Comparing Openstack
Comparing Openstack
Introduction
OpenStack is an essential component of every modern organisation which uses
private cloud infrastructure for better economics, improved security and a
higher level of flexibility. Over the last few years, it has evolved and become a
de-facto standard for implementing cloud computing platforms. It is one of the
three most active open source projects in the world with 1,003 developers from
188 organisations involved in the Ussuri release development. According to
451 Research’s Market Monitor from September 2019, it’s combined market
size worldwide is $7.7BN.
Red Hat’s mission is to provide open source technologies for enterprises. Its
business model is based on licensing their proprietary software components
added on the top of existing open source technologies, such as Linux or
OpenStack, and providing commercial support for them. Red Hat is one of the
main contributors to OpenStack, primarily focusing on exploring incubating
projects. The company provides a rich portfolio of enterprise software products.
Those include RHEL (Red Hat Enterprise Linux), Red Hat OpenStack Platform, Red
Hat VIrtualization, Red Hat OpenShift, Red Hat Ceph Storage, Red Hat Satellite
and Red Hat Ansible Automation Platform.
About Canonical
2
Canonical’s Charmed OpenStack
Distribution comparison
In the following section, we will perform a detailed analysis of the differences
between Red Hat OpenStack Platform and Canonical’s Charmed OpenStack. We
will evaluate how each distribution influences and impacts the business value
OpenStack can deliver to an enterprise. For a more detailed comparison, refer to
the appendix at the end of this whitepaper.
Licensing
One such OpenStack distribution is Red Hat OpenStack Platform. Although the
underlying engine, TripleO, is open source and hosted under the governance of
the OpenStack Foundation, Red Hat packages their OpenStack Platform with a
proprietary installer. This installer, called Red Hat OpenStack Platform Director,
provides an additional layer on top of TripleO and requires a subscription to be
purchased. This stack is shown in Fig. 1. Although Red Hat provides a trial period
during which organisations can try Red Hat OpenStack Platform for free, the
entire platform cannot be used after the trial period expires. This means that
organisations have to cover all of their environments, including production,
development and staging, with the subscription. This subscription, as we will
show in the next section, is expensive.
3
Red Hat OpenStack
Platform Director
OpenStack
Charms
TripleO
Subscription required
Fig.1. Red Hat OpenStack Platform and Canonical’s Charmed OpenStack software stack.
Pricing model
Although not all OpenStack vendors require a subscription for usage of their
distribution, all of them rely on the subscription to provide commercial support.
This is where the pricing becomes a decisive factor. Every organisation, regardless
of size, wants to receive the best possible services with the best economics.
Red Hat only supports OpenStack deployments made through the Red Hat
OpenStack Platform Director. Various types of subscriptions, including Red Hat
OpenStack Platform, Red Hat OpenStack Platform (without a guest OS) and Red
Hat Ceph Storage are available, depending on the node destination. Although
Red Hat’s pricing for OpenStack is not publicly available, an estimated price of a
Red Hat OpenStack Platform subscription unit is 5,000 USD per socket-pair. In
addition, all nodes in the cluster have to be covered with a RHEL subscription at
the price of 1,299 USD per socket-pair.
A per socket-pair pricing model means that a separate subscription is needed for
every 2 CPUs (central processing units). As a result, the cost increases as the
number of workloads grow, making this an expensive option when scaling. In turn,
Canonical uses a per host pricing model, regardless of the number of CPUs inside
of the host, under the Ubuntu Advantage for Infrastructure (UA-I) subscription.
This means that if the number of workloads grows, Canonical’s customers can just
switch to more powerful hardware instead of purchasing additional subscriptions,
as adding and removing hardware is fully automated, including bare metal
provisioning. They can also plan the deployment carefully without having to
purchase more subscriptions than they actually need during the initial roll-out of
the cloud.
4
The difference between per socket-pair and per host pricing model, based on a
compute node example, is shown in Fig. 2:
Canonical: 1 subscription
(1,500 USD)
Red Hat: 1 subscription
(˜6,300 USD)
CPU CPU
Canonical: 2 subscription
(3,000 USD)
Red Hat: 2 subscription
(˜12,600 USD)
CPU CPU
Canonical: 2 subscription
CPU CPU CPU CPU (3,000 USD)
Red Hat: 8 subscription
(˜50,400 USD)
CPU CPU CPU CPU
The biggest problem with the per socket-pair pricing model is the lack of
transparency and control over the costs compared to the per host pricing model.
