I think an update is in order to my original OpenStack blog post . Over the last six months, we’ve seen Icehouse, the software’s ninth release, come out of the gates, along with a very successful summit in Atlanta. So, let’s do a quick recap to see where it stands now.
By the numbers
There’s some really interesting information that has come out of surveys from the Atlanta summit . This is really interesting information on the size of installs, environments and what versions are currently in use. OpenStack components I mentioned in my previous post how OpenStack is written in Python (basically 2.x, but there’s some momentum to rewrite everything to 3.x within the next few major releases).
One of the strengths and challenges of such a large and encompassing cloud management platform is that it does a lot of things. All of these things are split out into modules: compute (“Nova”), object storage (“Swift”), block storage (“Cinder”), networking (“Neutron”), dashboard (“Horizon”), identity service (“Keystone”), image service (“Glance”), telemetry (“Ceilometer”), orchestration (“Heat”) and database (“Trove”). You can easily search online for more information on each of these, so I won’t get into the details here. This modular architecture may allow the community to work in a more agile way, but also comes back to pose challenges, as everything basically still needs to communicate with each other to provide a unified, seamless platform.
Another thing you might notice from the recently published numbers is that Linux rules as the underlying operating system of choice. So, if you want to gain a certain level of respect, and be able to show or prove you know your stuff, you’ll also need to have some advanced skills in Linux. Along with Linux being the leading operating system, KVM is the leading choice for hypervisor, with VMware and Microsoft Hyper-V (on Microsoft Windows) quite far behind on use.
There are several basic ways you can interface with OpenStack that come packaged: dashboard (web browser interface), command-line, Python API and a REST API (for custom interface development). For the most part, I typically lean toward the command-line interface, but will fall back on the dashboard when I’m not sure what command to use.
Setting up OpenStack, even in a controlled environment, can be quite a big task. Fortunately, if you’re looking to get something up relatively quickly, maybe for a proof-of-concept, there are some great tools out there. Tools like Red Hat’s packstack or devstack can get you up and running quickly with OpenStack. Personally, I’m a bit of a Red Hat fan, so I’m partial to following anything it does. If you’re somewhat familiar with Linux, you may know about Fedora (this is free, and more of a Linux desktop version) and CentOS (a free Red Hat Enterprise Linux clone), which are both supported by Red Hat and work with packstack. These automated installers are very useful, but remember that to consider yourself at the master level, you’ll want to install from scratch. I honestly haven’t tried this yet, but it’s just a matter of time before I do, in order to gain the whole experience. I find I learn the most when something doesn’t work as expected or I have to actually troubleshoot or read the documentation, or even install files before I can really grasp how it all works. Is OpenStack “better”? I’ve never been a big fan of trying to pick which technology is better. A few years ago, it seemed to be a debate whether Linux or Microsoft Windows was “better.” I was never interested in these opinion-based debates, rather, I like to use the best fit. In the private cloud space, VMware seems to have the bigger market share. I saw a recent reference on Twitter that compared VMwar e and OpenStack like this: If you have money, use VMware; if you have time, use OpenStack. Now, nothing is that simple, because most organizations have a limited amount of money, and they also have to consider that OpenStack can have integration challenges since it’s not as polished. It can be quite costly when something isn’t done right from the start, and considering that it might be beneficial to have a Python programmer help with any integrations, time isn’t the only factor you need to consider before implementing an OpenStack-based private cloud. It’s honestly hard to imagine OpenStack becoming a major player in the public cloud arena against the likes of Microsoft, Amazon or Google, for example.
Just like Hadoop, quite a few companies were maybe too excited to launch their own distribution. As reality set in, we saw one recent big partnership between Intel and Cloudera. There was also a relatively big acquisition by Red Hat (of eNovance), so there may be other shake-ups including an acquisition of RackSpace (one of the original project founders).
Make no mistake, if you think OpenStack is easy to grasp, you’re not wrong. But mastering it can be a challenge. After all that, if you’re still with me, Pluralsight is launching an introductory course to help you along with your learning experience. There’s nothing like having an experienced professional lead you through your training, and using their own insight to help you succeed. A very interesting resource that was announced at the Atlanta summit is a new publication called Superuser . It’s definitely worth checking out to supplement your learning experience. With the next release, Juno, planned for mid-October, OpenStack continues to move forward. With most technologies, any hint of slowing down in updates can be a sign of weakness, but this doesn’t appear to be a problem with OpenStack. Only time will tell if it’s here to stay.