- Who uses OpenStack?
- Commercial Offerings
- A Great Keynote video
- History and Beginnings
- My Experience
There have been a lot of “analysts” following OpenStack and evaluating it compared to other IaaS cloud management platforms. One of the most no holds barred reports came out a few months back from Gartner… big surprise there. For some reason companies still look to Gartner as some kind of tech bible for technology recommendations. If your company is OK with just being mediocre and is not really looking to be innovative then most of Gartner’s recommendations are probably fine to follow. I’ll drop the Gartner comments here. Below is a direct link to the report and another to a GigaOm article about the report.
Well maybe one more… It seems to me that Gartner should be going out and talking to a lot of the companies that are using OpenStack very successfully instead of fueling the “hate on OpenStack bandwagon”, which from what I can tell are people who have never even used OpenStack. Sure you have to be pretty technical to get OpenStack up and running, but we ARE talking about a cloud management platform… Your traditional IT department that manages Windows machines probably isn’t going to be technical enough to install, configure, and manage OpenStack. A team that is put in charge of evaluating options and building a cloud infrastructure HAS to be multi-disciplined, including software engineers/developers who are comfortable with reading/writing code and understand software architecture along with your traditional IT roles of system administration, networking, and storage.
My advice on analysts, take anything they say with a grain of salt since analysts don’t really USE the products they are making recommendations on.
Let’s cover some of the companies using OpenStack
Cisco – WebEx
- WebEx is a leader in on-demand collaboration and the second largest vendor of SaaS for business applications in the world. Earlier this year, the company placed a strategic bet on OpenStack to build a private cloud platform for many of its mission-critical apps. In collaboration with Mirantis, WebEx was able to successfully traverse the implementation path and is currently onboarding a number of production workloads onto its OpenStack cloud. In this talk, WebEx will share more about the project, rationale behind choosing OpenStack, some of the key road-blocks on the path to production and touch upon the future roadmap.
- Video: Case Study – WebEx Deploys Enterprise Operations Infrastructure with OpenStack
Ebay / PayPal
- eBay: Practical SDN deployment with OpenStack Quantum
- “We’re housing our OpenStack capabilities on dedicated infrastructure in three AT&T data centers today, with locations in Dallas, San Diego and Secaucus, New Jersey.”
- Sandia National Laboratories is a multi-program laboratory managed and operated by Sandia Corporation, a wholly owned subsidiary of Lockheed Martin Corporation, for the U.S. Department of Energy’s National Nuclear Security Administration under contract DE-AC04-94AL85000.
Argonne National Laboratory Magellan Cloud
- “We expected to see a performance penalty because of virtualization; however, the penalty was considerably smaller than we expected for many scientific applications. One of the most surprising findings from the evaluation project was the benefit scientific users derive from direct access to computational resources for their applications.”
- “Upon completion of the evaluation project, we decided to transition the OpenStack system into production to support the computational and infrastructure needs of the DOE Systems Biology Knowledgebase project.”
Cybera’s Space Weather Research Cloud
- “So, after careful investigation and much trepidation, we took the dramatic step of switching to OpenStack as our IaaS cloud platform. Another challenge we faced was dealing with multiple physical sites for our different cloud zones, each of which has different hardware, different operations staff and different policies.”
MercadoLibre – E-commerce
- 20M API requests / minute
- 50k requests / second peak on holidays
- 4Gb bandwidth / second
Some of the OpenStack vendors and service providers
- Dell – http://www.dell.com/OpenStack
- Rackspace – http://www.rackspace.com/cloud/private/
- RedHat – http://www.redhat.com/openstack/
- Canonical – http://www.ubuntu.com/openstack
- Nebula – http://www.nebula.com/
- Piston Cloud – http://www.pistoncloud.com/
- Cisco – http://www.cisco.com/web/solutions/openstack/
- Cloud Scaling – http://www.cloudscaling.com/
- Metacloud – http://www.metacloud.com/
- Rackspace Cloud – http://www.rackspace.com/cloud/public/
- HP Cloud – https://www.hpcloud.com/
- Dreamhost Cloud – http://dreamhost.com/cloud/dreamcompute/
The Spring 2012 OpenStack Summit was the first OpenStack event I ever attended. It was an amazing time and I got to meet and talk with lots of great people. Chris Kemp’s keynote was the best talk of the conference which built lots of excitement in the room (or at least it did for me). Chris Kemp served as CIO and CTO of NASA and since then has gone on to found Nebula. I encourage you to watch it.
