My name is Quentin and I am French guy from France. I have always been fascinated by aerodynamics. My first ambition as far as I remember was to become a Formula 1 aerodynamicist because I was amazed by the level of technology associated with this field (that make them stick to the road). I decided to study engineering in the heart of European aeronautics and aerospace (ISAE: Aerospace institute in Toulouse) because I told myself that the rules to make a plane fly were the same as to make a Formula 1 vehicle stick. During my education, I stumbled upon CFD and, as a profound computer geek/nerd, I fell completely in love. I was finally able to combine both of my worlds. I did a lot of structured meshing using ICEM and a lot of CFD using ANSYS Fluent. I continued my education here in Canada in Montréal and worked in a university lab for Bombardier (IDEA Chair: Integrated Design toward Efficient Aircraft at Polytechnique Montréal) where I continued my passion for perfect meshes on full aircraft configurations. My first position after my education was for an ANSYS partner, SimuTech Group, as a CFD Engineer doing consulting projects while providing support and training for ANSYS. I learned a lot and worked with a great team here in Montreal. I quickly started playing with large clusters, connecting machines together to get more power, optimizing, scripting, using job scheduler (because sadly I had to share my toys…) and I rapidly realized that most of the users have a very limited knowledge in that matter, or they do not have the time to learn about it. At first I was just giving advice on what to buy and how to configure small servers, but it was taking more and more time. So, with SimuTech we decided that I should pursue this on my own and I started my own company, Quantum HPC and offer complementary services to theirs. For a little more than a year now I propose services and HPC servers for CFD and FEA users and have become the official hardware partner of SimuTech Montréal.
- Location: Québec City, QC, Canada
- Current position(s): IT, Technician, Technical Support, System builder, Developer, Secretary, HR, Accountant, Manager, Boss and employee
- Current computer:
- Personal computer: Intel i7 960, 8Gb DDR3
- Dev server: Intel Core 2 Quad Q6600 (still very good for servers), 8gb of DDR3 Fedora 22 Server
- One word that best describes how you work: do-it-yourself
How do you know Pointwise?
I have tested several meshers, commercial and open-source, and am a follower of the hex mesh religion. I rapidly stumbled upon Pointwise because very few tools in the market offer blocking-based meshing (the other is ICEM I suppose?). I have to admit that I don’t have a lot of experience with Pointwise, but I am a fervent reader of your blog posts and especially “This Week in CFD” because it is always a pleasure to know what is going on in the more-than-discreet CFD scene.
What do you see are the biggest challenges facing CFD in the next 5 years?
Democratization! I know it’s really different in the US, but for what I have seen in Canada, proposing CFD sometimes feels like selling an electric car: it’s more efficient, it has a lot of advantages, and in the end it will cost less money but … “I’m good with what I have.” CFD is a powerful and underutilized tool, but things are changing fast. Along with the democratization, a lot of training has to be done to prevent the misuse of the CFD. Performing CFD in aerospace, I worked a lot on the importance of the boundary layer, the length of the farfield, even the sensitivity analysis on just the turbulent viscosity ratio in the farfield, and everything has a meaning and its own importance. Of course this isn’t true for all applications and it depends on the field, but there is a popular sentence in CFD that goes like: “garbage in, garbage out.” Democratization of CFD (and FEA), a wider range of courses in universities, access to information, and effective hardware and software support will result in a better understanding of this field.
What are you currently working on?
I am currently developing an HPC front-end for job schedulers capable of connecting to every job scheduler available on the market (Torque, PBS Pro, Windows HPC, LSF, OpenLava, Slurm, etc…) which will be available in two versions: a “regular” version for physical servers and a cloud version for clusters and send jobs to AWS, GCE or Azure. We are planning on releasing the self-service SaaS version based on AWS in a couple of months allowing one to easily deploy the required number of servers based on the number of cores and memory demanded along with 3D-acceleration for pre- and post- based either on DCV or VirtualGl. As a day-to-day user of CFD and FEA applications, the goal was to create a really easy-to-use application allowing users to submit jobs to a bare-metal or cloud cluster and have access to output logs and residuals for monitoring simulations.
Apart from development, I support several clients with servers mostly running ANSYS and CD-Adapco tools, using Virtualization, Remote Desktop with DCV and job scheduling, and work closely with my clients to improve their infrastructure.
What project are you most proud of and why?
My company. Being able to do what I love every day while helping clients reach their goals is very gratifying. Every day I am challenged to find new solutions to interesting problems and have the opportunity to test everything and build solutions adapted to my clients needs and improve those solutions based on their feedback.
Concerning CFD, I am particularly proud of the research I did back at university for Bombardier where I performed the complete analysis of a full-aircraft for each angle of attack and yaw angle using several automated routines and scripts which included the generation of a large series of meshes.
Are you reading any interesting technical papers we should know about?
Sadly I do not read a lot of technical papers nowadays. I find myself reading content from w3schools, StackOverflow, GitHub, API documentation, and other forums. I mostly need to learn about what my clients are using as an infrastructure in order to better support them.
What software or tools do you use every day?
- Docker for testing and developing, as well as libvirt and VirtualBox
- ownCloud for storage
- GitHub and Bitbucket for private code and public open-source code repos
- ANSYS and CD-Adapco with clients and for benchmarks
- NICE DCV, VNC, VirtualGL for remote desktop and support
- PBS Pro, Torque, Windows HPC, SLURM for job scheduling
- Freshdesk for support tickets
- OpenStack, AWS and Azure for cluster deployment
- MobaXterm and Putty for SSH
- Wave for finance, Bitrix for CRM and tasks, Outlook, Mailchimp for mailing lists, Google Analytics for SEO, and Excel for invoicing
What does your workspace look like?
My company is based at home for now and I access my servers from everywhere so I can move my workplace as I need. Mobility is the future where we can access our data and work from anywhere and commute only when we really need/want to. I can work from anywhere, but there is no place like home.
What do you do outside the world of CFD?
Starting a company…it takes some time! Outside that, we have recently moved to Québec City and I am trying to enjoy our beautiful Province. I am waiting on the winter for winter sports which includes snowboarding mostly, a little bit of ice skating and sledging, and chilling in the spa in -20C weather (it’s part of the sport!). Apart from that, one of my favorite hobbies is electronics, which is primarily audio equipment. I am a huge fan of high-end audio and especially tube amps which is at the opposite end of the digital world. I repair and build small audio equipment and I play with everything that you can repair.
What is some of the best CFD advice you’ve ever received?
“It’s good enough!” The trap in which most CFD/mesh users fall in, and where far too much time is spent, is our desire to always go deeper, more precise, and with more cells. In the end countless hours are spent for an unnoticeable difference. During a CFD analysis for an aircraft, in order to validate the results obtained using coarser meshes, I generated a 200 million cell grid. It took me weeks to adjust the blocking and generated the mesh in batch on a large-memory node. After a week or so running on 500 cores, I only had a 2% difference when compared with my 50 million cell grid, which itself had less than 1% difference with a 5 million cell grid. When the techniques and settings are appropriate, you can obtain “good enough” results using modest meshes. There is always a balance in CFD to achieve the best results in the required amount of time. This is why support during an analysis is crucial to challenge and discuss a case.
If you had to pick a place to have dinner, where would you go?
Our modest town of Québec City is beautiful and I really recommend The Chateau Frontenac. There is a beautiful veranda overlooking the St. Laurent which is completely frozen in winter and a nice fireplace. Also, the Auberge Saint-Antoine, built on the well-preserved ruins of the old-port, holds some really good jazz nights with very fine dishes. Call me when you get here and I’ll take you.