This week sees the 12th edition of the Multicore World conference being held in Christchurch, New Zealand, with an impressive array of speakers joining to discuss a wide range of HPC-related topics.
@nicolas.erdody has done a great job of organizing this conference once again; you can read a Q&A with him about the conference here:
Cendio is sponsoring the event again this year, so I have the pleasure of attending. Day 1 saw some interesting presentations, including from Dan Stanzione (Texas Advanced Computing Center) and Rupak Biswas (NASA):
One thing I learned from Dan’s talk, is that storage costs increased by >50% (will have to check the exact number) in 2024 - the first year ever that IT storage costs have increased, rather than decreased. This trend is largely being driven by the massive investment in AI, and is affecting hardware costs across the board.
If this trend continues then it has all kind of implications, for example with regards to optimisation. The best way to increase performance or capacity may no longer be to just throw more hardware at the problem; it may be to optimise systems to be more efficient instead. This has for a long time been considered the more expensive way forward, but that may be changing.
What’s the most efficient way of providing remote desktops to users? Which method gives you the highest user density per server?
