In this special guest feature from Scientific Computing World, Alison Kennedy discusses why self-help will not be enough to overcome the challenge of XXscale in the HPC industry.
Gradually, the climate is changing in HPC. Women are in demand from employers and the industry, by portraying itself as serving society, is becoming more attractive to women. Typically, for many women, an important criterion in selecting a career is that their work should have societal benefits.
One of the many barriers to achieving Exascale is the limited pool of skilled and trained manpower (literally) in the HPC industry. Well-documented shortages in personnel provide an opportunity (and the impetus) to increase the diversity of the pool of talent in HPC: in particular, to increase the number of women working in the field. Funding agencies, operating in straitened circumstances, are increasingly being asked to justify their investment in HPC by reference to economic and societal impact, leading to a change in the way HPC systems are perceived: from prestigious toys to useful tools.
All these factors are creating a climate in HPC in which women are in more demand.
So far, so good. Increasing the number of women in HPC, sharing pictures of kittens on Twitter…two things that no one is against. Last month, Scientific Computing World reported on some initiatives to encourage more women to take up careers in scientific computing and in HPC — an appropriate piece of timing, given that this is the bicentenary of the birth of Ada Lovelace, widely regarded as having devised the first computable algorithm and thus as the world’s first computer programmer.
And yet, progress is not all plain sailing. Just as for Exascale, XXscaling the industry will require serious discussion about whether incremental or disruptive change is necessary to achieve the hoped-for influx of women into the industry.
I first realized the scale of the challenge facing women in HPC thanks to an unknown American male attendee at the 2013 International Supercomputing Conference. Two female colleagues from EPCC were on booth duty when he came up to ask if the University of Edinburgh (our parent organization) was a ‘girls’ school’. This was, to him, the only plausible explanation of how a High Performance Computing centre could be represented at a major international event by two technically qualified women. (In fact, given our staffing profile and assuming that two staff members were selected at random, the probability of selecting two women was 5.4 per cent and the probability of selecting two men was 57.9 per cent.)
Women in HPC was set up in the UK in 2013 to support collaboration and networking, by bringing together female HPC scientists, researchers, developers, users, and technicians. The initiative also encourages women working in HPC to engage in outreach activities and improve the visibility of inspirational role models. While this networking activity is key, both in raising the profile of women and in giving women working in the industry a chance to interact with others facing the same challenges as themselves, there is a danger that it reinforces the view that it’s all about ‘sisters doing it for themselves’ and that self-help is all that’s needed for women to play an increased role in the HPC industry. It’s not. A far more focused, integrated, and planned approach is needed. Women in HPC can play a significant, though not necessarily a dominant role. But there are some uncomfortable truths which need to be faced before HPC can move forward in the gender-neutral way that we’d like it to.
Women in HPC is so much more than just a numbers’ game. While incremental change is important in improving the gender balance and facilitating the entry of more talented women into the industry, it is only a first step.
A major part of the work of Women in HPC in the immediate future will be focused on understanding the global gender demographics of the HPC community. At the moment, evidence is only anecdotal, but several women have independently said that they felt pressured by male bosses to take up management roles in HPC because of women’s perceived superior ‘soft skills’, while they would prefer to stay in technical roles. Similarly, women appear to be over-represented in areas such as user-support, and it’s not clear whether this is due to conscious decision or unconscious bias. Also, while researchers of both genders indicated to us that they much preferred working in mixed-gender teams rather than single-sex teams, there is anecdotal evidence that it’s the men who get to do the more exciting advanced coding to implement new science and analysis, while the women do the duller but worthy running of the scripts and submission of jobs.
So this raises some interesting questions for the HPC community in general and women in HPC in particular. Should the focus be on attracting more women into a profession where there is a growing demand for skills and expertise, even if there is a suspicion that employers regard recruitment of women as the magic bullet to boost numbers in hard-to-recruit areas of HPC? Should we accept the notion of some areas of HPC being regarded as ‘women’s work’ due to perceived differences between men and women? Do ‘hero programmer’ men want to work in mixed-gender teams so that they have someone willing to do the less glamorous but nonetheless vital tasks? Do women take the roles that they are offered because it’s what they want to do, or because they feel disadvantaged, or lacking in confidence to insist on the career paths they really want to pursue? Do we have the wrong kind of men working in HPC for progress in gender equality to be rapid?
The issue of increasing the number of women working in HPC is therefore more complicated than it initially seems. Do we assume that if we increase the number of women working in the industry that women will be given equality of opportunity to progress their careers in whatever directions suit them best? Does the problem lie with women for being willing to settle for a role (any role) in HPC or with conscious or unconscious bias about what women are good at? Does it matter if women in HPC (as in law and medicine and other professions where the number of female participants has risen sharply) are clustered in certain areas of work (typically areas which are less prestigious)?
I think that it does. It is vital in an industry where multi-skilled, interdisciplinary teams are seen as the model of the future that women are not there just to make up the numbers, but are given every support and encouragement in achieving their full potential in whatever roles in HPC teams they feel themselves most suited.
As Karen Spärck Jones, winner of the 2007 BCS Lovelace Medal said: “Computing’s too important to be left to men.” Nowhere is this more true than in HPC.
Alison Kennedy is executive director of the Edinburgh Parallel Computing Centre (EPCC) in Scotland, and a member of the board of directors of PRACE.