It is much easier to predict the growth rate of physical hosts than the growth rate
of socket-pairs. This is especially true in environments where virtual machines
have dedicated CPUs assigned, such as NFVI (network function virtualisation
infrastructure) scenarios. As a result, a per socket-pair pricing model can quickly
result in an ever-inflating TCO. Per host pricing allows enterprises to thoroughly
estimate annual operational costs and enables full transparency when budgeting
for future changes.
5
Ubuntu Advantage for Infrastructure (UA-I) is an enterprise subscription for
Ubuntu, covering all aspects of the infrastructure, including Ubuntu Server,
OpenStack, Ceph and Kubernetes. It includes commercial support, production-
grade SLAs (service level agreements), Kernel Livepatch Service and up to 10 years
of security patches. Ubuntu Advantage for Infrastructure is available in three
variants: Essential, Standard and Advanced, each of which offer different levels of
support. The annual cost of an Advanced subscription is 1,500 USD per host.
Consulting services
Red Hat’s approach towards consulting for OpenStack is to engage with potential
customers through an SDF (solution delivery framework) which consists of three
stages: discover, design and deploy. Across these stages, Red Hat consultants
provide products, services and custom engagements to help their customers
design and build their private cloud infrastructure based on the Red Hat
OpenStack Platform. The pricing is calculated individually based on consulting
units which are credits that can be redeemed for consulting services. This means
that the customer pays for work hours of Red Hat’s consultants instead of the
actual deliverables. This can lead to budgeting challenges as customers often do
not know how many consulting units are needed to meet their requirements. An
estimated price of Red Hat OpenStack Platform delivery is 10,000 USD per week
of engagement with a Red Hat consultant.
Private Cloud Build (PCB) is a consultancy package for Charmed OpenStack which
includes cloud delivery based on reference architecture and certified hardware
options. Additional services and add-ons, such as on-site design workshops, custom
architecture and non-standard SDN, and storage solutions are available under the
PCB Plus package. Using PCB leverages the Infrastructure-as-Code (IaC) approach,
making the deployments repeatable and drastically reduces the time required to
deploy OpenStack. PCB is available at a fixed price and costs 75,000 USD.
6
Support services
Managed services
This is exactly what Canonical offers for its Managed OpenStack service. Although
the cloud is deployed by Canonical’s field engineers as a part of the PCB service
and is maintained by Canonical’s operations team, customers can always request a
handover. This is also something that Canonical can help with by offering
comprehensive training services for customer’s operations team members. In
turn, Red Hat does not offer managed services leaving their customers on their
own with post-deployment maintenance. Moreover, as the cumulative cost of
UA-I and Managed OpenStack (5,775 USD per host) is lower than the cumulative
cost of Red Hat subscription for OpenStack (~6,300 USD per socket-pair), in fact
Canonical’s customers can get fully managed private cloud service at a better
price than Red Hat’s OpenStack customers.
7
OpenStack maintenance
Although fully managed services for OpenStack are a tempting solution for
many organisations, due to various reasons they are not always an option.
While most organisations rely on vendor’s consulting services to deploy
OpenStack, they usually maintain the cloud by themselves post-deployment.
How easy this maintenance is becomes a decisive factor. As OpenStack
maintenance includes lifecycle management, configuration, daily operations
and integration with external components, a lot of resources are required to
operate it. In most cases, a dedicated team is hired to maintain the cloud.
An organisation then needs to evaluate how many employees are required
which will impact its ongoing OpEx costs.
While Red Hat OpenStack Platform Director provides full automation of the initial
OpenStack deployment process, its support for ongoing OpenStack maintenance
is very limited. Most of the tasks, such as scaling out the cluster, have to be either
run manually or in a semi-automated way. In turn, Charmed OpenStack expands
automation to daily OpenStack maintenance by using a model-driven approach
based on OpenStack Charms. This is possible by abstracting the entire complexity
behind OpenStack and exposing the configuration of the cloud in the form of a
model which can be easily updated by the operations team. As a result,
organisations have to spend less time performing daily maintenance tasks. This
means that fewer FTEs (full-time equivalents) are required to operate the cloud
which results in significant cost savings.
Release cadence
Historically, Red Hat’s release cadence was not predictable at all. Although
starting with OpenStack Platform 16, Red Hat now commits to release new
versions of their OpenStack distribution on a regular basis, support for the latest
upstream OpenStack version is still coming with a few months delay. This is
because the release cadence of Red Hat OpenStack Platform follows the same
conservative approach that the company applies for RHEL; striving to ensure
stability over anything else.