Jump to 23:40 in video to start watching. This a very long video, but Kemp’s portion is definitely worth watching.
“We’re OpenStack, we are the ecosystem, and we can do whatever we want to do!”
A little OpenStack history
My experience with OpenStack
I have been using OpenStack for over a year at work. This started out as a very small team of 3 people. We learned how to install, configure, manage, and troubleshoot OpenStack with a little bit of remote help from an OpenStack consultant. Bare in mind, this was over a year ago when the Diablo release just came out and OpenStack documentation was pretty poor back then. Now, OpenStack documentation is much better and it is pretty straight forward for anyone comfortable with Linux to manually install (no automation tools) a multi-node OpenStack cluster and have it configured and running in a couple hours.
The software team I lead at work has been exclusively using OpenStack for over a year to do all of our software development on. The agility that a self service infrastructure provides to software development teams is priceless, and OpenStack delivers this very well. We are a small team in a big enterprise and OpenStack is getting a lot of visibility for the things we are able to accomplish by using it. We are able to develop new software applications or add new features fast because we aren’t held back from obtaining the compute resources we need, we can get them almost instantly. We don’t have to worry about “breaking” a production environment to test a code change because we can spin up development or test environments at will and have all the machines that make up the distributed system automatically provisioned with all software installed and fully configured. That means no administrator or developer/engineer has to spend hours or potentially days doing brain dead/numbing tasks of installing and configuring software. As I wrote previously, we pass in some userdata to the VMs on launch and Puppet (our current automation tool handles the installs and configs). Not only does OpenStack allow the team to have VMs on demand as needed, but it also provides the dynamic infrastructure that allows applications to scale themselves out horizontally as needed. This horizontal scaling is one of the main advantages of cloud because it allows your application to run using only the minimum resources (CPU, RAM, disk, etc) necessary until a spike in traffic/load comes along and a new VM gets created via a monitoring tool making a cloud API call. This new VM/node in the cluster gets provisioned automatically via userdata/Puppet as well. We also have Puppet install a small agent to monitor resource utilization on each VM. This is very easy to do and automate for each VM that is started and you can use any monitoring tool you like… Nagios, Zabbix, Graphite, statsd, Splunk, etc. If the monitoring tool you are using supports role based access control to the data being collected then you can easily create a multi-tenant monitoring tool, which is what I have done with Splunk and we have also done with Zabbix.
Price… it is impossible to compete with OpenStack on price. If you plan to build a cloud at any kind of capacity and scale then buying a licensed cloud product (not naming names) will put you in the poor house buying licenses for even just a few racks of hardware, and that’s before you even buy hardware to fill the racks. Imagine paying for cloud licenses for a whole datacenter containing 100s or 1000s of racks.
One of the greatest benefits of OpenStack is that it is not a “product”, rather it is a stack of components that allow you to tailor a solution that can solve many different needs and use cases for building a cloud. It provides a great API and abstraction layer for many pluggable implementations of hypervisors, block storage, object storage, networking, etc. Since it is not a “product” it does not dictate what automation and monitoring tools you must use, it gives you the choice to choose what works best for you and your teams or to stick with tools you are already familiar with. OpenStack gives you all the freedom and flexibility you could ask for and only requires that the team deploying it knows what they are asking to get from it.
Dream it, build it!