Opinion: The Violent Waters of HPC

Print Friendly, PDF & Email

In this opinion piece from The Exascale Report, Chief Editor Mike Bernhardt foresees serious implications from IBM’s cancellation of the Blue Waters contract at NCSA.

Supercomputers should be considered an important strategic asset, and funded as such. If the USA really wants to demonstrate world technological leadership, then adequate resources need to be committed. The failure of Blue Waters should be seen not just as a failure for IBM, but as a failure for the leadership of this country.

The Exascale Report is available only to subscribers, but this article has been made available as an exclusive PDF download for insideHPC readers.


  1. I find no end in irony in that this report faults IBM and NCSA/NSF for the failure
    of BlueWaters while the a story on to this web site reads….

    “IBM Patents Reveal Plans for 100 Petaflop Blue Gene/Q Super”

    If I am a major customer, do I care, can I trust IBM with a major HPC installation?
    They woefully took the BlueWaters project bid it and didn’t understand the full
    ramifications to what they had bid for the money offered. Meantime your telling
    us how they are going to build 100 PF systems>?

    On the customer side, NCSA obviously didn’t understand the limits of what they
    had, budget wise, vs. what they were asking for.

    If I was in charge of this mess; NCSA would have their funding pulled back,
    someone else could use the money more wisely, and IBM wouldn’t get any
    Government HPC contracts for at least 12 calendar months. There should be
    a price for failure at this level, but I am not seeing either NCSA or IBM paying
    that price, why is that? So contrary to what is stated in the report I don’t see
    any implications being imposed on either NCSA or IBM, that is the real
    story here.

  2. I very much agree with your “price for failure” comment. When you say “you’re tell us how ..” — I assume you are referring to the publication here – insideHPC. But insideHPC is just reporting the news – that a bunch of other sites have picked up as well. The publication is just the messenger. But, I do agree also with your comments about the irony of this situation – even though one system was based on the Power 7 and the other being talked about is based on Blue Gene/Q – two very different architectures – it still seems utterly ridiculous that IBM can cancel one program to hit 20 PFLOPS – and in almost the same breath – announce how they are going to build one to reach 100 PFLOPS.

  3. “you’re tell us how ..” was supposed to be “you’re telling us how…”

  4. The highest levels at NSF signed off on IBM’s proposal, so it wasn’t just NCSA. The cost to develop the platform was already paid for by DARPA and the hardware was shipping, so any “research” was already sunk costs. IBM just decided at the last minute not to deliver.