8
OpenStack Y LTS
OpenStack Y
OpenStack X
OpenStack W
OpenStack Victoria
OpenStack Ussuri
OpenStack Train
OpenStack Stein
OpenStack Rocky
OpenStack Queens
Tech preview
Ubuntu LTS release Standard Support
Matching OpenStack release support
Extended support for customers
Extended support maintenance
OpenStack upgrades
Software upgrades are an essential part of the maintenance process. The same
practice should apply to OpenStack. However, OpenStack upgrades are known to
be very complex. This is because OpenStack consists of various interconnected
components which have to work together to provide the service to users. Thus,
the upgrade procedure has to be executed very carefully as a failure of a single
step results in a failure of the entire process. As a result, many OpenStack vendors
have never supported OpenStack upgrades, forcing their customers to re-deploy.
Re-deployments usually translate to downtime and additional work - scenarios
that organisations usually want to avoid.
Historically, Red Hat was not supporting OpenStack upgrades at all. This was
partially due to the fact that they have been constantly changing the underlying
software stack for OpenStack installation. However, starting from Red Hat
OpenStack Platform 16 upgrades are now supported. Red Hat provides a manual
procedure which users can use to upgrade their cloud. This is not ideal, however,
as the procedure is very complex and time-consuming. Moreover, its complexity
increases as the size of the cloud grows. Not to mention very complex
architectures such as those used by telcos, for example.
Canonical was one of the first OpenStack vendors offering full support for
OpenStack upgrades. Canonical’s Charmed OpenStack can be upgraded in a fully
automated way. The operations team only has to initialise the upgrade process
manually and the rest of the work is performed by OpenStack Charms. This is
again possible by abstracting the entire complexity behind OpenStack and hiding
it from the operations team. Fully automated upgrades allow organisations to
stay up to date while meeting their availability goals. All of that results in less
work required to maintain the cloud and reduced operational cost.
9
Multi-cloud and hybrid clouds
Red Hat OpenStack Platform does not provide any integration capabilities with
public clouds out of the box. It is simply not designed for this purpose. As for
containers, Red Hat has been promoting their OpenShift PaaS (platform-as-a-
service) platform which uses Kubernetes underneath, but it is proprietary and
available per an additional license. This again leads to vendor lock-in and increased
operational costs as customers have to purchase a subscription for another
infrastructure component and are fully dependent on the vendor.
Last but not least, security plays an important role in every organisation. Red
Hat’s approach towards security has always been based on relying on mature
versions of the kernel and software. Being very conservative with regards to new
RHEL releases and adopting new versions of OpenStack with a significant delay,
Red Hat has prioritised stability over new features. But is stability the only
guarantor of security? Not really and Ubuntu Server is proof of that.
Ubuntu Advantage for Infrastructure customers also receive access to the Kernel
Livepatch Service which allows applying kernel patches on the fly, without
rebooting the operating system. This results in improved stability and resiliency
as OpenStack services are not affected. Finally, the entire Charmed OpenStack
can be integrated with Canonical’s Landscape service which provides the ability to
perform security audits and generate compliance reports. It also provides
traditional operating system administration capabilities for bare-metal machines,
virtual machines and containers.
10
Conclusion
Although both Red Hat OpenStack Platform and Canonical’s Charmed OpenStack
are production-grade commercially supported OpenStack distributions, they
differ substantially. Those differences result from distinct missions and business
models of the companies behind them. While Red Hat provides an OpenStack
distribution for enterprises, Canonical’s mission is to deliver an OpenStack
distribution that is deployable, maintainable and upgradable economically. At the
same time, Canonical provides consulting, support and fully managed services for
OpenStack, simplifies its deployments and operations, enables fully automated
upgrades and ensures a higher level of security.
11
Appendix
The following tables summarises the main differences between Red Hat OpenStack Platform and Canonical’s
Charmed OpenStack:
Support pricing Per socket-pair (~6,300 USD) Per host (1,500 USD)
Availability of consulting services Yes, based on consulting units Yes, at a fixed price
(~10,000 USD per week) (75,000 USD)
Supported SDN platforms OVN, OVS, Juniper Contrail, OVN, OVS, Juniper Contrail,
Cisco ACI, OpenDaylight Cisco ACI, Nuage
12
Learn more
© Canonical Limited 2020. Ubuntu, Kubuntu, Canonical and their associated logos are the registered trademarks
of Canonical Ltd. All other trademarks are the properties of their respective owners. Any information referred
to in this document may change without notice and Canonical will not be held responsible for any such changes.
Canonical Limited, Registered in England and Wales, Company number 110334C Registered Office:
12-14 Finch Road, Douglas, Isle of Man, IM99 1TT VAT Registration: GB 003 2322 47