[Beowulf] interesting article on HPC vs evolution of 'big data' analysis

Prentice Bisbal prentice.bisbal at rutgers.edu
Thu Apr 9 13:40:32 PDT 2015


On 04/09/2015 12:28 AM, Kilian Cavalotti wrote:
> On Wed, Apr 8, 2015 at 11:31 AM, Prentice Bisbal
> <prentice.bisbal at rutgers.edu> wrote:
>> I got annoyed by this article and had to stop reading it. I'll go back later
>> and try to give it a proper critique, but obviously disagree with most of
>> what I've read so far. Right of the bat, the author implies that Big Data =
>> HPC, and I disagree with that.
> It is annoying, indeed! :)
> But I guess it's at least partly because some of its points make
> sense. Like you probably, I don't agree that MPI is killing modern
> HPC, but it's definitely true that the best tool or language for HPC
> is the one that gets the science done.
Actually, it was the first few paragraphs where he compared 'big data' 
tools to MPI. He's comparing apples to oranges. Different horses for 
different courses. I finally read the full article, and will post a 
proper rant/review soon.

The latter part of the article, where he talks about MPI being 
abstracted at the wrong level, and being too low-level, are legitimate 
arguments, Its what first section that annoyed me.

>
> That being said, I feel like the underlying pitch is that MPI is too
> hard if you don't exactly know what you're doing. Well, newsflash, yes
> it is. There are probably better alternatives for non-programmers,
> higher-level languages which allow to focus more on the science than
> on the nitty gritty of passing messages. But it's like saying that C
> is so hard that it's killing innovation and that everybody should use
> Javascript instead: kind of, but not really.
>
> Cheers,



More information about the Beowulf mailing list