SC is arguably the main event of the year for the HPC/supercomputing community. And so it becomes an annual cauldron, relentlessly bubbling to the surface those issues that are most topical for the HPC world. In 2016, two of those issues were women and money.
2016 was the year, I think, that the theme of diversity in the HPC workforce broke forth to be acknowledged as a fundamental issue that affects HPC. Isolated appeals have been trying to raise the diversity failure and opportunity for several years. But it was only in 2016 that the efforts of a persistent few gained momentum, brought focus, and dragged the issue into the center of the HPC community's self-awareness. And, equally importantly, a sufficient proportion of the HPC community was finally willing to see the glaringly obvious and maybe even do something about it.
In essence, the issue is simple: the HPC workforce - from new entrants to leaders - is numerically dominated by white men. Yep, that includes me too. And the leadership roles are dominated by older white men. (I don't think I'm "older" yet - although I guess that is relative.)
Well, we could look at it ruthlessly and pragmatically. It means that HPC is not taking advantage of a broader possible workforce (women, ethnic diversity, etc.) when at the same time many are calling out that there is a shortage of talent available for HPC. (Actually, the claimed shortage of talent is a whole other argument, one for another blog, another day.) So get more women into HPC and this should help reduce the shortage of talent.
It is not just "hands-on-deck" either - it also means that a wider pool of ideas is not being brought into play. Who knows what technical innovations, competitive edges, or commercial advances have been forever dropped into the mists of history simply because they hadn't been given the chance to emerge from the brains of the women who never got the opportunity in HPC? It's not hard to see fundamental business reasons for fixing the diversity gaps.
Or, we could look at it from the point of "right and wrong". I'm not usually a big fan of arguing on the basis of right and wrong - nature/fate/reality/whatever is simply too inconsistent and unfair for us to have a chance of fighting back. But I fail to see good reason why any one person, or any group of people, should be treated badly/differently, merely because of gender, color, sexual preference, accent, religion, or whatever.
In fact, I'd like not to believe some of the stories I hear of treatment of "minorities" in HPC. (How did we end up using "minority" to describe half the population???) Eyes open on twitter, or chat to people at SC - the many stories include assumptions that a woman won't understand tech details on a booth exhibit, or that the woman is the PR person not the CTO, or inappropriate physical behaviour, etc.
And so initiatives like Women-in-HPC (WHPC) are doing excellent work to raise the diversity issue, expose some of the real stories, explore possible reasons behind the situation, help raise the profile of "unconscious bias", and allow "minority" groups to share experiences and help each other.
And I'm sorry, there is a big "but". Or two of them actually.
The first big "but" is, oddly, a result of the momentum of the WHPC effort. I noticed it a little at ISC16, and then saw it hammered into focus at SC16. I ran a day of tutorials at SC16. With diversity being a stated theme at SC16 and something I personally care about, I looked around the room. And noticed there were hardly any women present in my tutorial. Had I failed to make the topic sufficiently attractive to women? While my co-presenter was speaking, I nipped out of the room and ducked into the other tutorials nearby. Same story. Then further down the hall, I saw something that I had hoped to attend if my tutorial hadn't made a clash of timing. The WHPC workshop. Yep, most of the women were in one room talking about being women in the man's world of HPC.
There is huge value in discussing the challenges for women in HPC, and exploring ways to help change the current failures. I don't in any way dispute that. And seeing the agenda, I was quite disappointed I couldn't attend most of the WHPC workshop myself - more white men attending these event is a good step towards raising awareness of the underlying issues and helping to improve the outcomes.
But, the concern is that the success/timing of the WHPC workshop effectively forced women attending SC16 tutorials/workshops to chose between being a woman (go to WHPC workshop) or being an HPC person (go to tutorial or workshop most relevant to their technical interests/career development). It is important to note that this isn't just my observation - this point was also made to me by several other people (mostly women) during conversations at SC (and ISC). Indeed, the comment from some was specifically that they hadn't attended the WHPC events (or similar program items) because they were too busy doing their HPC day job (i.e., having meetings, or attending tech talks or whatever).
In fact, several people made comments to me during SC16 (and ISC16 and other HPC events this year), all roughly of the essence of "I just want to be judged as me, on my abilities, on my track record, on my potential for the future, not on the fact that I am a woman".
This leaves a nasty catch-22: how to find time to focus on the need to improve the realities for women in HPC whilst not taking time away from those women being (primarily?) HPC people. I don't have an answer.
And the second big "but"? I wonder if any readers have noticed yet? For the last several paragraphs, I have only talked about women, as if that is the dominant aspect of diversity that needs attention. I feel that the community conversation mostly end up doing the same - be it media, or events, or workshops, or diversity-corrected panels, or chat-around-the-fireside.
In part, this too is due to to the success of the WHPC effort that has created an easy focus on gender as a narrow part of the wider diversity issue. However, please let's not forget that the lack-of-diversity in our HPC community is real on many axes, not just gender. I know the WHPC team are aware of this and try to raise other diversity axes too, but the narrowing still creeps in.
If you have any suggestions to fix the issues I've raised here, or any context that would help other fix them, then please drop a note into the comments below, or contact me via twitter (@hpcnotes) or directly.