Comment Re:Nevertheless I am impressed (Score 1) 229
That's a bit excessive! It does have some nice things going for it, including a fairly nice API that's been binary and source compatible for decades.
This is no advantage in this case as there are decades old API which are nicer, standard, and as a bonus have open implementation (like MPI).
There's end-to-end Unicode support in all APIs, a nice event logging and tracing system, a nice performance monitoring system (WMI), various asynchronous file and socket APIs, including advanced copy-less APIs that can tie TCP streams to specific CPU cores, etc...
All of which are pretty much useless in HPC, and when they are useful, they're already there in better form on Linux. A nice event logging and tracing system, really? When people want maximum computing efficiency, they don't want all that.
Unlike Linux, Windows has a built-in volume snapshot system that supports application quiescing (not just cache flushing), exportable snapshots, advanced access-list support that is standard and consistent, etc...
All of which are useless in a cluster. If Windows snapshots are exportable to Linux LVM, this is a good thing. The same if Windows ACL support is standard with the POSIX ones, or at least the one in Linux.
I didn't know that, but at the same time who cares?
Really, the biggest issue with Windows is that the source is closed, so if you need something special for a cluster, you're out of luck. "patch tuesday" is only an issue on networks which are not controlled, and a supercomputer would use a dedicated, isolated network.
Which is nonsense because a FS like Lustre is still updated constantly, and often need tweaking to a particular hardware.
And no, the biggest issue with Windows is its efficiency. It's one of the worst ones in the Tsubame 2.0 system at Rmax being 52 % of the Rpeak. In part due to the use of GPU, but still. Even the first on the top 500, that also uses GPU, is at 54 %. Linux can go in efficiency to as high as 87 % and more.