Published at 08.01.2014
Table of Contents
anynines is based on Cloud Foundry, an open source platform as a service framework. Thanks to an installation automation technology called Bosh, one of Cloud Foundry’s advantages is infrastructure agnosticism. This means that there is no tight coupling of anynines to any infrastructure provider.
Cloud Foundry’s early ties to VMware, anynines started on a VMware infrastructure operated by our befriended datacenter Skyway Datacenter GmbH.
A little about the background. Back in the early days, Cloud Foundry has been funded and developed by VMware and hence first versions have been developed against and run on a VMware infrastructure.
So for us at anynines it seemed obvious that VMware will be the best supported infrastructure, so far. In addition, anynines’ approach was based on the lean startup paradigm. So the first experiments were about figuring out how hard it is to get a working Cloud Foundry setup. Which was surprisingly easy! So it was a logical choice to stay close to Cloud Foundry’s reference installation CloudFoundry.com and use VMware, at least in the first place.
However, in the meanwhile Pivotal has been founded and took over the Cloud Foundry development. They also drove forward a support for other infrastructures such as Amazon Web Services (AWS) and OpenStack.
When CloudFoundry.com moved away from VMware to AWS this was a signal that it’s also time for anynines to think about the mid- and longterm infrastructure of choice.
anynines is 100% European. The goal of anynines is to provide a cloud infrastructure that is 100% compliant to European and German privacy laws. It would therefore not fit our expectations to run on an infrastructure operated and/or owned by a US company. We love our american friends but there’s an inherent conflict between European privacy laws and the US patriot act. Unfortunately the eyewashing US – EU safe harbor does not solve this issue.
With AWS off the table, we needed to decide between a hosted VMware vs. a self-hosted OpenStack.
While VMware is a fine piece of technology, OpenStack was the match for the specific requirements of anynines.
We’ll yes, generally offering support for different infrastructures is one thing, switching infrastructure of a running system, is another challenge of a different magnitude. Not too many words of glory before it’s done but – at the time of this writing – we are absolutely sure it’s possible. We’ve done it internally a couple of times. We’ve been running Cloud Foundry test setups on OpenStack nearly since day one of anynines and we also tested the live migration several times.
So with respect to anynines still being in beta and the excellent architecture of Cloud Foundry, an infrastructure migration is feasible, for sure.
A wise man said in 2011: “You don’t get up in the morning and run OpenStack”. For sure, we’ve been through the valley of darkness since Diabolo. But fair enough, things have improved a lot and with more than a year of running OpenStack in the context of Avarteq’s* hosting department Enterprise-Rails.com we can be confident that we know how to run OpenStack as a solid foundation for anynines.
With that behind us we can look at the positive side of things.
One of the non-technical benefits of OpenStack is a much better philosophical match of OpenStack to the values of anynines. Offering transparency is important to us. So with both Cloud Foundry and OpenStack being open source software, you are free to see the software used to operate your applications. OpenStack is the fastest growing open source community on earth, so it’s good to know that there are thousands of high skilled people digging through its code.
The financial side of things has also impact on customers. For anynines, the operational costs of OpenStack are much lower than with VMware due to the absence of license costs.
We’ve been able to assemble the hardware servers exactly according to the specific needs of Cloud Foundry which leads to a better performance / € ratio.
The gain of cost effectiveness allows anynines to come up with a competing pricing from day one.
Running OpenStack on our own is also a gain in productivity. We’ve struggled through a set of infrastructure related issues in the past and with a chain of our anynines Cloud Foundry team, the remote VMware operators and sometimes the remote VMware support guys long communication latencies have eaten up significant parts of our productivity. Today, any infrastructure issue with impact on anynines immediately becomes priority 1 within a single team. Together with the decreased communication latency we’ve perceived a much better mean time to repair during error cases in both test and simulated scenarios.
OpenStack is about to become the industry standard API when it comes to cloud infrastructure. It’s a solid and fast growing community which makes any investment a lasting one.
OpenStack’s multi tenancy capabilities offer much potential for future service provider programs. anynines comes with a standard set of services such as MySQL, PostgreSQL, MongoDB, Redis. Of course there’s demand for more services. With the ability of having isolated networks, compute and storage resources, it will be possible to offer service providers access to the anynines OpenStack infrastructure to offer dedicated services which can then be integrated as anynines services.
OpenStack Swift is OpenStack’s answer to Amazon’s S3. An object based storage which can be interfaced using a simple REST API and scales up multiple petabytes. There’s already a beta implementation that allows you to store your application assets using the anynines Swift service. Having an object store was essential to offer a complete alternative for most applications to existing AWS deployments. With anynines Swift even your assets remain on EU territory both physically and legally.
In case you are a little familiar with OpenStack, the release anynines is based on at the time of this writing is havana.
OpenStack is a win. Leaves the question how it affects you.
The migration will be covered by the anynines engineers and won’t affect user applications.
There are some edge cases, customer applications with hard coded credentials, for example, that might have to be touched after the migration. Corresponding customers will be notified by our support team.
For all other cases the migration should happen without any effort on the customer side.
During the migration there might be small outages which will be announced on our status page, separately. Most likely, outages will mainly be caused by DNS changes.
Another impact you might recognize is that everything becomes a whole lot faster as the new infrastructure comes with more CPU power, more RAM and a 10 GBit/s backnet.
The migration to OpenStack is a large milestone for anynines towards the end of beta. It demonstrates the brilliant architecture of all involved technologies, especially Bosh and Cloud Foundry. Thanks to Pivotal for this fantastic software and help. Last but not least, thanks to all our customers. You are the reason we are working so hard and why we love our work!