VMware VS Canonical – Which Openstack is for You? A Technical Comparison
For anybody that has been following the Openstack project since its inception back in 2010, it has become evident that the community interest has gained significant momentum recently. This is down to a number of reasons such as the product maturity reaching a point where the original mission statements of the project are now a reality in providing an established, viable and effective solution. Also the number and variety of organisations contributing and backing the project, has improved the quality and depth of development and its evolution. A colleague and I attended the Openstack summit in Paris where it was clear Openstack in gaining more and more interest, with attendees listed as increasing by 75% over the last 3 summits with people from all over the world attending.
Openstack was originally established as a joint project between Rackspace and NASA to provide a free open source IAAS (Infrastructure As A Service) cloud computing platform. Since then there are now over 200 companies officially contributing including some of the big names from the IT world such as Canonical, IBM, Dell, Cisco, EMC, VMware … the list goes on. Now that there are numerous players realising the benefits of the Openstack architecture, there are several different offerings from separate vendors on how best to evaluate/deploy/manage an openstack environment. But which one is right for you … ?
The main challenges for anybody considering evaluating Openstack are:
- A steep learning curve of the various Openstack components and their interaction and dependencies, as well as the supporting technologies required such as back end databases and message queues, storage presentation such as CEPH.
- Deployment, management and operational procedures as it is a fundamental shift in thinking from the traditional model of providing compute resource.
Here at Gyrocom we have been following Openstack with great interest and have spent time evaluating some of the different approaches and methodologies out there, and developed key business partnerships based on our experiences. The two main players that we have aligned ourselves with are Canonical and more recently VMware. Both of these vendors have unique offerings that provide excellent tool sets to facilitate deploying and managing an Openstack cloud.
Canonical the company behind Ubuntu is and has been intrinsically linked to Openstack with the six monthly release schedule of their operating system being aligned with the release cycle of each new iteration of the Openstack project. Canonical spend a lot of time and research on OIL (Openstack interoperability Labs) to verify features and function of various technologies/vendors as relating to their reference architecture for Openstack to ensure reliable stable implementations. It’s no coincidence that approximately 75% of cloud based Linux deployments use Ubuntu !
Canonical provide several free open source tools that can be used together to provide a really powerful and flexible solution. These tools are:
MAAS (Metal As A Service) – A MAAS server manages the discovery and deployment of bare metal servers (can be physical or virtual). This allows for servers to be ‘commissioned’ within MAAS ready for deployment. The system allows for a very granular control of allocation of systems by using ‘tagging’ to denote the usage of a particular system based on available resources such as CPU/RAM or hardware architecture. MAAS then ‘pushes’ and configures an operating system to the commissioned nodes ready for operation.
JUJU – This is an orchestration and management product that integrates with MAAS to facilitate the deployment of applications to machines. Juju integrates with the MAAS server to allow applications/services to be installed and configured at the time of deployment through the use of Juju ‘Charms’. These ‘Charms’ define are used to define the required packages needed on a particular system/s and the associated attributes for configuration files. These Charms are also capable of configuring and managing the connectivity to other charms to allow complex relationships to be built in a hierarchical fashion. There are both Canonical verified charms available as well as those created by the community that can deploy complete solutions such as an Openstack cloud, simply from a template ‘bundle’ .
Landscape – Another product in the Canonical portfolio is a subscription service called ‘Landscape’ that acts as graphical management front end to your Ubuntu estate, for monitoring, patching deployment and plugs into MAAS/Juju/Openstack.
These three products provide a powerful and end to end solution from bare metal systems to a fully configured solution, and management.
Canonical also offer a ‘Bootstack’ solution that allows a faster path to innovation through Canonical themselves designing, deploying and managing a cloud on behalf of a customer, with the option of the customer taking over ownership once they feel confident to do so.
Gyrocom were very fortunate in that we were invited as the only VMware UK partner to participate in the VMware VIO (VMware Integrated Openstack) Beta program back in January 2015 before it’s general release in March this year. We have been a VMware partner for a number of years historically primarily in the server virtualisation arena, however we have also become one of the first partners to become certified in the VMware SDN (Software Defined Network) solution VMware NSX. A VIO solution is deployed in combination with NSX to provide a truly unique offering.
Historically around 90% of the Openstack deployments in the wild have been implemented with KVM as the hypervisor of choice to spawn cloud instances, with VMware gradually being introduced more recently in specific use cases. VIO however uses Vsphere ESX exclusively to provide both the infrastructure and compute resources for an Openstack cloud. This is a huge step forward for a lot of our customers whom only use VMware internally and haven’t had the time or skills in house to investigate Openstack. By using tools and technologies that customers are already familiar with provides a massive advantage to pre-existing Openstack options available.
In a similar manner to Canonical, VMware have removed a number of the complexities of deploying an Openstack cloud, by shielding the dependencies and interoperability elements of the deployment from the end user. Both NSX and VIO are deployed via an OVF template on to existing ESX hosts. The Vsphere web console via a Virtual Centre server is then used for all initial configuration/troubleshooting activities through the use of separate plugins for both NSX and VIO. This provides a single interface to interact with presented in a format already familiar to customers.
Once VIO has been installed and is running, all administration of the Openstack environment is completed from the standard Openstack interfaces such as the Horizon Web console, API and CLI. The changes made via the ‘Openstack’ interfaces are then implemented at the back end within ESX and NSX such as the creation/modification of new networks and spawning of cloud workloads as virtual machines.
The VIO OVF file deploys the VIO virtual appliance which consists of a management server and a virtual machine template. After answering a dozen questions about the infrastructure topology and configuration, the management server clones the template (the template incidentally is an Ubuntu image – inline with VMware’s own internal Openstack cloud!) to produce 13 virtual machines that create the Openstack cloud infrastructure.
One of the most attractive features of the VIO deployment is that HA/Resilience is included by default in the deployment; for example three MySQL instances are deployed with replication between each cluster node, also two ‘Cloud controllers’ are deployed with an additional pair of load balancers at the front end to remove a single point of failure. The technologies used to provide this enterprise grade resiliency are achieved using the same technologies recognised in the Openstack community such as Galera replication for MySQL and HAProxy for load balancing. So not only is the Openstack infrastructure a resilient deployment in its own right, there is also the established Enterprise features associated with VMware such as HA clusters, Vmotion, DRS etc …
Another feature of VIO is that is FREE to use for any VSphere Enterprise Plus customer and works hand in hand with NSX to provide a complete solution. Existing customers running OpenStack on VMware in their production environments such as: Nike; Adobe and Wells Fargo, have all professed their orgainsation’s enthusiasm for the potential heightened security benefits the pairing of these platforms could bring*. There are several blogs on our site discussing NSX if you require further information, please take the time to check them out.
Canonical and VMware both provide a simplified process to the highlighted challenges of deploying and managing an Openstack cloud, by streamlining and reducing the complexities associated with a successful deployment, whilst also providing a supportable model for the enterprise with reference architectures.
Both VMware and Canonical have adopted the Kilo revision of Openstack as their stable supported version of code and both vendors have kept their releases as close as possible to the mainstream ‘trunk’ releases keeping development and community driven content as potentially viable solutions rather than bespoke implementations.
Both of the different approaches discussed above offer distinctive benefits from deploying an Openstack cloud from scratch on a node by node basis, with a clear emphasis on the whom the target audience requirements either a infinitely configurable fluid implementation or a turnkey solution. So which ever approach appeals to you, once thing is clear; it’s a lot easier to deploy an Openstack Cloud than ever before …
Pre-sales Engineer at Gyrocom.