Women in HPC

10.20.15 -

Access’ Elizabeth Murray spoke with Toni Collis, applications consultant in HPC Research and Industry at Edinburgh Parallel Computing Centre (EPCC) at the University of Edinburgh, about her work with Women in HPC, a network dedicated to women in the field.

What do you do in your role at EPCC?

My role involves working with academics and industry partners to develop and improve software for supercomputers. I specialize in software used for molecular simulation and stencil-based codes, but have been known to work on a wide variety of topics. My work often involves working with academics with no prior software engineering or HPC training to help them improve their results so that they can make new scientific discoveries, which might not have been possible otherwise. I also take a similar role as part of the team that supports scientists using the UK National Supercomputing Facility, ARCHER. I run ARCHER training courses for academics using HPC, and I also teach part of EPCC’s MSc on High-Performance Computing and Data Science.

You founded Women in HPC. Why is your movement and others like it so important?

At Edinburgh, I am very lucky to work in a group that has a significant percentage of women, varying from 25 percent to 35 percent over the four years since I have been here. As I realized that EPCC is unusual in having such a high proportion of women, I looked around to see what was being done to gather data and address this both in the UK and internationally. I was surprised that nothing was being done to attract women, or even to simply record the proportion of the community that is female in HPC. The more “traditional” subjects such as physics and computer science, which have similar gender imbalance issues, are collecting data to understand the problem as well as developing initiatives to engage women in the subjects from primary school to the highest levels of academia and management. And I believe our field has the unique opportunity to come together with these more traditional subject areas, from natural sciences to humanities, to build an expansive collection of data and strong recruitment programs.

Initially Women in HPC was supposed to be a UK initiative, bringing together UK HPC women to facilitate these women in building a network of female peers. However, I was pleasantly surprised that the majority of the interest in the group was actually from outside the UK. Women in HPC has become a truly international initiative.

Why is it important to collect data on gender discrepancies in HPC?

There are many reasons for identifying and quantifying the issues affecting the proportion of women engaging with HPC. First, we need to know the current picture of HPC. We can gain so much understanding from simply learning what the current gender mix is: where women are currently working in HPC, are they spread evenly across the fields and sub-disciplines, and if not, why? But by identifying the current demographics of the HPC community, we can give clear advice as to what we should be doing, rather than just always assuming it is not our problem to solve.

Several conferences and HPC services have now shared data with Women in HPC, and it looks like women make up about 17 percent of the international HPC community. As I looked around at conferences I started to realize that women may represent as little as 5 percent of the HPC community, in particular niche areas. This both inhibits women from joining the community and limits the scientific output of any group. However, we are in a very early period of data collection so the picture may continue to develop, and we still do not have a good understanding of whether this is representative of the fields that produce HPC users, technicians, and academics. Our next challenge is to improve the proportion of women. The obvious activities here include helping women to network and meet potential mentors, with the additional aim of monitoring the effectiveness of these activities.

How are you going about collecting and understanding the data?

I am working with Lorna Rivera from I-STEM at the University of Illinois at Urbana-Champaign and Lizanne DeStefano from CEISMC at Georgia Tech to help understand why women are under-represented in HPC. By looking at the participants of HPC training programs, we hope to gain an understanding of why some women engage in HPC and others do not. And if there are significant gender differences in the way training is received, how self-assessment by men and women in HPC may vary, and therefore impact recruitment and selection activities.

As far as we know we are the first group to look at this for the HPC community; and although we are just in the earliest stages of collaboration, we hope that the information will prove valuable in helping decision makers develop best practice to broaden the diversity of the HPC community. For example, if we find that there is a significant gender difference in the way women and men present their abilities in HPC, this needs to be considered in recruitment and selection for activities such as funded places at education programs and conferences.

What things can NCSA and EPCC be doing now to recruit more women into HPC?

Unconscious or implicit bias is now a well-known issue that affects us all, particularly during recruitment, with people less inclined to recruit staff that are not “like them” in some form or other. Women are also just as likely to be biased against women as their male colleagues and we shouldn’t be afraid of admitting that. Being aware of your biases is one of the most important steps, and in recruitment we should be encouraging everyone to address this, particularly as there have been multiple studies showing that women are less likely to be offered jobs and mentoring opportunities, along with being given lower starting salaries than men.

I think one of the most important things to look at is the way we recruit future staff and advertise positions. One of the most common problems facing recruiters is getting women to even apply for a job in HPC. EPCC often recruits staff with the potential to develop key skills, rather than already having those skills. This has meant we are able to recruit more women and retain a high proportion of staff, both male and female.

Also, we should not be afraid of accepting that there are, on average, gender differences in the way we think and behave; therefore, for many women working in an environment where they may only interact with one or two women on a daily basis, providing them with the opportunity to build relationships with female peers from other institutions is so important. This is why Women in HPC now runs three workshops per year, bringing together women (and a few men) from the greater HPC community. One of the most fascinating moments for me was at our SC14 workshop hearing a male colleague of mine, who is incredibly supportive of what Women in HPC is trying to achieve, admitting that he felt uncomfortable to walk into a room full of women! I had forgotten that earlier in my career I have often felt uncomfortable walking into a room full of men.

What was your impression of NCSA during your recent work visit with I-STEM?

My overwhelming impression on visiting I-STEM and NCSA is that everyone is incredibly welcoming, friendly, supportive, and interested in what I am doing with the Women in HPC initiative. There is clearly a lot of interest in improving diversity in HPC, and this was a hot topic of discussion while I was in Illinois. I was lucky to visit during the Blue Waters Petascale Institute involving Blue Waters interns and XSEDE Scholars. It was fantastic to see the novel approaches to teaching HPC and in particular the evaluation process that took place. Of particular interest was the use of focus groups, primarily focus groups where the participants were all from a “minority” background, and the change in attitude of the participants. As a lover of HPC machines and technology I was also very lucky to be taken on a tour of both Blue Waters and Donna Cox’s visualization lab at NCSA, which was fantastic! I felt like a small child in a toy shop!

Looking to the future, what are you most excited about when it comes to HPC, in particular Women in HPC?

Women in HPC is now growing faster by the day as more people find out about what we are hoping to achieve. Our workshop at SC15 in Austin is likely to be the biggest Women in HPC event so far, bringing together women from around the world to discuss how we can improve the representation of women, and also providing early career women with the opportunity to present their work at the largest HPC conference in the world.

The growing participation in the Women in HPC initiative is also enabling us to develop an international network. I am particularly excited with our program to develop new partners and chapters internationally, having singed our first memorandum of understanding with Compute Canada in July this year. Having partners will allow groups to run local Women in HPC events that are special to their community, providing training where needed and evaluating practices for their effectiveness. As the initiative grows we hope to build a truly international organizing and steering committee to help diversify our activities beyond, improving what we do and opening up the community to an even greater proportion of women.