Cisco Blogs


Cisco Blog > High Performance Computing Networking

Platform Acquires HP-MPI

August 24, 2009
at 12:00 pm PST

In a move that will surely cause some head-scratching, Platform has acquired the intellectual property of the-MPI-previously-known-as-HP-MPI.The head scratching part is that Platform already owns Scali MPI. It’s no secret that they recently moved all Scali development to an engineering team based in China.Both Scali MPI and HP-MPI were (are!) strong individual products. Indeed, HP-MPI had deep/excellent coverage of ISV’s who sell MPI-based software (e.g., oil and gas simulation software packages, manufacturing applications, etc.). The PR directly states that this was a key point in the acquisition. Here’s the interesting bit:

Platform’s plan is to combine the superior performance of its existing MPI technology with HP-MPI’s broad ISV acceptance to create a single, comprehensive and powerful product named Platform MPI.

Since I have no “insider” information on this acquisition, I feel free posit a few speculations:

  • The press release could be read to mean that Platform intends to combine the HP-MPI and Scali MPI code bases — perhaps taking the best features of both…? Having done just about exactly that to create the Open MPI product, I know exactly how difficult this is. In our case, we ended up starting an entire new code base and cherry-picking the best features from several different prior MPI implementations to port to our new code base. I’m guessing that such an approach would not be attractive to Platform (or anyone) because it took a solid two years before Open MPI had a usable product. I would guess/assume that a similar timeline would be incurred by others undertaking similar endeavors.
  • Alternatively, Platform could have bought HP-MPI in order to reduce competition and have a single commercial MPI implementation that is popular with ISV’s. That, too, may be a bit of a challenge; the ISV’s I’ve talked with have been fairly happy with HP-MPI. Rebranding the Scali code base as “that-which-used-to-be-HP-MPI” might be difficult — even if some of HP-MPI’s more-attractive features are ported over to the Scali code base. Convincing ISV’s to effectively change to a new underlying MPI may be a difficult proposition. ISV’s like stability. Change is hard — it requires potentially large amounts of QA resources (and therefore, money).

I don’t really have a crystal ball to know what will happen, but now I know the MPI community will have something to speculate about until the American college football BCS later this fall.The only thing that is certain is that once the dust settles on this acquisition, the field of commercial Linux MPI implementations will have effectively been reduced to just two companies: Platform and Intel.

Comments Are Closed