliminating latency and security issues. Samsung SDS will review the architecture and technologies they are levering to make this solution possible.
Key take-aways from the presentation would be:
OpenStack high availability has seen a lot of progress since it was defined as an overarching design goal during the Folsom cycle. Not only is there now a reference architecture for highly available OpenStack infrastructure services, but OpenStack is also gradually growing native high availability features. In this session, we are giving an overview of the current state of high availability technology in OpenStack, and an outlook as we enter the Havana cycle.
In particular, this session covers >
Infrastructure HA:
High Availability features in OpenStack Compute (Nova):
High availability improvements in OpenStack Networking (Quantum)
Attendees should have a good general understanding of OpenStack components. High availability experience is a plus, but not required.
In this talk we will explain how to make some OpenStack services highly available in active / active mode, or active / passive for some of them. We will also approach the subject of shared storage and how OpenStack is using it. We will share our experience on production environments where High Availabity is one of our top priority.
Cloud Networking introduces several new concepts and practices that change the way traditional networks are being built and managed. Network architects, solution designers and application developers need to
understand these new networking capabilities to take advantage of the cloud. This talk aims to demystify cloud networking to the above audience by providing a deep dive analysis of the various cloud networking models and capabilities by using specific networking scenarios from the OpenStack Quantum service and Amazon EC2. We will also examine how these capabilities could be leveraged to build fault-tolerant cloud applications.
Existing approaches to delivering persistent block storage in OpenStack focus on integrating existing SAN/NAS hardware solutions, using Distributed File Systems (DFS), or using simple Direct Attached Storage (DAS) with Cinder. There is another alternative: scale-out block storage nodes with intelligent scheduling. This is the same approach that Amazon Web Services (AWS) uses for Elastic Block Storage (EBS) and it's worth taking a close look at the pros and cons. This presentation will explore the differences between SAN, NAS, DFS, DAS, and EBS. We will look at the implicit and explicit contracts that users and operators get from the different approaches and look at a variety of failure conditions. EBS may not be right for some clouds, but for many it's an important and viable alternative to the existing approaches.
Eventlet is a core library that Openstack depends on for network communications. Eventlet is not magic, though often it is treated as such. In this talk, I will dispell the magic, and discuss best practices for using Eventlet.
With the Grizzly release comes many new and exiciting features for Windows Server 2012 and Hyper-V. We will discuss new features including
The current implementation of VMware VC compute driver for OpenStack uses one proxy server to run nova-compute service to manage a cluster. In this session, we would cover the changes implemented to enhance VMware VC Compute driver so that it runs as a Proxy Compute Service to manage multiple VMware vCenter Clusters and Resource Pools as compute nodes. These proposed changes are in line with nova Bare metal proxy driver.
Highlights of these changes:
OpenStack Grizzly will finally include support for elastic load balancing. Quantum LBaaS project provides a standardized REST API that abstracts diverse hardware and software-based load balancers. This allows administrators and applications to instantiate and configure virtual and physical load balancers on demand.
The talk will walk attendees through key features of Quantum LBaaS and will include a live demonstration of managing HA-proxy instances. We will also discuss supported load balancers and go over the future roadmap.
We present the Openstack architecture that integrates Openflow based software defined networking (SDN) enabling automation and provisioning of network services spanning virtual switches (OVS) and physical switches. The talk will provide a deep understanding of the architecture components and the interactions. We will also discuss the unique benefits of SDN/openflow vis-a-vis Openstack, compare with existing Quantum supported networking architectures and share our deployment based experiences.
This presentation will be an in-depth critique of the existing OpenStack networking approach, with a focus on how the Nova network controller is more of a hindrance than a help. We will also discuss the changes in Quantum's functionality required to close the gap, and alternative solutions. How can we make networking in OpenStack robust, high performance, and fault tolerant? What do typical large scale networks look like and what lessons can we learn from them? Is there an approach to networking we can take that is the same with a handful of servers as it is with hundreds of racks?
This session will consist of three, 10-minute lightning talks from OpenStack Networking experts followed by 10 minutes of Q&A. A summary of each lightning talk is provided below:
Quantum Plugin and Extensions for Cloud Applications
Mohammad Banikazemi, IBM
This presentation covers challenges in developing a Quantum plugin for Meridian, a service-level network model that provides high level connectivity and policy abstractions for cloud applications. Although the current Meridian implementation leverages OpenFlow, the services it defines are amenable to a variety of implementations including overlay networks. The Meridian architecture and implementation is described briefly. Key challenges in the design and implementation, including orchestration of network tasks on large networks, efficient handling of dynamic updates to virtual networks are then discussed. Next, the Quantum plugin for Meridian, which maps the basic Quantum constructs to the Meridian network model is presented. Finally, a set of extensions to the base Quantum API that allow the entire set of Meridian features to be exploited is described. These features include support for flexible and dynamic insertion of middle boxes. The presentation identifies some strengths as well as some weaknesses of the current Quantum design.
SDN deployment using Floodlight with Openstack Quantum and Openvswitch
Damian Igbe
The SDN revolution has started and while the shape of things to come is still unclear, this is the best time to delve into the technicalities of SDN. It is only those who really understand this buzzword now that will shape the future of network virtualization. This paper aims to explore SDN using Floodlight Openflow controller along with Quantum and Openvswitch. The emphasis will be on Floodlight as the Openflow controller based on experience experimenting and integrating to Quantum and Openvswitch. A hands-on demo will be provided to highlight the essential configuration steps required to get Floodlight up and running and talking to Openvswitch and Quantum in a multi-tenancy deployment.
Extending Quantum and the OVS plugin for physical network orchestration
Andre Pech, Arista Networks
Quantum currently does not allow for the simultaneous configuration of virtual and physical switches as part of tenant network provisioning. When using OVS, the user is required to manually provision the physical network that provides connectivity between VM's in a tenant network.
We've extended Quantum and the Quantum OVS plugin to allow for the registration of plugins ("hardware drivers") to coordinate the physical network based on the topology of the virtual switches, enabling a fully automated deployment of new tenant networks. We've also exposed how the virtual tenant networks map onto the physical network, providing users with increased visibility and better troubleshooting ability.
In this talk, we'll go over the changes we've made within Quantum, do a demo showing the orchestration of the physical network in response to user actions in OpenStack, and discuss what we see as potential future integrations of the physical network infrastructure into Quantum.