The two years missing from the archive

When we put together the first conference in 2006, high performance computing in Ukraine consisted of a few clusters scattered across institutes and a few dozen people who knew how to work with them. Most of those people had never met. The whole idea of HPC DAY was to put them in one room for one day.

Over the next eight years the conference grew, from a small gathering to nearly three hundred attendees in 2013. International partners arrived, then vendors, then joint university and industry projects. The motto in 2013 was “Integration”: we brought HPC DAY and the academic HPC-UA conference tightly together and hoped they would eventually merge into one stronger event. In the invitation the following year we wrote that more than two and a half thousand people had passed through the conference over eight years. It looked as though the curve would keep going up.

In the autumn of 2014 we met once more. This time HPC DAY was not a standalone event but a technology day inside the HPC-UA conference, in the hall of the Institute of Cybernetics. Four talks. The announcement said it plainly: resources are smaller this year, but we will still deliver the important technology update, and we expect fifty-odd people, because times are hard. Against 2,750 attendees over the previous years that read like a surrender. It was in fact a decision not to stop.

And then we could not do even that. For two autumns running there was no conference. We never announced a closure — it simply became clear each year that there was neither the money to gather nor a reason to: running an event about buying supercomputers in a country at war made no sense.

The conference returned in 2017, but it was a different event, in a different country, facing a different market. Cloud resources had made some of the 2013 arguments obsolete and research groups had found machine time abroad. In 2018 HPC DAY ran as a track inside a larger conference, and there it ended.

This archive exists so that the two empty years stay visible. It would be easy to list the dates and let the history read as unbroken. It was not unbroken, and the gap explains more about it than any of the successful years.