Thanks to Northeastern CCIS and their generous alumni, I had the amazing experience of attending the LISA15 conference this past month (Nov 2015) where I learned about new and interesting technologies, met industry experts, and participated in a competition to build a shield that would protect astronauts on Mars.
LISA, which stands for Large Installation Systems Administration, is an annual conference dedicated to the improvement of system reliability and improving work culture. Topics ranged from DevOps practices to automated provisioning of OpenStack nodes to testing your company’s infrastructure by intentionally breaking it. This was my kind of conference, and it was unlike any conference I have ever attended in the past.
One of the notable talks I attended was a presentation about how Google does DiRT (Disaster Recovery Testing). DiRT involves intentionally disrupting parts of your product’s infrastructure to determine if your failover and redundancy policies actually work. This was something new to me, and as the presenter, Kripa Krishnan, pointed out “You do all that work to make your product HA, but then you don’t regularly test it?” While the talk contained some technical notes about ways to safely attempt to break your infrastructure, its focus was on the logistics required to effectively run disaster recovery tests, namely the benefits behind DiRT and the challenges you might face from management when trying to run tests are designed to have a “higher than average” potential to bring down your product for end users. Learning about the logistical challenges and requirements to effectively run DiRT testing was by far one of the most interesting topics out of a conference that I thought was going to be purely technical.
Another interesting talk was for the openstack ironic project. This presentation directly correlated to work I have been doing to bring OpenStack to CCIS and addressed one of the pain points that we have experienced: provisioning. One of the amazing things about OpenStack is that almost any computer can act as a node in an openstack cluster. There are few hardware requirements, even desktop machines that might be retired from service can add to the capacity of an OpenStack cluster. However, provisioning devices can be a time intensive process, especially when the install process for OpenStack is as complex as it is. Ironic takes care of this, in an interesting way. Ironic actually interacts with the onboard controller for the desktop or server that you are provisioning, whether that be iDRAC, ILO, vPro, or something else. Ironic will take care of powering on the device, telling it to boot to the network, downloading the linux image to install, installing linux, and then provisioning the desired OpenStack services on that hardware however you choose. Pretty neat and something I will be investigating further.
However, LISA wasn’t just talks, there was also a coding competition to protect a stranded astronaut on Mars from the sun’s deadly radiation by activating a shield with limited batteries. This competition was run by EMC code who provided a central game server that simulated the shield and radiation levels. Our job was to build a server that ran several docker containers. We had to build the images that would support these containers, and the containers would ultimately run a few programs we modified that would communicate with the central game server. It was an amazing experience to participate in this competition, especially because it was scored based on the devops principles we applied to our server and that we started the competition a day late, with only 3 hours to go.
LISA15 was an amazing conference, and as a student, it was even better. LISA16 will be held in Boston, so if you are a student in Boston, look into attending, it is a great learning and networking opportunity!