Part one of a two part interview series with tinkering whizz-kid and OneKloud CTO.
What’s up Xavier, can you introduce yourself?
I compiled my first Linux Kernel at eight years old. As a kid I used to disassemble just about everything I could get my hands on. At ten years old I modified the electronics of my first walkie-talkies to try to tune into the police frequency. Then I started coding, the first thing I did was was to clone my mom's cell-phone sim card so I could use her phone without her knowing. My second project was to clone the television’s satellite card so I could watch cable on my computer… I’ve always been into tinkering, fiddling. My studies were math/electronic oriented in high school and then I went on to an engineering school and then got a Masters 1 degree from the Dominican University of California, University in San Rafael, California. Today my whole house is totally controlled by computers, from the front door to the lighting and everything in between. I’m also a massive fan of raspberry pi and arduino.
I am French / Spanish, I was born in France to Spanish parents, so all my summer holidays were in Spain, on my mother’s side, Madrid and my father’s, Barcelona. I speak spanish fluently. I am married and have two kids, 3 and 4 years old.
Can you tell us about coming up with the idea for OneKloud?
In June 2014, I was appointed CTO of a American-Canadian company called SociableLabs, I had 20 developers in Montreal and 10 in San Françisco. The company’s AWS budget was $60,000. My first mission was to reduce this spend by negotiating reserved instances with the Amazon sales team. Six months later I realised that my developers weren’t using the instances I’d negotiated, but instead they were just using whatever they felt like. Not long after that, an Amazon sales guy told me that they were going to block our production account because the keys used to launch these Amazon servers (access codes, password keys) had been leaked on Pastebin —a tool for sharing pieces of code in public or private. One of my interns in Montreal had shared a few pieces of code there with keys in it. A bot has found these keys exposing us to DDOS attacks or Bitcoin Mining. This situation cost us $50,000 in one day, let’s not forget the annual budget was $60k! Then a loss of turnover cost $100k, because Amazon cut our key, so we had to rollout all the keys in our code and deploy all the production and staging environment. And add on the time it took to find out who had made the stupid mistake!
So I tried to find a solution to prevent this happening again, with products like Cloudability ou Cloudchekr. These were good. They allowed me to have a dashboard showing what had been spent and displayed my different accounts —but this, after the money was already spent.
The second scenario that lead to the creation of OneKloud: Same company. One day, one of my developers in San Francisco who I had working with on big data processes—an amazon tool called AWS Elastic Mapreduce—made a mistake he shouldn’t even have even been able to. Basically we run AWS EMR (ElasticMapReduce) processes on 20 machines to process large data files for 10 minutes then turn them off. You get billed by the minute and it’s very expensive. But it’s a way to process data extremely rapidly. The problem was that this developer set up 200 (not 20) of these Mapreduces and went off to take a leak, coming out of the bathroom his colleagues grabbed him for Friday afternoon happy hour, hey it was 4pm. A couple of beers, a few rounds of ping pong … it’s already 6pm so he closes his laptop and leaves.
The result? Monday morning and a 10min process had turned into 80 (redundant) hours with 200 machines and not 20. The cost? $140,000. The company's cloud budget was $3 million per year. On Monday I spent my day looking for the developer responsible from my team of 50 and trying to piece together what had happened.
Following that, I did some CTO and business consulting and mentoring. I saw a lot of companies and talked about my idea of finding a solution to control cloud budgets. I found that people were really interested in and receptive to my idea and often had their own stories about dealing with the cost control aspect of AWS. I’ve seen companies recruit 3 engineers, that’s 3 salaries at $150k a year, to sort out this kind of capacity management problem.
Because I don’t like the consultant side of things —I like technical challenges but not selling missions, I told Eric Didier about my idea, alongside the stories above. Then I told him, either we start our own company, or go to Google.
What are some of the biggest challenges for companies working with IaaS, from your point of view as a CTO?
A CTO has the role of managing the technical teams. They are the technical captain of the technical boat. With the massive increase in Iaas and Saas usage this role has changed, we now spend our time sorting through the invoices of the different tools we use; and auditing these tools, in terms of security, that were previously inside the company and are now hosted at by a third party. The challenge is to modernize and innovate, while using third-party services whose development cycle is out of our control.
And what about companies in the process of migrating to the cloud, how do expectations meet reality?
Keeping one’s own infrastructure is becoming increasingly expensive and because this isn’t companies’ key expertise, it distracts them from fully concentrating on their offer. Using the cloud accelerates development and limits maintenance.
Mass migration is already underway with private clouds, especially since new container technology such as docker lets code run everywhere. Infrastructure has become a commodity. Differentiation can be made in the performance of the infrastructure and the price. Migration will become easier in the future software environment, both intercloud and migration to a new environment. Docker has a special program for this and invests a lot in container based legacy app migration. See this link for more information.